While traditional brown energy remains a significant energy source for AI data centers, clean and renewable energy continues to grow in capacity.
Goldman Sachs believes 40% of new data center capacity will be from renewables.
Hyperscalers and data centers are adopting a mixed energy portfolio, combining brown, clean, and renewable energy to balance emissions while ensuring 24/7 reliability.
As the artificial intelligence (AI) revolution drives the growth of AI data centers, the topic of energy continues to gain prominence. AI data centers cannot function without energy sources, and how that power is generated seems just as important as how reliable the power is. Energy sources are commonly labeled brown, clean or renewable. Goldman Sachs says AI data center power consumption demand is expected to grow by more than 160% by 2030 from 2023 levels. Let’s take a look at each of these sources and how efficient and reliable they can be for AI data centers.
What is Brown Energy?
For decades, the world has relied heavily on brown energy from fossil fuels like oil, coal and natural gas. These dominant fuels powered the Industrial Revolution and remain the primary sources of electricity for the grid today, with coal and natural gas generating nearly 60%.
Coal-fired power plants are some of the worst offenders as they release enormous amounts of carbon dioxide and greenhouse gases into the atmosphere. Thermal efficiency (TE) measures how effectively a fuel’s heat is converted into electricity. Coal has some of the lowest TE efficiency, around 33%.
Natural gas is the top fuel source for powering the electric grid at 43% and is cleaner than coal but still contributes carbon emissions when it's burned. TE is between 35% to 42% on a simple cycle gas turbine and up to 62% when using combined cycle gas turbines (CCGT), which use the exhaust heat to boil water in a steam turbine, adding the extra 20% to 25% TE.
Join thousands of investors who trust I/O Fund’s expert stock analysis on AI, semiconductors, cryptocurrency, and adtech — sign up for free! Click here!Join thousands of investors who trust I/O Fund’s expert stock analysis on AI, semiconductors, cryptocurrency, and adtech — sign up for free!Click here!
The environmental impact of burning fossil fuels has paved the way for the clean energy and decarbonization movement. The combustion process produces carbon emissions, which contribute to air pollution and climate change. As AI data centers consume massive amounts of electricity, they’re also trying to meet clean energy initiatives.
What is Clean Energy?
Clean energy has much less environmental impact, producing low to zero greenhouse gas emissions. It also generates less pollution and leaves a smaller carbon footprint. Some examples of clean energy are:
Solar power generates electricity using sunlight and photovoltaic panels with TE between 15% to 22%.
Wind power generates electricity by harnessing kinetic energy from the wind with turbines with mechanical efficiency (ME) between 35% to 50%; since they don’t operate on thermal cycles, there isn’t heat conversion.
Hydropower generates electricity from water flowing through turbines, again no thermal cycle but the ME is 85% to 95% since the momentum of the water converts to power with almost no loss.
Geothermal power uses the heat emanating from the earth’s core with a low TE ranging from 10% to 23%
Nuclear energy generates power through fission, heating water to create steam that drives turbines, with TE averaging between 33% to 37%.
For AI data centers, the most practical clean energy sources come from nuclear, solar and wind power. While the efficiency of hydropower is exceptionally high, it isn’t practical for data centers due to factors like heavy capex to build dams and reservoirs, environmental impacts and geographical limitations. Clean energy enables data centers to lower their carbon footprint and enhance their environmental reputation.
What is Renewable Energy?
Renewable energy comes from natural processes that are replenished at a faster rate than consumed, such as solar, wind, and hydropower. While renewable energy is typically clean, meaning it generates low carbon emissions, not all clean energy sources are renewable. For example, nuclear power is considered clean but not renewable.
While solar and wind power are renewable and clean energy, they aren’t available 24/7. They would require a battery (storage) system to match the 24/7 reliability of nuclear and natural gas. As for supply costs, renewable energy sources are actually cheaper than generating electricity from natural gas. According to a report by Goldman Sachs, solar energy costs $25 per megawatt-hour (MWh) compared to CCGT natural gas at $37/MWh. But there's a reason natural gas costs more: reliability.
“In practice, though, utility-scale solar plants only run around 6 hours per day on average, while wind plants run for an average of 9 hours per day. There is also day-to-day volatility in the capacity of these sources, depending on the radiance of the sun and the strength of the wind.” Solar is not effective during cloudy and overcast days, whereas nuclear and natural gas plants can run around the clock every day regardless of weather. Goldman Sachs believes that 40% of the new capacity built to support data center power demands will be renewables.
The Mixed Energy Sources Portfolio Approach
Many hyperscalers and data centers have adopted a mixed energy portfolio approach utilizing brown energy and clean or renewable energy to balance the emissions and maintain a green stance.
Goldman Sachs Infrastructure analyst Jim Schneider commented, “Our conversations with renewable developers indicate that wind and solar could serve roughly 80% of a data center's power demand if paired with storage, but some sort of baseload generation is needed to meet the 24/7 demand.” While the baseload power preference is nuclear, building them is just too difficult, which makes natural gas and renewables the most practical solution short-term.
The I/O Fund recently entered five new small and mid-cap positions that we believe will be beneficiaries of this AI spending war. We discuss entries, exits, and what to expect from the broad market every Thursday at 4:30 p.m. in our 1-hour webinar. For a limited time, get $20 off a Monthly Pro plan with code PRO20OFFget $20 off a Monthly Pro plan with code PRO20OFF[Learn more here.]Learn more here.]
Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.
Core Scientific is a major Bitcoin miner leading the transition to high-performance compute (HPC) data centers with 1.3 GW of contracted powered infrastructure.
The Company signed 12-year hosting deals with AI Hyperscaler CoreWeave to provide 590 MW of HPC infrastructure valued at up to $10.2 billion.
CoreWeave will front the $750+ million in capex funds to modify Core Scientific’s data centers as an anchor client.
Core Scientific’s HPC hosting revenues could surge 21X per quarter when the full 590 MW of critical load go online in 2027.
ROTH MKM expects Core Scientific to reach a $1.58 billion run rate by 2027.
The I/O Fund will be holding a very tight stop on any position we initiate, and the stock would be for Advanced Market Signals only, indicating it qualifies for more advanced investors who are comfortable trading daily/weekly.
Core Scientific (NASDAQ: CORZ) is a digital infrastructure company that operates bitcoin mining and hosting services, and high-performance compute (HPC) hosting services through its nine purpose-built data centers. As one of the largest Bitcoin miners in North America, operating 171,000 mining rigs (164,000 owned), the Company is positioning itself for significant growth in the AI space.
Its strategic shift to high-performance computing (HPC) hosting is particularly compelling, allowing it to mitigate Bitcoin’s volatility while capitalizing on the surging demand for AI data centers. By securing high-margin HPC hosting contracts, the company is poised to tap into one of the most lucrative and rapidly growing markets in technology. While other Bitcoin miners are starting to catch on, attempting to transition to AI data centers, Core Scientific has a clear first mover advantage reinforced by lucrative multi-billion dollar 12-year contracts with upside revenue potential of $10.2 billion with AI hyperscaler CoreWeave with a trajectory aimed at generating 21X HPC hosting revenue growth in 2027.
Core Scientific’s Value Proposition for HPC Hosting Customers
Core Scientific provides many attractive value propositions to hyperscalers:
Specialized Power Infrastructure: HPC customers require more power than conventional data centers can offer. AI and HPC workloads require 6 to 10 times more electricity to operate. Traditional data centers are accustomed to single racks consuming 10 to 15 kW of power. Current AI racks push 80 KW as they will soon draw 120 kW to 150 kW in the next generations, with 200 kW within several years. Core Scientific has the existing infrastructure with nearly 900 MW of capacity for HPC hosting in addition to the 400 MW for Bitcoin mining capacity. , with 200 kW within several years. Core Scientific has the existing infrastructure with nearly 900 MW of capacity for HPC hosting in addition to the 400 MW for Bitcoin mining capacity.
Scalable HPC Infrastructure: They are actively transitioning their facilities to cater specifically to AI workloads with infrastructure optimized for machine learning and deep learning applications. Its Denton, Texas, facility is undergoing a $6.1 billion expansion, boosting its MW capacity by 97 MW to 394 MW with 47 more acres to 78 acres to host one of North America's largest GPU supercomputers for AI computing. It's being converted entirely for HPC hosting.
Location: Core Scientific operates 9 Application Specific Data Centers strategically located near major internet hubs in Georgia, Kentucky, North Dakota, North Carolina, Oklahoma, Alabama and three in Texas. Its new leased (with an option to buy) site in Alabama offers 11 MW of critical IT load and is scalable up to 66 MW of critical IT load, which Core Scientific is in discussions with potential new clients to contract for HPC hosting. They are developing a state-of-the-art 100 MW data center in Muskogee, Oklahoma.
Maintenance and Repair: Offering 24/7 around-the-clock monitoring, support and maintenance, Core Scientific is one of the largest application-specific integrated chips (ASIC) repair centers in North America, servicing their own and customer’s fleet of 171,000 bitcoin mining rigs. Parlaying from ASICs, they plan on using their expertise and manpower to replicate it and expand their offering on the GPU side.
CoreWeave: An Early Believer in the Core Logic’s HPC Transition
CoreWeave is an NVIDIA-backed AI Hyperscaler, providing AI cloud services by offering GPU clusters for HPC and AI workloads. CoreWeave is a specialized cloud provider offering an optimized platform for GPU-intensive tasks. They build and operate their own data centers equipped with a massive scale of over 300,000 Nvidia GPUs. They currently have 28 operational data centers and plan to open 10 new data centers in 2025, leasing a significant portion of their capacity with Core Scientific.
On March 6, 2024, Core Scientific announced an initial long-term deal with anchor customer CoreWeave to provide up to 16 MW of data center infrastructure for their HPC and AI workloads at their tier 3 data center in Austin, Texas. This helps to substantiate the pivot from Bitcoin mining to offering AI/HPC infrastructure, stating a “strategy shift” to AI may be good for a temporary stock price spike, but actually signing up customers is another story.
CoreWeave was already a GPU hosting client from 2019 to 2022, hosting thousands of GPUs. Despite the potential value of the March 6 deal being worth up to $100 million, it didn’t move the needle much for the stock price, which still traded under $4.00, selling off to $2.95 the following week.
CoreWeave Ups the Ante and Fronts the Capex Funds in $3.5 Billion Deal
Core Scientific was successful in delivering 16 MW of capacity more than 30 days ahead of schedule at its Austin, Texas, data center. This prompted more deals. On June 3, 2024, CoreWeave signed several 12-year contract deals securing 200 MW of infrastructure to host CoreWeave’s NVDA GPUs. Additionally, CoreWeave will fully fund (not pay for) the capital investments (capex), estimated around $300 million, required to modify Core Scientific’s “existing infrastructure into cutting-edge application-specific data centers customized for dense HPC." CoreWeave will put up the capital for the modifications and Core Scientific will credit them 50% of their hosting fees until it’s paid back fully.
Regarding CoreWeave paying for the capex, here is what was stated on the call:
“Yes. Thanks, Brett. I mean really, the difference in the CoreWeave deal is 100% funding of the CapEx. They were able to significantly buy down their rates. And I think as we look forward, what we're seeing for 2025 is frankly rather unique. If you're able to deliver capacity in 2025 and 2026 right now — we're definitely seeing those lease rates be much higher than we expected, especially given that many of these folks are willing to cover some portion of the CapEx of the build-out. So we're excited about where lease rates are going, and we believe we'll be able to extract a significant amount of value from the demand that we're currently seeing over the next few years.”
CoreWeave Contracts a Total of 502 MW Generating $8.7 Billion Over 12 Years
Once the 200 MW of HPC is operational Core Scientific estimates they’ll receive around $290 million annually or more than $3.5 billion during the initial 12-years terms of the contracts. CoreWeave exercised its options and signed for an additional 70 MW on June 25, 2024, and $105 million of capex funding, equating to an additional $1.23 billion for Core Scientific during its 12-year term. In August 2024, CoreWeave signed another 112 MW contract beginning in 2026.
In October 2024, CoreWeave exercised the rest of its options and signed another 120 MW hosting contract for 12 years for a total of a full 502 MW of critical IT load with a total revenue potential of $8.7 billion over the 12-year terms of its contracts. The average annual run rate is $725 million. HPC hosting revenues are expected to start flowing in 2025 with 200 MW delivered by the end of 1H 2025, up to 270 MW delivered by the end of 2H 2025, up to 382 MW by the end of 1H 2026 and up to 500 MW by the end of 2H 2026. CoreWeave is expected to fund capex costs of $750 million, which Core Scientific will credit 50% of the hosting fees until fully repaid.
CoreWeave: Providing 250,000+ NVIDIA GPU-powered AI Infrastructure For Lease
As Core Scientific’s largest anchor customer, it’s important to take a look into this client. Core Scientific is a direct benefactor of CoreWeave’s success. What’s good for CoreWeave is also good for Core Scientific.
CoreWeave is an AI hyperscaler that has evolved from a crypto miner that leased space and power (NVIDIA GPUs) from Core Scientific to an AI hyperscaler powerhouse with a roster of high profile clients including Microsoft, Meta Platforms, IBM, Cohere, NVIDIA and OpenAI. CoreWeave is a specialized cloud provider focused on offering scalable AI cloud infrastructure including access to over 250,000 NVIDIA GPUs, low-latency networking and high-bandwidth storage optimized for the massive computational workloads of AI training and inference and ML. The company stands out, as NVIDIA stated, “… CoreWeave has launched NVIDIA GB200 NVL72-based instances, becoming the first cloud service provider to make the NVIDIA Blackwell platform generally available.”
CoreWeave builds infrastructure that can scale at a moment’s notice that can go from zero GPUS to 10,000 GPU working on the same job within a minute.
In 2024, Microsoft accounted for nearly 62% of CoreWeave’s revenue (with Meta accounting for 15% according to H.C. Wainwright), which surged 737% YoY from $229 million to $1.9 billion. Customer concentration concerns were eased a bit with the signing of a $11.9 billion deal with OpenAI, who will also become an investor owning $350 million of stock. NVIDIA holds a 5% minority stake in CoreWeave, which will be going public in 2025.
Core Scientific Gets a Game Changer Deal with CoreWeave
The revenue potential of CoreWeave’s contracts is $8.7 billion over their 12-year terms, equivalent to $725 million annually once fully online. That equates to $181.25 million of quarterly HPC revenue, up from $8.5 million in Q4 or 21X potential, by early 2027. When compared to overall revenue, this is 2X growth on a quarterly basis.
The 21X growth in HPC and 2X growth in overall revenue is before the additional 70 MW $1.2 billion expansion deal announced at its Denton, Texas facility in its Q4 earnings release, resulting in the cumulative revenue potential of over $10 billion from CoreWeave, with 75 to 80% cash gross profit margins according to the Company. The full 590MW contracted to CoreWeave is expected to come online in 2027. This would be up from 250 MW expected in 2025.
CoreWeave Announces $1.2 Billion Expansion at Denton, Texas Facility
On Feb 26, 2025, coinciding with its Q4 2024 earnings release, Core Scientific announced a $1.2 billion 70 MW expansion at the Denton, TX site, for CoreWeave. The 70 MW of additional contracted power at the Denton site increases the full critical IT load to approximately 260 MW. The agreement increases CoreWeave's total contracted HPC infrastructure with Core Scientific to approximately 590 MW across six sites.
Under the terms of our Agreement with CoreWeave with respect to this additional 70MW, Core Scientific is responsible for funding $104 million of the additional required capex ($1.5M per MW), with CoreWeave responsible for the additional capex associated with the expansion. The company also retains the option for two additional five-year renewal terms.
“Looking ahead, we now expect to have delivered approximately 250 megawatts of HPC capacity to CoreWeave by the end of this year, with the full 590 megawatts coming online in early 2027. This represents a shift from our previous timeline and reflects both the size and complexity of the project, particularly the addition of an incremental 70 megawatts of critical IT load.“
Adam Sullivan also added this, “So from what we're seeing on CoreWeave's demand side is significantly stronger than what we saw in 2024. There's a lot of things going on in the market today that we're seeing that's actually driving continued demand and flow into CoreWeave. And so we're excited about continuing to expand with them at Denton. And Denton is going to be one of the largest supercomputers in the United States, and it's going to be a flagship asset for CoreWeave.“
One note of caution: Core Scientific has all the makings of a hypergrowth stock and this includes immense risk. The company is recently out of Chapter 11 Bankruptcy and has to raise cash to fund operations, which means taking on debt. The I/O Fund will be holding a very tight stop on any position we initiate, and the stock would be for Advanced Market Signals only, indicating it qualifies for more advanced investors who are comfortable trading daily/weekly.
Looking Beyond the Q4 2024 Headline Numbers
Core Scientific reported disastrous-looking Q4 2024 earnings results based on headline numbers, with an EPS loss of ($0.60), missing consensus estimates for a loss of ($0.09) by ($0.51). Revenues fell 33.1% YoY to $94.93 million, missing consensus estimates by ($2.14 million). Yet, the stock gapped over 10% following its release.
The reason is that just beneath the surface, Core Scientific is setting up to solve one of the biggest issues the United States and the AI market face: power supply. The company is going through a transition period as it moves into the AI/HPC data center markets supported by AI hyperscaler CoreWeave, who just signed a five-year $11.9 billion deal with OpenAI.
$224.7 Million Mark-to-Market Adjustment Shouldn’t Spook Investors
The initial sting of the reported ($265.5 million) GAAP loss in Q4 2024 may sound like bad news, but $224.7 million of it is a non-cash mark-to-market adjustment on warrants; just accounting noise. The “actual” Q4 net loss was ($31.8 million), not ($265.5 million).
Core Scientific issued warrants, which are considered liabilities under GAAP accounting rules since the Company has to deliver stock at the exercise price. If the stock rises in value, the Company has to post a larger liability, but it doesn’t mean they are taking any actual losses. The Company issued two tranches of warrants at $6.81 x 98.3M shares for Tranche 1 (CORZW) exp. January 23, 2027, and $0.01 x 81.9M shares for Tranche 2 (CORZZ) exp. January 23, 2029, as part of its plan to emerge from Chapter 11 bankruptcy in January 2024.
The Company still receives the funds when the warrants convert as they issue the required shares. There are no real losses. In fact, it’s relatively good news since the higher the stock price rises, the deeper the “paper losses” appear until the warrants are all exercised or expired and taken off the books. However, that presents a dilution issue of an additional 180.2 million additional common shares.
The 116 Million Shares from Warrants Remaining Might Spook Investors
During 2024, Core Scientific received $4.4 million in proceeds from 646,109 shares of Tranche 1 warrants being exercised. On December 24, 2024, 60.9 million Tranche 2 warrants were exercised for $600,000. This leaves 116 million warrant-related shares remaining of potential dilution on the remaining warrants. Core Scientific has 294 million shares outstanding as of February 20, 2025.
As of February 20, 2025, the pro forma diluted share count is 501 million shares. This includes the current 294 million shares outstanding along with 207 million additional unissued shares that include Tranche 1 and Tranche 2 warrants of 116 million remaining, convertible notes of 70 million shares and 21 million shares of restricted stock and reserve shares.
Revenues Sink as Company Converts Bitcoin Data Centers to HPC Data Centers
The company’s Q4 revenue fell by (33.1%) YoY and (0.44%) QoQ to $94.93 million, primarily due to the decline in self-mined Bitcoin to 974, down from 3,042 in the year ago period. The Company has been converting some of its Bitcoin mining data centers to HPC data centers and actively “sunsetting” Bitcoin hosting contracts as it transitions to HPC hosting. The Bitcoin halving event also occurred in April of Q2 2024, thereby causing Bitcoin revenue to shrink on a YoY basis further magnifying the deceleration. Revenue fell short of estimates by (2.2%).
Analyst expect revenue to fall (48.26%) YoY to $92.77 million in Q1 2025, and fall (30.03%) YoY to $98.73 million in Q2 2025.
Full-year 2024 revenues rose 1.6% to $510.7 million.
Analysts expect FY2025 revenue to fall (3.71%) YoY to $491.75 million.
Revenue Segments: Bitcoin Revenues Drop in Preparation for HPC Revenue Acceleration
As Core Scientific transitions from Bitcoin self-mining and hosting to HPC hosting, the revenue segments can be expected to drop in the Bitcoin segments and rise in the HPC hosting segment. The quarters may look predominantly worse until the Core Scientific HPC revenues start to ramp up as they go online. Based on analyst estimates, Q1 2025 may be the final “kitchen sink” quarter before revenues reaccelerate.
Margins Consistently Expand Through 2024
Q4 gross margin was 5%, compared to 27.7% in the same period last year.
Q4 operating margin was (41.9%), compared to 2.8% in the same period last year. However, EPS is showing a rebound on the horizon as higher margin HPC hosting revenues increase.
GAAP EPS Trending Towards Positive After Mark-to-Market Adjustments on Warrants
Q4 GAAP EPS was ($0.60) compared to ($0.11) in the same period last year. The EPS miss was primarily due to the $244.7 million non-cash market-to-market (MTM) adjusted on the warrants.
Analysts expect GAAP Q1 2025 EPS to improve to ($0.10).
Analysts expect GAAP Q2 2025 EPS to improve to ($0.07).
Analysts expect GAAP Q3 2025 EPS to improve to ($0.05).
Analysts expect GAAP Q4 2025 EPS to improve to $0.01 as CoreWeave's data centers come online.
Full year 2024 GAAP EPS was ($4.39) compared to ($0.65) last year.
Analysts expect full year 2025 GAAP EPS to improve to ($0.24).
Analysts expect full year 2026 GAAP EPS to improve to $0.40 as more of CoreWeave’s data centers come online.
Cash Grows as Core Scientific Issues $1.09 Billion in Convertible Senior Notes
Core Scientific closed Q4 with $836.2 million in cash and $1.09 billion in debt. The debt is comprised of two convertible notes. In August 2024, The Company issued $460 million in convertible notes due 2029, which enabled the Company to refinance its debt from a 12% interest rate to 3% while increasing its cash position and removing covenants to allow the Company to accumulate Bitcoin. The conversion price is $11.00 at a rate of 90.9256 shares per $1,000 in principal, which brings a total 41.82 million shares issued upon conversion.
In December 2024, Core Scientific priced an upsized $625 million convertible senior notes offering due 2031. The conversion price is $22.49 at a rate of 44.4587 shares per $1,000 in principal. This brings a total of 71.61 million additional common shares upon full conversion.
The Implications of Not Being Investment Grade
Its worth noting that there are implications of not being investment grade especially when needing to raise cash. Considering Core Scientific emerged from Chapter 11 bankruptcy in January 2024, this status alone shapes their cash-raising strategy. Being non-investment grade tends to mean higher borrowing costs, but Core Scientific was able to cut their interest rate from 12% to 3% by swapping out the debt with convertible notes.
It’s worth noting that Core Scientific’s 3% interest rate is impressive for a company just out of bankruptcy implying the institution(s) are very confident in Core Scientific’s strategy. However, that route also comes with its potential share of dilution (41.82 million new shares) if shares are converted at $11.00 per share. Core Scientific has the option to redeem early if the stock trades 130% above the conversion price for 20-30 trading days ($14.30) after the initial non-call period August 2027.
The additional $625 million convertible also comes with dilution (27.79 million new shares) but at a higher conversion price of $22.49 and no interest rate. However, Core Scientific achieved this funding with 0% interest implying very high confidence that Core Scientific will either redeem the notes at maturity or that the shares will surge above the conversion price enabling them to convert shares for a profit before then. Both convertibles are senior unsecured obligations, therefore in the event of bankruptcy or default, unsecured creditors rank below secured lenders.
Valuation
The Company trades at a forward P/E ratio of 12.27. The trailing twelve month (TTM) P/S ratio is 4.34 and forward P/S is 12.27. The five-year average P/S ratio is 5.08. The P/S ratio peaked at 9.3 in November 2024.
Management highlighted their strategic pivot from Bitcoin mining to HPC hosting. The Company delivered 500MW of capacity through 12-year agreements worth $8.7 billion, expanding its HPC infrastructure to over 1.3GW of contracted power. Its key initiatives include accelerating capacity expansion and targeting significant new site acquisitions, including projects in Auburn, Alabama, and Denton, Texas.
In Q4, the Company secured approval to expand its gross capacity at its site in Denton, Texas, by nearly 100MW, which equates to nearly 70MW of critical IT load. Denton is on track to host one of the largest GPU supercomputers in North America. The Auburn, Alabama, site currently has 11MW of critical IT load, and the Company is actively working with Alabama Power to secure a much larger power agreement. They are deferring significant capital deployment until they finalize negotiations with prospective customers.
CEO Adam Sullivan said this.
“Today, we announced a significant expansion of our relationship with CoreWeave at our Denton facility, which will bring that site to full capacity. This new agreement adds approximately 70 megawatts of critical IT load and represents approximately $1.2 billion in additional contracted revenue over a 12-year term. With this latest expansion, our total contracted value with CoreWeave now exceeds $10 billion, an amount that includes our Austin, Texas agreement, and covers roughly 590 megawatts of critical IT load once fully online. Of that total, just over 570 megawatts reflect the capacity we're converting at existing sites to HPC, where we expect 75% to 80% cash gross profit margins.”
However, Core Scientific will put up the CapEx to receive full HPC rental payments rather than 50% payments, with the other 50% being a CapEx credit for the upfront CapEx spent by CoreWeave. Sullivan said this.
“Under this newest agreement for the additional 70 megawatts, we will fund $1.5 million in capital expenditures per megawatt, whereas in prior agreements, CoreWeave covered those costs. In return, we will benefit from full rental payments during the first two years of the contract because there will be no CapEx credit associated with this new agreement.”
It’s worth noting that analysts may be considering 2027 full delivery as too ambitious considering as evidenced by the lowering of revisions. There are execution risks that may be out of their hands including grid delays and securing power with utilities (IE: working with Alabama Power to secure more power to the Auburn site), funding capex or negative developments with CoreWeave.
Prioritizing Customer Diversification and CoreWeave Timeline to Come Online Fully
Diversifying its customer base is a key priority as it aims to reduce CoreWeave's share of revenues to under 50% of critical IT load by 2028. The Company is in active discussions and remains confident in its ability to diversify its HPC customer base. The Company exited the year with 15MW of critical IT load. New block ASIC chips are expected in 2H 2025, which will refresh some of its Bitcoin mining fleet. Otherwise, there are no plans for any further CapEx spending in 2025 for its Bitcoin mining business.
CEO Adam Sullivan reiterated their top priority of diversifying new customers.
“We are in active discussions with dozens of new customers, including the vast majority of hyperscale providers in several large enterprise companies. Demand remains strong, but we're seeing considerably more due diligence compared to the first half of 2024. This heightened scrutiny reflects the influx of new market entrants who make ambitious capacity promises yet lack the tangible power agreements to back them up, much like the recent situation where a hyperscaler canceled contracts with companies that overstated their available power.”
The Company expects to have nearly 250MW of HPC capacity to CoreWeave delivered by the end of 2025. The full 590MW is coming online in early 2027. Core Scientific believes they can add another 300MW of capacity across existing sites by the end of 2027.
Earnings Call Q&A:
The Goal of Reducing CoreWeave’s Concentration of Revenue under 50%
Core Scientific is actively trying to diversify their concentration of revenue from CoreWeave.
Jeffries analyst John Peterson:
“Okay. And then I appreciate the goal of wanting to bring CoreWeave down to less than 50% of revenue by the end of 2028. I think that would require you to procure a lot more power this year in addition to signing on additional customers. So maybe just talk through the milestones that you need to hit throughout this year to be on track to do that.”
Adam Sullivan:
“We talk about the ability to continue to expand at existing sites. And that's a competitive process because we are getting direction in terms of how much additional power we're going to be able to achieve at some of our existing sites and then some of our new sites as well. Very attractive locations. Our focus today is on building blue-chip assets. And we want to have those blue-chip assets with blue-chip clients. And so that's where our focus is today. And we're going to continue to execute and acquire more sites to bring more capacity online to secure more contracts and achieve our goal of getting them below 50% by 2028.”
At 75% to 80% margins, 590 MW would yield $637.5 to $680 million, with a midpoint of $658.75 million after 2027 (assuming all 590 MW comes online). If CoreWeave is 50% of critical load by 2028, total HPC capacity needs to double to 1,180 MW from producing more power and acquiring more sites. It would require Core Scientific to assume more hyperscalers sign under similar 12-year terms to CoreWeave. Sullivan stated how diversifying its customer base was a top priority, “Starting with diversifying our customer base, this is the top priority for the company this year, and the goal is to sign enough contracts so that CoreWeave represents less than 50% of critical IT load by the end of 2028.” Sullivan mentioned 700 MW was available.
Nick Giles:
“So, appreciate your target that CoreWeave represents less than 50% of critical IT, but that implies that you sign at least the same amount with other customers, but you do have 700 megawatts that you've outlined between existing and new sites by 2027. So, should we assume that the delta would be new customers as well, or could that kind of 130 be split between a new customer and maybe one more tranche with CoreWeave?”
Adam Sullivan:
“Yes. So, we've outlined the 300 and the 400 number that's critical IT load megawatts, so about 700 megawatts. As we look forward if we have 590 of CoreWeave contracts the 700 available to us is really where our focus is going to be on executing new clients. So that's part of our goal to get them below 50%, to have enough capacity available and saleable for us to be able to bring them down to that level.”
How the Deep Seek News Only Made People Want to Move Faster
The Deep Seek news was a head fake as actual demand increased, and it only made people want to move faster.
Adam Sullivan:
“We've seen much more specific requests around locations in terms of developments and where they would like to build. But overall, the Deep Seek news for hitting the public markets rather hard from everything that we've seen on the actual demand side, demand continues to increase, and those conversations continue to progress very well.”
Diversifying the Customer Base Beyond Hyperscalers
While Core Scientific makes headlines when deals are made with name-brand hyperscalers, enterprise customers could also fill in pieces of the void to improve diversification.
Greg Lewis:
“Could you talk a little bit about you mentioned enterprise customers potentially. It's something that seems to be we're hearing more about beyond just the hyperscalers. As maybe you broaden out the customer base beyond just the hyperscalers, which it seems that latency is a big issue for them. Maybe scalability is a big issue for them. As you kind of look at potential enterprise customers, does that open up sites maybe in your portfolio and elsewhere that maybe under hyperscaler footprint wouldn’t work but through enterprise it might?”
Adam Sullivan:
“And so, we're looking at having hyperscale at the very least as anchor, potentially a single tenant. And if they're serving as an anchor, being able to fill out the rest of the capacity with enterprise clients as well. So, the demand, does it open up more sites with enterprise? Absolutely. But we're focused on blue chip assets with blue chip clients, which includes both of those groups.”
Delivery Times and Securing Power Agreements is a Competitive Advantage
Sullivan pointed out that many while demand remains strong, they are seeing considerably more due diligence compared to the first half of 2024 due to the influx of new market entrants that make “ambitious capacity promises” but actually lack the “tangible power agreements to back them up.” Sullivan referenced what may have been the rumored Microsoft cancellation of commitments with CoreWeave due to “delivery issues and missed deadlines.” Microsoft outright denied the cancellations.
Adam Sullivan:
“This heightened scrutiny reflects the influx of new market entrants who make ambitious capacity promises yet lack the tangible power agreements to back them up, much like the recent situation where a hyperscaler canceled contracts with companies that overstated their available power. Our proven track record and secured power agreements set us apart in this environment, and we won't be expanding our footprint unless we have a high degree of confidence in our ability to deliver for additional customers.”
When pressed about the rumor of Microsoft cancelling capacity with CoreWeave, Sullivan responded.
“I can't comment specifically on any relationship between CoreWeave and Microsoft other than what they've spoken about publicly. But, I mean, CoreWeave's continuing to expand. You're seeing it not only with Core Scientific, but really across the globe and internationally. So, from what we're seeing on CoreWeave's demand side is significantly stronger than what we saw in 2024. There's a lot of things going on in the market today that we're seeing that's actually driving continued demand and flow into CoreWeave. And so, we're excited about continuing to expand with them at Denton.”
Is Core Scientific in Discussions with Other Hyperscalers?
Needham analyst John Todaro inquired about discussions with other hyperscalers and CoreWeave. Sullivan noted they are in talks with a majority of hyperscalers in conversations with large enterprises. The customer conversations are continuing to evolve throughout the early part of 2025. Sullivan was asked if he saw any demand changes across inference and training workloads on the back of Deep Seek headlines.
Adam Sullivan:
“Denton was a site that we were really slating for CoreWeave. We did have conversations with some other hyperscalers and other clients on those megawatts. As we talked about the 300 megawatts potential at other existing sites, we're in conversations today with other potential customers around that. There's really no guarantee that anything like that would go to CoreWeave, because what we do want to do now is really focus on continuing to diversify our client base, and our existing sites are great campuses for us to do that.”
Elaboration on the Delays
Sullivan mentioned there were some delays from changing some of the designs to fit for the equipment further impacted by the constrained supply chains going out into 2026.
Adam Sullivan:
“And one of the things that we wanted to ensure that we achieved was that we had the right equipment on the right schedules for the site plans that we had. And so, that required us to change some of the designs to fit for the equipment that was available to deliver on the timelines that we set forward. And so, there was just some incremental delays there. But overall, we have high confidence in where the delivery schedules that we've put forward today. And we believe we're going to be able to hit those timelines.”
Management now expects critical IT load to be 250 MW, including the 16.5 MW, down from 270 MW plus 16.5 MW.
Brett Knoblauch:
“Thanks, guys. Really appreciate it. Maybe just quickly on the delays, if you will, or the pushback in timing. Just want to make sure I heard you right. You're now expecting critical IT load this year to be 250 megawatts. Does that include the 16.5? And before, you guys were expecting, I think, 270 plus the 16.5.”
Adam Sullivan:
“Yes, thanks, Brett. Yes, that's correct. That's really a push out of just one 40-megawatt building out into early 2026. And you're absolutely right. That number does include the 16.5 megawatts.”
Core Scientific Implements a Utility First Process For Evaluating Expansion Sites
For data center site selection, there is a shift away from large remote training sites towards locations that are closer to major metropolitan areas. This has been driven by demand and the need for proximity as the Company expands into new markets, which include the East Coast. However, prioritization is based on reliable utility partnerships.
Rosemarie Sison:
“Just to follow up on that comment that you made, Adam, about proximity to major metro areas. Would that mean that you're potentially looking at expanding out of the markets that you're in right now possibly into the East Coast or the West Coast as those opportunities present themselves?”
Adam Sullivan:
“Yes, absolutely. Thank you for the question, Rosemarie. I mean, we are building one of the larger data centers on the East Coast right now. And so, we have a lot of confidence in our ability to continue to expand in new markets. This is something where we're going to be one of the larger providers in the Dallas market. We believe something similar in the Atlanta market as well. So, we're definitely looking at continuing to enter into new cities. But albeit that looks a little bit different because we might have less familiarity with the utilities in that location. A point on that is we currently operate with seven utilities. We're continuing to expand our relationships across that base. And so, we're taking a very diligent process, a utility-first process, when we're evaluating entering new locations to ensure that we have a strong partnership and relationship with that utility so that we know that we have that firm power available when we go take them to a client.”
Conclusion: Solid Gameplan, Execution is the Key
Other Bitcoin mining companies are adopting Core Scientific's pivot to HPC hosting. However, Core Scientific's game-changing contracts amounting to over $10 billion in revenues over 12-year terms with CoreWeave give them a first-to-market advantage fortified by $10.2 billion in revenue potential from an AI disruptor.
CoreWeave, backed by NVIDIA as an investor and customer, is likely the leading hyperscaler in the market, positioning itself as a first mover in the AI data center space. Given NVIDIA’s potential preferential treatment toward CoreWeave, especially compared to competitors like Amazon who are building with custom silicon, CoreWeave is primed to lead the way in scaling AI infrastructure.
Hyperscalers will likely follow in CoreWeave's footsteps. This dynamic reinforces the notion that Core Scientific's strategic pivot to HPC hosting could be bolstered by CoreWeave's leadership in the hyperscaler space, further underscoring that what’s good for CoreWeave is also good for Core Scientific.
CoreWeave was initially interested in acquiring Core Scientific for $1.02 billion or $5.75 per share in June 2024, but was rejected and they decided to back them as they expanded their data center footprint. The downside to this relationship is the very limited customer concentration, as CoreWeave is their largest HPC hosting client. Core Scientific’s near-term future lies with CoreWeave. CoreWeave is expected to generate $10 billion from Microsoft as a client by the end of the decade.
As a potential lottery ticket element for investors, CoreWeave could revisit another acquisition attempt for Core Scientific after its IPO, where it would have additional cash and stock to use as currency. The initial acquisition attempt in June 2024 was for $1.02 billion in cash or $5.75 per share, which Core Scientific rejected stating that the offer “significantly undervalues the Company.” With an estimated $35 billion valuation, CoreWeave could make a much more attractive acquisition offer for less than it would be paying Core Scientific over its 12-year term leases.
Core Scientific has a solid game plan to accelerate its quarterly HPC hosting revenue by at least 21X in two years. As with any great game plan, the flaw always lies in the execution. Analyst estimates forecast one more kitchen sink quarter to go before revenues turn back up as HPC hosting revenues start to ramp up. The potential for more than doubling the outstanding shares to 501 million shares upon full conversion and vesting of restricted stock is concern down the road, but for now the game plan looks solid; the execution is the key.
Welcome to the I/O Fund’s new Discovery Tier, where we cover a new stock idea on a weekly or bi-monthly basis. We are excited to bring you more coverage from the I/O Fund team geared toward new idea generation only.
Jea Yu, Equity Analyst at the I/O Fund, contributed to this article.
Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.
Alibaba is one of the hottest AI stocks in the market in 2025, with shares up more than 60% year-to-date. The surge in price is due to Q3 results showing Cloud revenue accelerated with AI revenue up triple-digits for a sixth quarter in a row. Notably, the triple digit growth is on much lower revenue than what United States Big Tech companies are reporting, with leader Microsoft at 13X higher AI revenue. AI is also not meaningfully contributing to revenue nor earnings despite six consecutive quarters of triple digit growth.
Despite the advancements that Alibaba is making in AI and tens of billions it is committing to invest in AI infrastructure, AI remains but a small portion of Cloud revenue, far below the scale of US-based peers. An ultra-competitive Chinese AI market engaging in a pricing war likely is hindering the country’s AI growth potential, as Alibaba is touting tens of thousands of users and millions of downloads with small dollar growth to show for it. I break this down and more below.
Alibaba has Made Visible AI Advancements
Alibaba remains committed to its dual-prong strategy of e-commerce and AI + cloud. The company has recently highlighted multiple advancements on the AI front, with a handful of its models showing “industry-leading” performance. Shares rose again on Tuesday as Alibaba struck a partnership with Manus AI to roll out a Chinese version of Manus’ rapidly popular AI agent.
China’s AI market is rapidly heating up with fierce competition from Alibaba, DeepSeek, Baidu, Tencent and Bytedance, among others. Alibaba is aiming to take a leading position in the AI market and announced recently that its Qwen2.5-Max mixture-of-experts model outperforms DeepSeek’s V3, Meta’s Llama 3.1 and OpenAI’s GPT-4o. In Q3’s earnings call, Alibaba teased the release of a new deep reasoning model built on Qwen 2.5 Max.
Alibaba says its Qwen model outperforms DeepSeek’s V3 and Llama 3.1-405B across major benchmarks.
Last week, Alibaba announced a new 32 billion parameter reasoning model, QwQ-32B, that it says outperforms DeepSeek’s R1 reasoning model, with 1/20th the parameters. QwQ-32B is the latest iteration of QwQ, which was first launched in November 2024 and “designed to enhance logical reasoning and planning by reviewing and refining its own responses during inference,” which a report from VentureBeat says allowed it to outperform OpenAI’s o1 on math benchmarks AIME and MATH, and scientific reasoning tasks.
Alibaba is committed to expanding access to its AI models and tools, with its Qwen-2.5 series of models available via APIs as well as its genAI development platform Model Studio. Developers also can access Alibaba’s family of multimodal models – its vision understanding model Qwen-VL, its visual generation model Wanx2.1, its audio language model Qwen-Audio, and its coding assistance model Qwen-2.5coder.
In Q3’s earnings call, Alibaba’s management highlighted that this approach is paying off, with more than 90,000 Qwen-based derivative models developed globally at the end of January, which made Qwen “the most popular among developers across the major model families.” Management added that more than 290,000 companies and developers accessed Qwen APIs through Alibaba Cloud's Bailian platform.
Alibaba Outlines Significant Capex Outlay to Support AI
To support its growth in AI and become a competitive global player, Alibaba outlined a plan to “aggressively invest in AI infrastructure,” with planned capex over the next three years exceeding what it has spent over the last decade. Management’s plan calls for spending of 380 billion yuan, or ~$52.4 billion, over the next three years predominantly for AI.
Management first discussed increasing capex in the June 2024 quarter, when expenditures nearly doubled YoY to ~12 billion yuan, or $1.6 billion. Management noted in the June quarter that confidence in this level of investment was driven by strong demand, and expected to stay around this level for the next few quarters.
However, capex quickly accelerated — in the September 2024 quarter, capex rose more than 310% YoY to 16.9 billion yuan, or $2.4 billion, and in the December 2024 quarter, capex rose 330% YoY to 31.4 billion yuan, or $4.3 billion. For the nine months of this fiscal year, capex totaled 60.3 billion yuan, or $8.3 billion, rising more than 246% YoY.
Capex Plans Well Below US Big Tech
This 30 billion yuan/quarter rate is a bare minimum of what’s needed to come close to management’s spending targets, signaling heightened capex will continue for the foreseeable future. However, at $52.4 billion over the next three years, Alibaba is spending only a fraction of what US Big Tech firms are spending, which could impact its competitive positioning.
For 2025, Amazon, Microsoft, Meta and Alphabet outlined plans for approximately $320 billion in capex, predominantly for AI, which is 6x more than what Alibaba is spending. Of the four, Meta has earmarked the least towards capex, at $60-65 billion, but even at the low end, Meta’s one-year spending is more than 15% higher than Alibaba’s 3-year plan.
In 2024 and 2025, Big Tech is on track to spend more than $550 billion in capex, or more than 10x higher than Alibaba’s plan in just two years as opposed to three. So while Alibaba’s AI spending commitment looks high compared to its historical investments, it’s spending just a mere fraction of what Big Tech is. The US is not likely to back down when it comes to AI dominance, which we covered in the article “DeepSeek Creates Buying Opportunity for Nvidia Stock.”
Alibaba Remaining Competitive on AI Pricing
China’s AI market is engaged in much fiercer competition than in the US, and this is evident within the pricing structures of AI models. Alibaba and rivals are pricing models at mere fractions of the cost of US-based competitors, in an effort to win over customers from each other and remain ahead in the broader global AI race.
China’s AI market has been in a pricing war since early 2024 – Alibaba had cut prices by up to 97% in May 2024. ByteDance further intensified the price war in December 2024, when it slashed prices for its new Doubao model with vision understanding capabilities to $0.00041 per 1,000 tokens, 85% lower than the industry average. Alibaba mirrored this move within two weeks, matching Doubao’s $0.00041 price for its Qwen-VL Max model.
Alibaba’s Qwen models are priced much cheaper than leading models from OpenAI and Anthropic as China engages in an AI pricing war.
Alibaba’s Qwen2.5 Max is priced at $0.0016 per 1,000 input tokens and $0.0064 per 1,000 output tokens, while its Qwen Plus is offered at $0.0004 per 1,000 input tokens and $0.0012 per 1,000 output tokens. Qwen Turbo is priced at $0.00005 per 1,000 input tokens and $0.0002 per 1,000 output tokens.
For comparison, ByteDance’s Doubao 1.5 Pro is priced at $0.00011 per 1,000 input tokens and $0.00028 per 1,000 output tokens. DeepSeek’s V3 model is priced $0.00014 per 1,000 input tokens and $0.00029 per 1,000 output tokens, while its R1 models is priced higher at $0.00057 per 1,000 input tokens and $0.00227 per 1,000 output tokens. Baidu recently announced that it is aiming to make Ernie free for all users by the start of April this year.
This pricing structure is allowing Alibaba to remain quite competitive in the Chinese AI market and more so on the global market, as China’s models are significantly cheaper than OpenAI, Anthropic and others, with Meta and Mistral two of the lower-cost competitors.
OpenAI’s GPT 4.5 is currently one of the most expensive models available, at $0.075 per 1,000 input tokens and $0.15 per 1,000 output tokens – that’s up to almost 50x more expensive that Qwen2.5 Max. OpenAI’s o1 is priced at $0.0015 per 1,000 input tokens and $0.06 per 1,000 output tokens, while the much smaller GPT 4o mini is priced comparatively to Chinese rivals.
Pricing is Alibaba Stock’s Achilles Heel for AI
However, it is this pricing structure and ongoing price war in China’s AI model market that is Alibaba’s Achilles heel. While the low-cost structure is enabling Alibaba to remain extremely competitive in the face of rising competition both domestically and globally, it’s serving as a bit of a hindrance to growth, with AI revenue likely still quite below the $1 billion mark, where US tech giants are touting multi-billion dollar AI revenue streams.
Alibaba first outlined the growing demand for AI model training and AI infrastructure services in its June 2023 quarter. The following quarter in September 2023, Alibaba laid out a two-pronged strategy for driving AI growth in the cloud.
Management said that they will aim to “build the most open cloud in the AI era, providing stable and efficient AI infrastructure for all industries and enabling all sectors to go intelligent,” and “build an open and prosperous AI ecosystem.” This was underscored by its Qwen family of models, its model application development platform Bailian, and its open-source platform ModelScope. The September quarter saw ModelScope’s cumulative downloads more than double sequentially, from 45 million in July 2023 to 100 million, attracting more than 2.8 million developers.
Discussing the March 2024 quarter results, CEO Eddie Wu explained that AI is serving as a primary driver for the revenue growth that is being seen in the broader Cloud segment:
“If you look at the overall revenue growth of the Cloud business today, most of that is already being driven, I would say by AI and AI-related new products. So going forward a lot of the incremental growth we can expect to see in the Cloud business will be related to investments the customers are making in AI. But also there is a complementary effect because the more that customers invest in and make use of AI the more demand they will also have for other of our various cloud offerings.”
For the December 2024 quarter, Cloud Intelligence revenue accelerated 6 points sequentially to 13% YoY to ¥31,742 million, or $4.35 billion, up from 7% growth in the September quarter. This marks the steepest sequential increase in what has been a rather gradual acceleration from 2% YoY growth in the September 2023 quarter.
Alibaba stock’s Cloud revenue growth accelerated to 13% YoY in the December quarter, aided by AI.
In its December 2024 quarter results, Alibaba noted that AI product revenue “maintained triple-digit year-over-year growth for the sixth consecutive quarter,” starting in the September 2023 quarter. AI helped drive an acceleration to the double-digits for Cloud Intelligence revenue growth.
There are a handful of stats that support increasing adoption of Alibaba’s AI products – more than 90,000 Qwen-based derivative models had been developed globally at the end of January, while more than 290,000 companies and developers have accessed Qwen APIs through Alibaba Cloud's Bailian platform. Qwen’s models have been downloaded more than 7 million times, and ModelScope has attracted more than 4,000 AI models and 5 million developers.
The I/O Fund specializes in covering lesser-known AI stocks on our research site with trade alerts and weekly webinars. Learn more here.The I/O Fund specializes in covering lesser-known AI stocks on our research site with trade alerts and weekly webinars. Learn more here.Learn more here.
Yet despite this and six quarters of triple-digit growth, Cloud revenue is only growing in the low-double digits, implying that AI’s contribution to revenue remains quite small. A rough estimate places AI’s contribution in the mid-single digit percentage of Cloud revenue, with revenue possibly around the $200-275 million range as of the December quarter. This would put Alibaba’s AI run rate between $800 million to $1 billion.
Compare this to Microsoft, where its AI run rate on Azure surpassed $13 billion last quarter, up 175% YoY. Microsoft is also showing rapid growth for AI platforms and tools – the number of Azure OpenAI apps running on Azure databases more than doubled YoY last quarter, while Azure AI Foundry reached 200,000 MAUs within two months. Microsoft’s Phi family of small language models have been downloaded more than 20 million times, nearly 3x more than Qwen. 160,000 organizations have used Microsoft’s Copilot Studio, creating 400,000 custom agents last quarter, up 2x QoQ.
AI’s Impact on Alibaba’s Cloud Margins
Alibaba noted last quarter that a shift to higher-margin cloud products, including AI, has aided EBITA growth in its Cloud segment. For the December quarter, adjusted EBITA rose 33% YoY to ¥3,138 million, or $430 million, decelerating dramatically from 89% YoY and 155% YoY growth in the prior two quarters.
Adjusted EBITA margin was 9.9% in the December quarter, up from 9% in the prior quarter. It’s hard to argue against the beneficial impact of AI on EBITA margin for Cloud, with margins beginning to expand significantly as AI embarked on its six-quarter stretch of triple digit growth in the September 2023 quarter, aside from a hiccup in the March 2024 quarter. Since that point, margin have risen nearly 5 points and are knocking on the double-digit range.
Alibaba stock’s Cloud adjusted EBITA margins have expanded as AI drives a shift to higher-margin products.
Despite the margin expansion and strong EBITA growth from a shift in product mix towards higher-margin cloud offerings and AI products, Cloud’s share of consolidated adjusted EBITA is still quite small. For the last four quarters, Cloud’s share has hovered between 5% to 6.6% of consolidated adjusted EBITA. While this was an improvement from 0.9% in the June 2023 quarter and 3.3% to 4.5% in the second half of 2023, segment adjusted EBITA growth has decelerated sharply to the lowest level in the past seven quarters, suggesting EBITA contribution may follow and plateau.
Alibaba’s Cloud Intelligence adjusted EBITA growth has decelerated more than 120 points in two quarters to 33% YoY.
What this means is that despite the six-quarter string of triple-digit growth for AI revenue, there’s minimal impact to the bottom line from this AI revenue surge at the moment. Earnings estimates for this fiscal year and next were relatively unchanged through much of the second half of 2024, within a 3% range, only rising in February following a 10% earnings beat in the December quarter.
Alibaba’s EPS estimates for this fiscal year and next were relatively unchanged through most of 2024 despite AI growth. Source: YCharts
AI revenue is not yet at the scale where it is meaningfully contributing to revenue or earnings, though Alibaba’s commitment to spend significantly on AI after witnessing six quarter of triple-digit growth is more positive for the long-term rather than the short term.
Valuation Reaching a Peak
Because AI is not driving the revenue scale or profits that we are seeing here with US Big Tech, Alibaba’s valuation is getting pricey, trading at peak levels from the past three years. Alibaba not only is facing tough competition from Big Tech but also from within domestic peers, with Tencent and Baidu both reporting strong AI growth.
Alibaba is trading at peak valuation levels from the past three years. Source: YCharts
Alibaba is currently trading at 15.1x forward earnings and 2.35x forward revenue, both at or just below peak valuations since early 2022. This rapid rerating has likely been driven predominantly by AI enthusiasm, given that a majority of the multiple expansion occurred following DeepSeek’s rout in late January.
For comparison, Baidu trades at a 40% discount to Alibaba on both metrics, at 9x forward earnings and 1.7x forward revenue, with its AI Cloud revenue rising 26% YoY, double the rate of Alibaba’s. Baidu’s Ernie handled 1.65 billion daily API calls in December 2024, with external API calls up 178% QoQ. Baidu’s Wenku platform reached 94 million MAUs, up 216% YoY and 83% QoQ.
There’s also risk that China remains behind the US when it comes to AI and monetization, with Tencent VP Martin Lau laying out three reasons why it lags behind US peers despite AI revenue reaching 10% of Cloud revenue. He explained that China does not have nearly as large as an enterprise market as the US, and within that, the SaaS ecosystem “is not really that vibrant in China.” He added that fewer AI startups in China are purchasing less compute, another reason the US leads. These three reasons are why Lau believes that AI revenue is starting to scale, but not exploding as it is in US.
Conclusion
Alibaba is one of AI’s top winners so far in 2025 with shares rising more than 60% YTD on AI enthusiasm as the giant has released highly-competitive Qwen models and struck partnerships with leading Chinese AI firms. However, the rally is likely front-running AI revenue to a significant degree, as Cloud’s low growth and management’s comments about AI revenue imply that AI is still growing off quite a small base.
Alibaba is quickly ramping AI investments to better compete on the global scale, but its AI run rate far lags that of US peers, with Microsoft recently reporting 175% YoY growth to a $13 billion AI run rate and Amazon and Google both reporting in the multi-billion dollars. Alibaba has a lot of ground to cover to get into the same realm as Big Tech on AI, as its AI run rate is still likely below the $1 billion mark.
The I/O Fund has recently added five new small and mid-cap positions poised to benefit from the ongoing AI spending war. Join us every Thursday at 4:30 p.m. for our exclusive 1-hour webinar, where we cover market entries, exits, and key insights on the broader market. Take advantage of our limited time monthly promotion for up to $110 off Pro. Learn more here.The I/O Fund has recently added five new small and mid-cap positions poised to benefit from the ongoing AI spending war. Join us every Thursday at 4:30 p.m. for our exclusive 1-hour webinar, where we cover market entries, exits, and key insights on the broader market. Take advantage of our limited time monthly promotion for up to $110 off Pro. Learn more here.here.
Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.
Vistra is an independent power product (IPP) with 41 GW capacity, operating the second-largest fleet of nuclear plants in the United States.
Vistra has closed power purchase agreements (PPAs) to supply Amazon with 200 MW and Microsoft with 405 MW of renewable energy for their data centers in Texas and Illinois, respectively.
The Company is in talks with two unnamed hyperscalers for natural gas plants co-located with their data centers.
Co-location involves building data centers or power plants on-site to provide electricity directly to the data center, bypassing the electrical grid. This is known as "behind the meter."
Regulations on co-location builds are materializing as hyperscalers await the precedent-setting decision on Talen Energy/Amazon co-location reliability solution vs FERC/AEP/Exelon push for cost equity as a major catalyst.
Vistra has said they are in talks with hyperscalers and data center developers for nuclear power but hasn’t announced any deals yet.
Vistra Corp. (NYSE: VST) is the largest competitive power generation and retail electric company in the United States. It’s an integrated independent power producer (IPP) based out of Irving, Texas, serving five million customers across 20 states and Washington, D.C., with a power generation capacity of 41 GW (41,000 MW). It has the second-largest competitive nuclear fleet in the country, behind Constellation Energy. Texas is a hotspot for data centers and chip manufacturers.
The company has power purchase agreements (PPA) with Amazon and Microsoft and is actively “speaking with all the major hyperscalers” and actively engaged with the major data center developers, according to Vistra CEO Jim Burke during their Q4 conference call. As of March 1, 2025, Vistra energy has 41 GW capacity, comprised of 27 GW for natural gas, 6.4 GW for nuclear, 6 GW for coal and 2 GW for renewable and battery/energy storage (1 GW), which includes solar assets with 350 MW already online.
The Energy Harbor Acquisition Bolstered Vistra’s Nuclear Footprint
On March 1, 2024, Vistra completed the acquisition of Energy Harbor Corp., adding an additional 4 GW of 24/7 nuclear generation and 1 million retail customers. The Company paid $3 billion in cash and a 15% equity stake in a new subsidiary comprised of all of its nuclear fleet (6.4GW) called Vistra Vision. The transaction closing followed receipt of all required regulatory approvals, including the from the Federal Energy Regulatory Commission (FERC) in February 2024.
In Q3 2024, the Company bought back the remaining 15% minority interest stake in Vistra Vision for approximately $3.248 billion from affiliates of Nuveen and Avenue Capital Management LLC. The buyback increased its nuclear ownership stake to the full 6.4GW capacity, not to be mistaken with adding any more capacity or increasing its licensed output (uprate). This results in keeping all the proceeds they were distributing to the minority stakeholders in Vistra Vision, which will result in a bump up in net income moving forward, however, they still have to pay $3.248 billion for it.
Vista expects to pay in five installments of $1.18 billion on December 31, 2024, $114 million on June 30, 2025, $1.0 billion on December 31, 2025, $54 million on June 30, 2026, and $900 million on December 31, 2026. The net present value of the purchase price as of December 31, 2024, discounted at a 6% interest rate, is $3.085 billion.
Complying with the Electrical Grid Operators and FERC
Here’s a quick summary of the many acronym organizations in the electricity market. As an IPP, Vistra actively deals with the electrical grid operators where they sell their electricity. Vistra sells wholesale electricity to markets managed by the PJM Interconnection (Pennsylvania, New Jersey, Maryland), which covers 13 states. PJM is an independent system operation (ISO) and regional transmission organization (RTO). They run the wholesale electricity markets for the regions, balancing the needs of market participants and ensuring grid reliability. They coordinate the flow of electricity from the power generators to local utilities.
ERCOT (Electric Reliability Council of Texas) performs the same ISO and RTO functions in Texas, managing high-voltage transmission grids. Texas is the only state that owns its power grid. ERCOT was the first ISO in the United States. Vistra also owns power plants and sells electricity to markets managed by ERCOT. The ISOs and RTOs are regulated by the Federal Energy Regulatory Commission (FERC).
Co-Location and BTM are Growth Drivers for Vistra
Front of the meter (FTM) refers to power produced by the power plants that flow to the electrical grid and then to the data center or customer. Behind the meter (BTM) refers to electricity that’s generated and directly delivered to the data center, bypassing the grid. Co-location deals where utilities like Vistra install gas power plants on-premise for the hyperscalers are BTM, as the power is generated on-site and delivered directly to the data center. Vistra can repurpose existing gas plants or construct new builds. BTM and co-location are Vistra’s AI hyperscaler growth strategy, but so far, there haven’t been any deals announced yet.
The time to power depends on the type of gas plant, whether it's repurposing an existing gas plant or constructing a new gas plant. Repurposing an existing gas plant can take six to 24 months, using spare capacity. New builds can take three to five years due to gas infrastructure, permits, construction and commissioning (IE, Building up to 860 MW of new gas plants in West Texas).
Vistra is a Nuclear Energy Powerhouse Primed for AI Data Centers
As of March 5, 2025, Vistra operates four active nuclear plants with a combined power capacity of over 6.4 GW, which is enough zero-carbon baseload capacity to power 3.2 million homes across PRJM and ERCOT markets. These include:
Camanche Peak Nuclear Power Plant located in Texas, comprised of two units with 2,400 MW total capacity, with licenses extended to 2050 and 2053, approved by the Nuclear Regulatory Commission (NRC) in July 2024.
Beaver Valley Nuclear Power Plant in Pennsylvania is comprised of two units with a 1,800 MW total capacity. NRC licenses expire in 2036 and 2047.
Perry Nuclear Power Plant in Ohio is comprised of one unit with a 1,300 MW capacity. NRC license expires in 2046.
Davis-Nesse Nuclear Power Plant in Ohio is comprised of one unit with 894 MW capacity. NRC license expires in 2037.
Beaver Valley, Perry and David-Neese Nuclear Plants were acquired from Energy Harbor for $3.4 billion in March 2024. The nuclear plants are operated under Vistra Vision, and the fossil fuel segment operates under Vistra Tradition. Nuclear energy comprises just over 15% of its total energy production. Vistra Energy’s nuclear fleet operates at 92% of its maximum capacity (aka capacity factor) in Q4.
Nuclear Energy and Gas Power Plants Talks with Hyperscalers
Vistra signed a 200 MW power purchase agreement (PPA) with Amazon for a Texas solar facility in October 2024. Vistra signed a 405 MW PPA with Microsoft for a solar project in Illinois during Q3 2024. Incidentally, Vistra hasn’t announced any major nuclear power purchase agreements (PPA) with major hyperscalers yet. However, the Company is in early discussions with some hyperscalers about nuclear uprates and some new builds as well.
Co-location is the trend of building data centers or power plants near data centers to alleviate pressure on the electrical grid and reduce transmission loss to ensure maximum efficiency.
Vistra is in discussions with two unidentified hyperscalers to build new natural gas power plants co-located with data centers. During its Q3 2024 conference call, Vistra's Head of Strategy, Stacey Dore, said this.
“There's a lot of interest, obviously, in the nuclear side. However, we have ongoing conversations with several different development companies about a handful of our gas sites, both in PJM and in ERCOT. And we're in early discussions with some of the hyperscalers about nuclear uprates, and some new build as well as Jim mentioned. And then, finally, we're in discussions with two particular large companies about building new gas plants to support a data center project.”
Diore also cautioned that these talks aren’t an overnight decision, “As we've said before, the diligence process for these deals takes a long time. It's an intense effort because these are very long-term commitments to purchasing power.” The announcement if and when these two hyperscalers closed the deal should be a catalyst for the stock.
How Natural Gas is Used to Generate Electricity
Vistra generates 66% or 27 GW of its energy capacity from natural gas using gas-fired power plants. The primary method for large-scale natural gas generation (1MW or more) is combined-cycle gas turbines (CCGT), which connect a gas turbine with a steam turbine. The natural gas is burned in a combustion chamber with compressed air to spin the turbine connected to a generator producing electricity. The exhaust heat can climb up to 1,000 degrees Fahrenheit, which is used to boil water to generate steam, driving a second turbine to generation addition electricity.
Simple-cycle turbines operate with just the natural gas turbines, without the steam turbines. Natural gas combustion turbines generate 35% to 42% of direct electricity. The steam turbines add another 20% to 25%, yielding total thermal efficiencies between 55% to 67%, which is the energy output vs fuel input. As natural gas is delivered through pipelines, around 92% is actually delivered, as 8% energy loss is common. Coal plants yield around 33% thermal efficiency. Nuclear plants yield 33% to 37% thermal efficiency, similar to or slightly better than coal. The main difference is the carbon emissions, which are virtually none, making nuclear the cleanest option of the three.
On a smaller scale (325 KW), we wrote about how Bloom Energy Servers (BES) can reach 85% to 90% thermal efficiency through its non-combustion, electrochemical reaction method using solid oxide fuel cells and heat capture.
“BES is designed to work with existing carbon capture utilization and storage (CCUS) and combined heat and power (CHP) technologies. CCUS mitigates emissions from natural gas as BES generates a pure stream of CO2 that can be used or sequestered. CHP allows the exhaust heat generated by BES (operating at a core temperature of 1,500 degrees Fahrenheit or 800 degrees Celsius) to be channeled and made available for use, further increasing the efficiency of the system.“
AI Applications are Driving AI Data Center Power Needs
AI applications require much more electricity to operate, depending on the applications. AI training and inferencing drive power demand, “Wells Fargo is projecting AI power demand to surge 550% by 2026, from 8 TWh in 2024 to 52 TWh, before rising another 1,150% to 652 TWh by 2030. This is a remarkable 8,050% growth from their 2024 projected level. AI training is expected to drive the bulk of this demand, at 40 TWh in 2026 and 402 TWh by 2030, with inference’s power demand accelerating at the end of the decade. In this model, the 652 TWh projection is more than 16% of the current total electricity demand in the US.”
As IO Fund pointed out, “The Electric Power Research Institute forecasts that data centers may see their electricity consumption more than double by 2030, reaching 9% of total electricity demand in the US. The IEA is projecting global electricity demand from AI, data centers and crypto to rise to 800 TWh in 2026 in its base case scenario, a nearly 75% increase from 460 TWh in 2022. The agency’s high case scenario calls for demand to more than double to 1,050 TWh.”
Vistra Leverages Clean Energy Tax Credits and Incentives
Vistra currently does and could leverage many forms of energy credits to help squeeze every bit of margin. Its customers can also benefit from tax credits under these programs. Here are some of the most lucrative credits:
The Inflation Reduction Act (IRS) of 2022 introduced a Nuclear Production Tax Credit (PTC) of up to $25 per megawatt-hour (MWh), which runs for 10 years through 2032 for facilities in service prior to January 1, 2023. The PTC kicks in went prices are below set limits. Vistra recognized a $545 million benefit from the nuclear PTC in Q4 2024.
The Investment Tax Credit (ITC) was extended by the IRA through to December 31, 2024, under section 48, offering a 30% credit for projects started prior to then with up to four years to go online. New projects starting on January 1, 2025, and after shifts to section 48E, the Clean Electricity Production Credit, which applies to zero-emission projects through 2033
The Clean Electricity Production Tax Credit (PTC) under section 48E replaced the traditional PTC with technology-neutral clean electricity PTC, which offers up to 2.75 cents per kilowatt-hour (kWh). Projects that begin construction before 2033 and meet wage/apprenticeship rules can receive the credit. Vistra can receive PTC for any upgrades (capacity increases) or new builds after 2025. It could have the potential for up to $50 million annually in solar and scaling higher with new projects. Vistra’s solar projects with Amazon (200 MW PPA) and Microsoft (405 MW PPA) are likely claiming the ITC of 30%, which covers installation costs. It’s 600 MW planned battery story planned in Texas also qualifies. Nuclear uprates could also apply if started post-2025.
The Energy Community Tax Credit Bonus adds a 10% tax credit (10% points) to the existing ITC for projects in energy communities and areas with close coal plants/mines. Vistra operates in many former coal mining regions in Ohio, Pennsylvania and Texas.
The Trump administration has mentioned it plans to rollback many IRA provisions, which could cap or limit tax credits and incentives moving forward.
Financials: MTM Caused a Non-Cash Surge to Financial Metrics
Note that Q3 2024 shows a surge in net income to $3.465 billion, up 159% YoY from $1.335 billion in Q3 2023, driven primarily by unrealized mark-to-market (MTM) gains on derivative positions and the addition of Energy Harbor. This jolted various metrics, including gross and operating margins, GAAP EPS and gross profits. Still, since most of it was a non-cash gain, the operating cash flow didn’t surge proportionally. The mark-to-market gains were non-cash, but the addition of Energy Harbor gains were cash and are here to stay moving forward. Vistra expected $700 million in contributions from the Energy Harbor business for 10 months. The adjusted EBITDA excludes the impact of unrealized gains or losses on derivatives, which makes for a better measure of operating performance. CFO Moldovan clarified this in the Q4 conference call.
“Including the nuclear production tax credit, our adjusted EBITDA was more than $850 million above the midpoint and more than $600 million above the top end. Notably, the 10-month contribution from Energy Harbor, including the nuclear PTC, exceeded our $700 million expectation by approximately $200 million.”
Revenue is Lumpy, But The Energy Harbor Acquisition is Accretive
Q4 revenue rose 31.16% YoY but fell (35.8%) QoQ to $4.04 billion, beating the single analyst estimate by 3.2% or $124 million, driven primarily by the inclusion of results from the Energy Harbor acquisition and an increase in revenues due to the estimated nuclear PTC recorded in the quarter. The negative QoQ was primarily due to the surge in Q3 "driven primarily by unrealized mark-to-market gains on derivative positions."
Adjusted EBITDA from Ongoing Operations: Bypassing the MTM Noise
Due to the MTM unrealized gains on derivatives, the adjusted EBITDA from the ongoing operations metric provides a more accurate picture of the operations. Q4 adjusted EBITDA from ongoing operations was $1.985 billion, up 104.2% YoY and 37.85% YoY. This was an improvement from Q3 adjusted EBITDA of $1.44 billion, down (10.73%) YoY and up 1.84% QoQ.
Net income for the full year 2024 increased by $1.32 billion, driven primarily by unrealized MTM gains on derivative positions, the addition of Energy Harbor and an increase in revenues due to estimated nuclear PTC recorded in Q4. Ongoing adjusted EBITDA for the full year 2025 increased $1.516 billion YoY primarily due to the inclusion of results from the Energy Harbor acquisition and estimated nuclear PTC recorded in Q4 2024. Full year 2025 adjusted EBITDA from ongoing operations was $5.656 billion. Management guided full year 2025 ongoing operations adjusted EBITDA of $5.5 billion to $6.1 billion, with the midpoint of $5.8 billion. Management has "high confidence" in an adjusted EBITDA midpoint opportunity above $6 billion in the full year 2026, as its hedge ratio has increased from 64% to 80% since Q3.
The hedge ratio is the percentage of future (2026) expected electricity generation in megawatt hours (MWh) that is already locked in at a fixed price (IE, 80% is locked in at a fixed price) through derivatives like futures, options or swaps. The unhedged portion (IE: 20%) is exposed to market prices, which may rise, thereby raising EBITDA or potential fall and sinking EBITDA.
Margins: MTM and PTC Surge Q4 Improvement by 348.7%
Q4 gross margin was 39.6%, up 348.7% YoY and down (28.14%) QoQ, largely due to the MTM unrealized gains on derivatives, as mentioned earlier. The large YoY surge was due to the near doubling of gross profit to $1.6 billion in Q4 2024 compared to $850 million in Q4 2023, partially driven by the $545 million nuclear production tax credit (PTC), which reflects the full-year credit and the inclusion of results from the Energy Harbor acquisition. However, the jump was primarily based on a non-cash event, which investors shouldn't mistake for organic growth.
Lumpy Cash Flow as Debt Reaches Highest Level of 2024
Q4 cash flow reached $1.353B as operating cash flow rose to its highest level of five quarters at 33.5%. However, cash flow has been lumpy, ranging from $312 million in Q1 to a peak of $1.702B in Q3. Free cash flow closed Q4 at $923 million but was just as lumpy at ($153 million) in Q1, peaking at $1.017 billion in Q3 2024. Vistra closed Q4 2024 with $1.19 billion in cash and cash equivalents. The debt reached its highest level in five quarters, closing the year at $17.49 billion. Net debt to adjusted EBITDA is 2.9X. The Company plans on executing $2 billion in stock buybacks in 2025 and 2026.
Conference Call: Potential 10% Nuclear Capacity Increase by 2030
CEO Jim Burke reviewed 2024 events, including acquiring three nuclear sites and 1 million retail customers. They also completed a 20-year license renewal for the Camanche Plant nuclear power plant and secured Amazon and Microsoft PPA agreements with its renewables pipeline. The pre-Q&A can be summed up with the following:
Vista is in the early stages of the development of two natural gas peakers, power plants using natural gas during high demand periods, "peak times," totaling up to 860 MW of capacity. They are targeting mid-2028 for commercial operations.
Vistra continues to execute its zero-carbon growth strategy by leveraging existing land and interconnections to develop solar and energy storage projects opportunistically. They brought two solar and energy storage facilities online at its Coffeen and Baldwin, Illinois, sites. These facilities are part of the Illinois coal-to-solar and energy storage initiative. Vistra has begun construction of its Oak Hill, Texas, site for its contract with Amazon and the Pulaski, Illinois, site for its contract with Microsoft. Once they go online, it will add more than 600 MW of renewable capacity to Vistra’s portfolio.
Vistra has engineering studies in process with initial estimates indicating the potential uprates across their nuclear fleet of nearly 10%. Uprates increase a nuclear plant's power output without having to build new reactors. This can be performed with technology upgrades and improved turbines. It means they have the potential to add an extra 640 MW of additional capacity to their portfolio by the early 2030s when they expect to go online. This could equate to an extra $258 million a year at $50/MWh x 5.16 TWh, and even higher towards $361 million annually, with hyperscaler PPAs paying a premium of $70/MWh.
Texas policymakers are concerned about grid reliability and the challenges of accommodating rapidly growing energy demand, particularly from AI data centers. Despite these concerns, market reforms to incentivize new generations have been limited, raising worries among both generators and large-load customers, including data centers. While it's unclear if data center customers will alter their decisions during ongoing legislative discussions, there are potential solutions.
Vistra plans to spend $700 million on solar and energy storage products in 2025, which includes solar projects for Amazon and Microsoft.
Q&A: Gas Power Plants and Nuclear Plants for Data Centers
Analysts had tunnel vision during the Q&A; data centers. Right off the bat, Management was asked about the primary impediment to getting the deal done with the hyperscalers. CEO Burke responded that it’s not as easy as just signing a contract. The “flavor of the deal” matters.
Burke assured, “So you can assume that we're speaking to all the major hyperscalers and that we're actively engaged with them and the major data center developers. But there are flavors of complexity. So, the virtual PPA, which would be a front-ended leader, would be a relatively straightforward deal to execute. We have a number of discussions going on with those. Those do not offer, we think, the same margin potential as the more complicated deals, which do involve co-location, whether it's with existing assets or new assets.”
Burke further elaborated on co-location deals (colo-deals), which include risk-sharing for 10 to 20 years.
“And you've seen not many deals have been announced that have actually been co-location related. We think colo-deals offer a lot of benefits for not only the data center customer and our fleet but the market overall — the overall customer base because it not only provides speed to market for the customer but can also result in more transmission build-out that the grid absorbs today.”
Front-of-the-meter (FTM) power purchasing agreements (PPAs) with data centers are a lower-margin product. Burke elaborated that the highest margins come for co-lo deals offering speed to market using an existing resource. Burke noted that the sweet spot is understanding their value proposition.
The complexity of these deals has sparked elevated discussions in regulatory and policymaking circles, with FERC and Texas recently addressing co-location issues. As customers seek clarity on the rules, the timing of any announcements will depend on the resolution of these regulatory discussions in PJM and Texas. Although front-of-the-meter virtual PPAs are still a possibility, co-location deals—where the load and generation asset are close—remain the ideal. The company is actively engaging in these discussions and is optimistic about progress, though further clarity is needed before any deals are finalized. Burke said the Comanche Peak opportunity is considered the most attractive and fastest to execute in its portfolio.
Status of Gas Power Plants for Co-Located Data Centers
Morgan Stanley analyst David Arcaro asked about the prospects of potential gas co-location. Vistra Head of Strategy Stacey Dore started by saying they are seeing interest in their existing gas sites from data center developers at this point. Since a grid connection is typically needed at a gas power plant, the regulatory approval process applies even for a front-of-the-meter interconnection, which adds more time and complexity, especially when adding batteries and other backup generators to replicate reliability. Their gas lines don’t have as much land associated with them as a nuclear site, so the challenge of where to build the data center exists but continued to say:
“Having said that, we're progressing on a lot of those conversations on a handful of our sites in detail, working on agreements to bring those projects to fruition. And in addition to that, we are in a number of conversations about building new gas for data centers as well. So we have a number of conversations going on that are at the papering stage. And as Jim referenced earlier, those agreements can be complex, but we are optimistic about our ability to bring those projects to a close.”
CEO Burke believes a cap and floor and likely to be approved in the next two auctions. While some worry about the grid doubling by 2030, Vistra anticipates a more moderate peak demand growth in the 3% to 5% range rather than double-digit increases.
Talen/Amazon Deal will Set Tone for AI Data Centers and Colocation
The Talen and Amazon deal highlights the resource adequacy issue, sparking a conversation about whether co-location or transmission charges should be addressed. While co-location helps minimize strain on the electrical grid as electricity is funneled directly to the data center bypassing the grid, critics argue it actually deprives consumers of capacity while shifting transmission costs to them unfairly. American Electric Power (AEP) and Exelon made the case that Amazon was getting a “free ride” by co-locating adjacent to Talen’s Susquehanna Nuclear Power Plant and shifting up to $170 million in grid costs to their customers, while depriving them of electric capacity.
Vistra supports efforts to require customers larger than 75 MW to shed load during critical peak hours, and the customers they’re talking to are preparing their designs for this. But when the talk of remote disconnect switches arises, it's unusual and only would exist for these customers, which gives them pause.
Are Microgrids and Off-Grid Energy Servers the Solution?
Shedding loads during critical peak hours could be done through micro-grids. As IO Fund wrote in Bloom Energy: Fuel Cells for the Booming AI Data Center Trend Bloom Energy: Fuel Cells for the Booming AI Data Center Trend in the Discovery Tier, Bloom energy servers (BES) can be stacked. “The 325 kW base blocks can be duplicated and scaled up to multiple MWs for any project. They can also be used as the primary power source. They can be used off-grid or parallel as a microgrid. BES has a high density of 100 MW per acre.” The ideal has gained traction as evidenced by the game changer deal made with Bloom Energy and American Electric Power (AEP) to procure up to 1GW of Bloom’s solid-oxide-fuel-cells (SOFCs) for their hyperscaler customers.
Customers Waiting on Transmission Charges
Burke insists customers need clarity on transmission charges, which was a big choking point with Amazon and Talen. Amazon claims they shouldn’t have to pay transmission fees since their data center is co-located adjacent to Talen Energy’s Susquehanna nuclear plant. The data center would be powered behind the meter directly from the nuclear plant, bypassing the grid. However, FERC rejected the amended interconnection service agreement (ISA) to increase to 480 MW with a potential up to 960 MW. American Electric Power (AEP) and Exelon argued that diverting that much power would reduce the electricity available to the 65 million residents in the regional PHM grid, which could destabilize the grid, and shift up to $140 million in costs onto other ratepayers, essentially giving Amazon a "free ride." The case could set a precedent with co-location agreements around the country. Burke said this:
“They just want to know, are they going to get something that's commensurate with the transmission and grid utilization. These are revisiting potentially of the four coincident peak methodology. That probably does need revisiting because some customers are able to either reduce their load or turn on back up and minimize their transmission exposure, which means those costs go to other customers, including potentially residential customers. And since we serve nearly 5 million customers, we're sensitive to that as well.”
Seaport analyst Angie Storozynski bluntly asked management why they hadn't heard of any gas deals. Dore stated the co-location deals with existing assets are waiting on regulatory clarity as well. Whether gas or nuclear, regulatory clarity is what customers are waiting on.
“I mean the same uncertainty that is applying to behind-the-meter or co-located deals in PJM, for example, with the FERC proceedings, would apply whether the asset is nuclear or gas. Because the question really that's being asked is, what is the transmission charge, if any, that has to be paid on those deals.”
Vistra Expects FERC Clarity in the Second Half of the Year
Dore explained that Vistra isn’t waiting for full clarity from FERC or Texas before announcing deals, but ongoing legal proceedings (Amazon/Talon) are raising questions around risks like changing laws. The company is continuing work on projects like Beaver Valley and Comanche Peak, progressing without pause despite regulatory complexities. Jim Burke added that while co-locating load with plants raises concerns about grid adequacy, the company believes that co-location can benefit the grid by reducing transmission build-out and enhancing efficiency. Customers want to be seen as contributors to economic development, and co-location offers a faster market entry.
Despite regulatory delays, the company remains committed to advancing these discussions. Dore further noted that FERC's recent order sets a positive timeline for co-location and recognizes that there is no resource adequacy difference between front-of-the-meter and behind-the-meter loads. The challenge lies in balancing the customers' desire for fast deployment with policymakers' concerns about grid stability. The company hopes for clarity from FERC within the next 5-6 months to move co-location projects forward.
Conclusion:
Vistra is technically a utility company. Historically, these stocks have been considered defensive income plays due to their stagnant growth and consistent dividend yields. The AI boom has triggered interest in utility company stocks on a convincing thesis that power consumption is the chokepoint with AI. Utility companies will make more money with the explosion of power consumption driven by AI driven by data centers.
For utility companies, there are a number of ways to grow: acquisition (IE: Energy Harbor), rising energy prices (rate hikes), increasing capacity (uprates), increasing customers (IE: Energy Harbor = one million new customers), regulatory incentives (PTC, ITC, CEPTC, ECTC), cost reductions and new service offerings (co-lo). Co-location is the highest margin option with hyperscalers, but they are waiting on the Amazon/Talen FERC ruling in 2H 2025 to set a precedent on transmission fees and whether co-located data centers have to pay their share for grid maintenance and upgrades even without using the grid.
Vistra confirmed they are in talks with hyperscalers for co-location and new builds. However, they haven’t announced any nuclear PPAs so far or named the two hyperscalers getting gas plants. Announcements with hyperscalers will be a catalyst for the stock. Despite the stock being up 87% on a trailing twelve-month basis, the P/E is reasonable at 15.56.
Vistra was just a normal utility company until its announced the completion of the Energy Harbor acquisition on March 1, 2024, adding three nuclear plants and one million additional customers, this sent the stock surging from the $54.69 level gaining momentum on the AI data center “halo” as the markets turned to power producers as the next growth segment. Vistra announced Q2 earnings on August 8, 2024, revealing the two long-term PPA with Amazon and Microsoft, which were the first major hyperscalers they contracted. This sent VST stock on an upward trajectory from $80.46 to a peak of $199.84 on Jan. 23, 2025, just before the Q4 2024 earnings results, which triggered the downward trajectory as the Company only has two hyperscaler under contract (AMZN, MSFT), but are in talks with two hyperscalers for co-located gas plants. Investors put the cart in front of the horse, but shares are pulling back in time for the reveal of some more hyperscaler PPA news.
Investors should see how the stock reacts to announcements with hyperscalers, as it will either put in a new floor on the stock price or trigger a sell the news reaction as the AI “halo” dissolves. For now, it’s a waiting game for announcements and regulatory decisions.
Welcome to the I/O Fund’s new Discovery Tier, where we cover a new stock idea on a weekly or bi-monthly basis. We are excited to bring you more coverage from the I/O Fund team geared toward new idea generation only.
Jea Yu, Equity Analyst at the I/O Fund, contributed to this article.
Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.
AI data centers are expected to consume up to 9% of all electricity generated in the United States by 2030.
AI GPUs have already tripled their power consumption, with Nvidia Blackwell raising the bar as the GB200 is expected to use up to 2,700 watts of power, up from 250 watts for the earlier A100.
Power consumption is the chokepoint for AI data centers that are desperate to lock in power purchase agreements (PPAs) to procure long-term power.
The artificial intelligence (AI) revolution is driving an intensely competitive race across various facets, from GPUs and ASICs to CPUs, storage, LLM models, and beyond. However, one critical component stands out as the key enabler for AI's future: power.
Simply put, AI cannot exist without the electricity that powers its applications, making energy the chokepoint for AI data centers. Electricity must be generated to keep these data centers running at full capacity, and there are five primary types of fuel sources: coal, natural gas, solar, nuclear and fuel cells.
AI and Data Centers Are Driving Up Power Consumption:
Power consumption is rising with each generation of GPUs. Nvidia’s A100 max power consumption was 250W. Its H100 GPU consumes 350 to 350W and up to 700W with SXM. IO Fund wrote, “Nvidia’s upcoming Blackwell generation boosts power consumption even further, with the B200 consuming up to 1,200W, and the GB200 (which combines two B200 GPUs and one Grace CPU) expected to consume 2,700W. This represents up to a 300% increase in power consumption across one generation of GPUs with AI systems increasing power consumption at a higher rate.”
Join thousands of investors who trust I/O Fund’s expert stock analysis on AI, semiconductors, cryptocurrency, and adtech — sign up for free! Click here!Join thousands of investors who trust I/O Fund’s expert stock analysis on AI, semiconductors, cryptocurrency, and adtech — sign up for free!Click here!
AI data centers need to meet this power demand, which is expected to surge with each generation of GPUs, which are now expected to roll out annually (rather than bi-annually). A single rack is used to consume 10 to 15 kilowatts (KW) of power, but current AI racks are now averaging 80 KW. The next generation is expected to draw 120 KW to 150 KW towards 200 KW within the next few years. This will drive data center consumption by 160% by 2030, where data centers could draw 9% of all the electricity produced in the United States.
Using Thermal Efficiency to Rank Fuel Sources
AI data centers are actively trying to secure long-term power purchase agreements (PPA) and are even exploring nuclear energy options, as evidenced by the 20-year PPA Microsoft signed with Constellation Energy. Hyperscalers are also co-locating data centers near power sources to ensure reliable electricity.
To ensure reliable power, hyperscalers are increasingly seeking long-term power purchase agreements (PPA) and exploring various energy options, including nuclear power. This was underscored by the 20-year PPA that Microsoft signed with Constellation Energy marking the largest-ever PPA in its history. Constellation Energy will launch the Crane Clean Energy Center, restoring the Three Mile Island Unit 1 nuclear reactor, which will add 835 MW of carbon-free energy to the electrical grid.
While electricity is generated from various energy fuels, not all fuels produce the same amount of energy. Thermal efficiency measures how effectively a fuel is converted into electricity. For example, a thermal efficiency of 25% means that 75% of the fuel is lost as heat. Higher thermal efficiency means more energy is converted into electricity, while lower efficiency results in greater energy loss. Let’s see how the energy fuel sources size up.
Premium members receive real-time trade alerts on Nvidia (including 9 we sent under $20 in 2022), as well as deep-dive research on numerous AI stocks and crypto that we track closely. Learn more herehere
Sizing Up the Five Types of Energy Fuel
These are the five energy fuel sources and their general thermal efficiency. These are general thermal efficiencies, which can vary based on location, technology and method:
Solar PV: While solar photovoltaic (PV) panels produce renewable clean energy, they are also the least efficient, averaging 15% to 20%, or 17.5% midpoint. They generate electricity directly from the sunlight. Of course, sunlight is not available 24/7, and efficiencies can drop during overcast and precipitation.
Coal: The world has been phasing out burning coal due to the high levels of carbon emissions. Coal-fired plants burn coal in a boiler, producing steam which flows into a turbine, spinning a generator to produce electricity. The U.S. Energy Information Administration (EIA) suggests 16% of the electricity in the United States is powered by coal. Coal has a thermal efficiency of 33%.
Nuclear: The thermal efficiency of nuclear power plants ranges between 33% to 37%, with a midpoint of 35%. Interestingly, nuclear is not much more efficient than coal as it also uses steam from nuclear fission to spin turbines that generate electricity. However, nuclear plants don’t produce any carbon emissions, making them the cleanest energy option.
Natural Gas: Gas plants produce electricity through two methods. The simple cycle gas turbine works through combustion by burning the natural gas in a combustion chamber to spin a turbine connected to a generator that generates electricity. This alone has a thermal efficiency between 35% to 42%, with a midpoint of 39%. Many gas plants use combined cycle gas turbines (CCGT), which use the exhaust heat up to 1,000 degrees Fahrenheit to boil water in a steam turbine, which adds an extra 20% to 25%, midpoint of 23% efficiency for a total of 62% thermal efficiency. There are carbon emissions generated from burning the gas.
Solid Oxide Fuel Cells (SOFCs): SOFCs can use natural gas, biogas, propane or methane to generate electricity. Natural gas is the preferred method, which produces an electrochemical reaction to produce electricity facilitated by a solid ceramic electrolyte. The thermal efficiency is up to 65%. However, using a combined heat and power (CHP) system will use the exhaust heat of up to 1,500 degrees Fahrenheit to be channeled to a steam turbine, adding an extra 20% to 25% for a total efficiency midpoint of around 87%.
This is the Clear Winner in Thermal Efficiency
As depicted in the chart, SOFC has the highest thermal efficiency when used with heat capture to add an extra 20% to 25%. SOFCs emit carbon when using natural gas as a fuel source, but not as much as gas turbines since there is no combustion, just an electrochemical reaction. For zero carbon, hydrogen can be used as a fuel source. However, it takes electricity to make the hydrogen that will produce electricity, which defeats the purpose for now. Also, while hydroelectric power can have efficiencies as high as 90%, the problem is the location (need to be near large masses of water) and cost (dams are expensive). They don't generate enough electricity (from KWs to hundreds of MWs) to be cost-effective.
As the cost of hydrogen falls and the infrastructure supports it, then it may become a more readily used energy fuel. Natural gas already has the infrastructure and continues to penetrate as a preferred energy fuel. SOFCs are a cleaner way to generate electricity with higher thermal efficiency to boot using natural gas. Investing in the right energy infrastructure is essential to powering the next wave of AI innovation.
Stay on the leading edge of AI with I/O Fund’s high-performing tech portfolio, which had 7 positions outperform the Nasdaq-100 in 2023 for a cumulative return of 131% since inception. In 2024, the I/O Fund has 10 positions beat the indexes, many held at high allocations, and we are prepping for a strong 2025. Join our free newsletter here or subscribe to Advanced Market Signals for weekly webinars, entries, exits and real-time trade alerts.free newsletter here or subscribe to Advanced Market Signals for weekly webinars, entries, exits and real-time trade alerts.
Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.
TSMC released its monthly revenue for February on March 10th. Revenue grew by 43.1% YoY and down (-11.3%) MoM to NT$260.01 billion. February 2025 revenue was the new high for the month. In U.S. dollar terms revenue grew by 37.3% YoY to $7.93 billion using the average exchange rate of 1 US dollar to 32.78 NT dollars. This strong performance, coupled with Foxconn's impressive 56.4% year-over-year monthly revenue growth in February, suggests robust and sustained demand for AI.
A closer look at TSMC's monthly revenue reveals that month-over-month figures can be volatile. The February decline is likely attributable to seasonal factors, such as the Lunar New Year holidays and fewer working days. For context, February 2024 also saw a MoM revenue decrease of (-15.8%). This suggests that the recent month-over-month decline is not unusual and should be considered within the context of seasonal trends.
The management had provided an update last month that they expect the Q1 revenue to be near the lower end of the guidance range of $25 billion and $25.8 billion due to the Taiwan earthquake in January. Importantly, TSMC maintains its strong outlook for the full year 2025, anticipating revenue growth in the mid-20% range in US dollar terms, driven by robust AI demand.
Sustained sequential HPC growth
As the leading foundry for AI accelerators, TSMC is riding the enormous wave of demand from Big Tech. The chipmaker’s high-performance computing (HPC) revenues rose 19% QoQ to a record $14.25 billion and accounted for 53% of revenue in Q4, surpassing the 50% mark for the third time. The sequential HPC growth for the six consecutive quarters is a further testament that there is no AI demand slowdown and demonstrates sustained momentum in the AI sector. Management expects AI accelerators to be the strongest driver of the HPC platform growth in the next several years.
The above chart shows that TSMC’s HPC sequential revenue growth tells us that they're a few quarters ahead of Nvidia's bigger quarters like the 2022 sequential growth before the launch of Hopper Architecture.
TSMC is experiencing explosive growth in its AI segment, with revenue tripling in 2024. Management expects this remarkable growth to continue, projecting a further doubling of AI revenue in 2025.
Management expects AI accelerators to grow mid-40% CAGR for the next five years and expects AI accelerators to be the strongest driver of the HPC platform growth and the largest contributor in terms of the overall incremental revenue growth in the next several years.
The chart below further emphasizes the strength of TSMC's high-performance computing (HPC) segment, with revenue reaching a record $14.25 billion in the most recent quarter. This represents the largest sequential increase to date, surging by approximately $2.26 billion. This data underscores the significant growth trajectory of TSMC's HPC business, driven by robust demand for AI accelerators.
TSMC's Advanced Packaging in High Demand: NVIDIA Leads the Charge
TSMC also reported a surge in Advanced Packaging due to the strong demand for Nvidia’s Blackwell chips. NVIDIA has reportedly secured over 70% of TSMC's CoWoS-L capacity for 2025, with shipments expected to exceed 2 million units and grow by more than 20% each quarter, according to a report from Economic Daily News. The report also estimates that advanced packaging revenue accounted for approximately 8% of TSMC’s revenue in 2024 and is expected to exceed 10% in 2025.
TSMC, which is struggling to meet the strong demand for advanced packaging, plans to double production capacity this year to 75,000 wafers a month, according to a report from Taiwan Economic Daily. Furthermore, TSMC is projected to continue expanding CoWoS production in the coming year, reaching 90,000 wafers per month in 2026.
Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.
TSMC released its monthly revenue for February on March 10th. Revenue grew by 43.1% YoY and down (-11.3%) MoM to NT$260.01 billion. February 2025 revenue was the new high for the month. In U.S. dollar terms revenue grew by 37.3% YoY to $7.93 billion using the average exchange rate of 1 US dollar to 32.78 NT dollars. This strong performance, coupled with Foxconn's impressive 56.4% year-over-year monthly revenue growth in February, suggests robust and sustained demand for AI.
A closer look at TSMC's monthly revenue reveals that month-over-month figures can be volatile. The February decline is likely attributable to seasonal factors, such as the Lunar New Year holidays and fewer working days. For context, February 2024 also saw a MoM revenue decrease of (-15.8%). This suggests that the recent month-over-month decline is not unusual and should be considered within the context of seasonal trends.
The management had provided an update last month that they expect the Q1 revenue to be near the lower end of the guidance range of $25 billion and $25.8 billion due to the Taiwan earthquake in January. Importantly, TSMC maintains its strong outlook for the full year 2025, anticipating revenue growth in the mid-20% range in US dollar terms, driven by robust AI demand.
Sustained sequential HPC growth
As the leading foundry for AI accelerators, TSMC is riding the enormous wave of demand from Big Tech. The chipmaker’s high-performance computing (HPC) revenues rose 19% QoQ to a record $14.25 billion and accounted for 53% of revenue in Q4, surpassing the 50% mark for the third time. The sequential HPC growth for the six consecutive quarters is a further testament that there is no AI demand slowdown and demonstrates sustained momentum in the AI sector. Management expects AI accelerators to be the strongest driver of the HPC platform growth in the next several years.
The above chart shows that TSMC’s HPC sequential revenue growth tells us that they're a few quarters ahead of Nvidia's bigger quarters like the 2022 sequential growth before the launch of Hopper Architecture.
TSMC is experiencing explosive growth in its AI segment, with revenue tripling in 2024. Management expects this remarkable growth to continue, projecting a further doubling of AI revenue in 2025.
Management expects AI accelerators to grow mid-40% CAGR for the next five years and expects AI accelerators to be the strongest driver of the HPC platform growth and the largest contributor in terms of the overall incremental revenue growth in the next several years.
The chart below further emphasizes the strength of TSMC's high-performance computing (HPC) segment, with revenue reaching a record $14.25 billion in the most recent quarter. This represents the largest sequential increase to date, surging by approximately $2.26 billion. This data underscores the significant growth trajectory of TSMC's HPC business, driven by robust demand for AI accelerators.
TSMC's Advanced Packaging in High Demand: NVIDIA Leads the Charge
TSMC also reported a surge in Advanced Packaging due to the strong demand for Nvidia’s Blackwell chips. NVIDIA has reportedly secured over 70% of TSMC's CoWoS-L capacity for 2025, with shipments expected to exceed 2 million units and grow by more than 20% each quarter, according to a report from Economic Daily News. The report also estimates that advanced packaging revenue accounted for approximately 8% of TSMC’s revenue in 2024 and is expected to exceed 10% in 2025.
TSMC, which is struggling to meet the strong demand for advanced packaging, plans to double production capacity this year to 75,000 wafers a month, according to a report from Taiwan Economic Daily. Furthermore, TSMC is projected to continue expanding CoWoS production in the coming year, reaching 90,000 wafers per month in 2026.
The I/O Fund has a live portfolio of 10 to 15 positions. This portfolio is the culmination of all analysis by a team of analysts led by the Lead Tech Analyst Beth Kindig, who frequently appears on all Tier-1 media. Our highest convictions can be found listed by percentage of allocation. To view our entire portfolio, upgrade to Advanced Market Signals. Learn more here.
Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.
After posting its first annual decline in deliveries in 2024, Tesla continues to face major hurdles to growth in 2025. There are shockingly large declines in Europe and China so far this year, coupled with automotive margins that hit a low in Q4 with more margin pressure likely in Q1. Management has also quietly shifted its tone on 20% to 30% delivery growth, with other segments offering no reprieve as Tesla continues to eat into its gross profit to push deliveries higher.
While optimism has risen for robotaxi services and Optimus robots, neither of the two look to be major drivers of growth in 2025, with initial use cases likely to be internal or small-scale in nature. Tesla continues to hype up a more affordable model, though questions remain about its ability to do so profitably as Tesla has made more progress cutting selling prices than it has cutting production costs.
Stay on the leading edge of AI with I/O Fund’s high-performing tech portfolio, which had 10 positions outperform the Nasdaq-100 in 2024, many held at high allocations, and we are prepping for a strong 2025. Take advantage of our limited-time monthly promotion for up to 20% off our Pro service here to access the I/O Fund’s portfolio, library of research, webinars, and more. here to access the I/O Fund’s portfolio, library of research, webinars, and more.
Tesla’s Automotive Growth Stagnates in 2024
Automotive growth stalled in 2024, with Tesla recording a (1%) YoY decline in deliveries for the year to 1.79 million vehicles. Q4 capped off the end to a rather tumultuous year for Tesla, as Tesla sold down a significant amount of inventory after Q1 2024 saw its first YoY decline in quarterly deliveries in four years.
Tesla stock sees its first annual decline in deliveries as growth stagnated in 2024.
China was a strong area of growth for Tesla in 2024 and in Q4, with Tesla’s deliveries in China reaching a record high of 82,927 vehicles in December and 196,902 vehicles in Q4, up nearly 16% YoY and also marking a fresh record. For 2024, China deliveries exceeded 657,000, rising 8.8% YoY. As a result, China accounted for 39.7% of global deliveries in Q4 and 36.7% in 2024, up more than 3 points from 33.4% in 2023.
For 2025, Musk had estimated vehicle deliveries could grow 20% to 30% YoY in Q3, which would correlate to deliveries between approximately 2.15 million and 2.33 million, or an average of at least 550,000 deliveries per quarter. Interestingly, this target was not repeated in Q4, with Tesla saying now that it only expects to “return to growth” in 2025. This subtle shift in tone is easy to miss, but it suggests that Tesla may be on track for 1.85 to 2.0 million vehicles this year, technically returning to growth but at a much lower rate than prior commentary.
This year looks to be off to a challenging start, with early data from across Europe showing plunging sales, while China sales accelerated their decline in February to notch a (29%) YoY drop for the first two months of the year.
ACEA data showed that Tesla registrations fell 45% YoY in January 2025 in the EU, Iceland, Liechtenstein, Norway, Switzerland and the UK, reaching a two-year low, despite broader EV sales rising 37%. In February 2025, registration data showed declines of (42%) to (48%) YoY in Scandinavia and France, while Germany saw sales decline a whopping (76%) YoY after falling (60%) in January.
Tesla stock is facing a tough road ahead in China as sales have slumped to start 2025, with February plunging -49% YoY.
China sales dropped (12%) YoY in January, but accelerated this decline in February, with preliminary data from the CPCA showing sales down (49%) YoY to 30,688 vehicles, the lowest monthly volumes in the country since August 2022. This also marks a (51%) MoM plunge from January’s 63,238 vehicles, and a more than (67%) plunge from December 2024. For comparison, rival BYD’s February sales surged 161% YoY for BEV and PHEVs, with global sales up 56% YoY in the first two months. CPCA data also showed the NEV market rose 82% YoY in February, with Tesla lagging the market by a 131 point difference. Tesla is now offering an 8,000 yuan (~$1,100) insurance subsidy on Model 3 vehicles in China in an effort to revive demand.
Given this weakness already in Q1 in core regions, there are whispers that deliveries could fall to significantly below 400,000. There are mounting indications that Tesla is facing a demand problem, not only within plunging sales across multiple markets worldwide, but also in more aggressive financing perks. Tesla recently launched new financing and free lifetime supercharging perks this week to boost demand, offering 0% APR or zero due at signing for Model 3s and discounts on older Model Ys.
If Q1 does come in weak due to global sales weakness and transitionary impacts to production from the refreshed Model Y, Tesla will have to make up substantial ground in the back half of the year to reach its optimistic targets, as Musk’s prior 20% to 30% volume growth target would already be pushing the upper limits of Tesla’s installed manufacturing capacity.
Tesla’s Margins Have Faced Significant Pressure
Tesla has prioritized affordability to drive growth in delivery volumes and prevent inventory build-ups through 2024, with this coming at the expense of margins. In Q4, CFO Vaibhav Taneja reaffirmed these priorities, saying Tesla is still committed to reducing inventory and vehicle production costs – as expected, this came at quite a cost to margins. Management has also discussed for multiple quarters the plan to launch more affordable models in the first half of 2025, but there’s limited evidence that Tesla can do so in a margin-accretive way that quickly.
Taneja explained in Q4’s earnings call that Tesla was “able to get our overall cost per car down below $35,000, primarily by material costs,” despite increased depreciation as Tesla transitions to its refreshed Model Y. Calculations show that COGS per vehicle declined just (1.1%) sequentially in Q4 to ~$34,716, or a reduction of $390 from Q3.
Tesla’s actions to aggressively sell down inventories, which declined nearly $2.5 billion sequentially in Q4 to $12 billion, were possibly due to “attractive financing options but also other discounts and programs which impacted ASPs.” As a result, ASPs fell (5.2%) sequentially to $39,818 in Q4, a decline of approximately ($2,174) from Q3. This was the largest QoQ decline in ASPs since Q1 2023. Essentially, Tesla reduced production costs at less than 1/5th the rate of ASPs in the quarter.
Tesla’s average selling prices declined more than 5% QoQ in Q4 2024, while production costs declined just 1.1% QoQ.
Over the last three years, ASPs have been declining at a faster rate than COGS, pressuring margins quite substantially in the process. Tesla said that it saw a new record for deliveries in the highly competitive Chinese market, which (as we have discussed in our analyses Tesla Sells 33% Of Vehicles Below Average Cost, BYD Pulls Ahead in November 2023 and Tesla’s China Market Share Continues To Slide in December 2023) are detrimental to ASPs and likely a factor in the larger QoQ decline.
Putting this all together, automotive gross margin dropped more than 3.5 points sequentially to 13.59% in Q4, more than a full point below Q2’s 14.65% margin. This was visible within Tesla’s growth rates – automotive revenues declined (8%) YoY and (1%) QoQ despite a 2% YoY and 7% QoQ increase in deliveries. This was also mostly expected given management had stated that sustaining margins in Q4 would be challenging.
Tesla stock witnessed automotive gross margin (excl. regulatory credits) fall to a fresh low at 13.59% in Q4 2024.
On a per-vehicle basis, Tesla’s average gross profit was ~$5,102 in Q4, down more than (29%) YoY and (26%) QoQ due to the sharper decline in ASP. This is a far cry from the $14,000+ gross profit per vehicle Tesla was recording in late 2021 and early 2022.
As it stands, Tesla risks its per-vehicle gross profit falling below $5,000 in 2025 unless it can quickly drive production costs below $34,000 per vehicle or reverse its decline in ASPs. The aforementioned 0% APR financing perks and other promotional discounts are likely to weigh on ASPs, as was the case in Q4.
Margin Issues to Persist in Q1
Tesla has outlined more headwinds in the first half of 2025, and energy storage has been unable to offset automotive weakness recently, facing similar growth headwinds in Q4 – revenue and gross profit increased just 1.5% and 0.7% from Q2 despite deployments being more than 17% higher.
For Q1, there’s not likely to be much relief on the margin front, as management said that production of the refreshed Model Y kicking off in February will “result in several weeks of lost production” in Q1, and that “margins will be impacted due to idle capacity and other ramp related costs” that will ease once production is ramped. Energy storage is also likely to see margin pressure in Q1, with Shanghai production set to ramp with both Powerwall and Megapack remaining supply constrained.
Sign up for I/O Fund's free newsletter with gains of up to 2,250% because of Nvidia's epic run – Click hereSign up for I/O Fund's free newsletter with gains of up to 2,250% because of Nvidia's epic run – Click hereClick here
Questions also remain about Tesla’s ability to launch and ramp a more affordable model as promised this year. Tesla has stated for multiple quarters now that it remains on track to launch this vehicle in the first half of 2025, with it being a cornerstone to the previous 20-30% delivery growth forecast given management’s intense focus on improving affordability for customers to drive delivery growth in 2024. Tesla can tout Q4’s production costs as the lowest on record, but the bigger picture shows that Tesla has made very little progress in actually reducing production costs over the past three years. In fact, production costs have declined just (5%), or ~$1,830, since the end of 2021, or a little more than a $150 reduction per quarter on average.
If Tesla’s goal is to make an affordable model at a $25,000 price point and make it profitable at scale, production costs would need to be more than 30% lower than current levels. A 30% reduction in production costs from Q4’s level would equal ~$24,300, or a gross margin of under 3%. To produce a $25,000 vehicle at a ~15% margin, production costs would need to come down to ~$21,750, or nearly 40% below its average cost. Simply reshuffling logistics scheduling to reduce quarter-end weighting of deliveries or relying on materials costs coming down in the face of tariffs is not enough.
And if this is truly something that is feasible to do within the year, it begs the question, why hasn’t Tesla done this yet? There is little evidence that Tesla can flip a switch and bring to market a sub-$30,000 or $25,000 vehicle in the first half of the year without significant damage to margins.
The Bigger Picture at Play for Tesla Stock is Eroding Earnings
While I have heard numerous times that discussions on margins are short-sighted and that the bigger picture for Tesla is the long-awaited autonomous driving and robotics growth curve, I want to make clear that margins have led to a significant erosion in Tesla’s earnings power over the past few years.
At the beginning of 2023, Tesla was expected to earn $8.50 in earnings per share in 2025. At that time, automotive gross margin had actually contracted 5 points YoY, down from 29.2% in Q4 2021 to 24.3% in Q4 2023. To put it another way, Tesla was generating more in automotive gross profit at half the scale.
In Q3 2021, Tesla generated $3.67 billion in automotive gross profit, or $3.24 billion excluding leasing and regulatory credits, with deliveries of 241,391. In Q4 2024, Tesla generated $3.29 billion in automotive gross profit, or $2.39 billion excluding leasing and regulatory credits. This gross profit and margin erosion is why adjusted EPS peaked at $4.07 in 2022 and has since dropped to $2.42 in 2024.
Now, 2025’s EPS forecast stands at just $2.85 at the beginning of March, more than (66%) lower than the estimate from two years ago. It’s also a rather sharp decline this year, down more than (12%) from $3.25 in mid-January.
Tesla stock’s EPS estimates have fallen -12% so far in 2025. Source: YCharts
This has been primarily caused by margin contraction and lower automotive gross profit, which have dragged operating margin much lower. Operating margin peaked at 16.8% in 2022, before contracting to 9.2% in 2023 and now to 7.2% in 2024, with Q4’s operating margin at 6.2%.
Robotaxis, Optimus Inconsequential to Tesla’s Growth in 2025
Robotaxis and Tesla’s Optimus humanoid robots are two core anchors for a majority of the multi-trillion-dollar valuation thesis, with Musk throwing out in Q4’s call that Optimus has “the potential to be north of $10 trillion in revenue.” However, for 2025, even if both are achieved, they’re likely to be only for internal use and not ready for commercialization in a way that will contribute to growth.
Tesla says that it is planning to launch unsupervised FSD as a paid service in Austin, Texas in June, with Tesla’s fleet testing the service out at its factory, from the end of the production line to the destination pick-up parking spot. Musk explained that Tesla is aiming to have unsupervised services with its internal fleet in multiple cities by the end of 2025. Tesla is also seeking the first permit to pave the way for an approval for robotaxi services in California.
However, given that FSD is still Supervised for consumers, analysts had questions about the progress Tesla is making on reaching full autonomy (ie. hands-off, eyes-off), with management explaining that it is close but not there yet:
“We need to be very confident that the probability of injury is low before we allow people to check with their email and text messages. … We're in this perverse situation where people will turn the car off autopilot so the computer doesn't yell at them, check the text messages while steering the car with their knee and not looking out the window. … If you have any problems with the system and when people are not looking, that is a dangerous thing. And that's what we're trying to avoid. The capability is getting there, but it's not fully there.”
Though Tesla laid forth that goal to have a robotaxi service running as soon as this summer, and discussed its ramp profile in Q3, the company provided no major update in Q4 on the Cybercab, its purpose-built robotaxi. Tesla said that it would be aiming for volume production in 2026 with a ramp to at least 2 million vehicles per year, potentially as high as 4 million in the future.
For Optimus, Musk threw out a rather sensational $10 trillion revenue figure, though he shared more details about the robot and Tesla’s production projections. Musk explained that the current production line Tesla is “designing is for roughly 1,000 units a month of Optimus robots. The next line would be for 10,000 units a month. The line after that would be for 100,000 units a month.” He added that Tesla will “probably not” succeed in manufacturing 10,000 in 2025, which is what its internal plan targets, but will aim to ramp production significantly faster than its automotive side.
Musk also predicted that it would not “take very many years before we're making 100 million of these things a year,” though initial use cases for Optimus will be inside Tesla factories, with commercial sales not expected until at least the second half of 2026. If anything, scaling production of either robotaxis or Optimus this year will likely add to costs and impact the bottom line.
Note on Tesla’s Capex
What’s interesting is that even with the $10 trillion revenue potential (and 100 million production numbers) for Optimus and 2 million for Cybercab, Tesla’s capex is not expected to meaningfully increase from 2024’s levels over the next three years. For 2025, 2026 and 2027, Tesla said that it expects capex to be at least $11 billion, compared with $11.34 billion in 2024. Musk had said in Q4 that Optimus’ AI training needs are “probably at least ultimately 10x of what's needed for the car,” suggesting significant amounts of compute would be needed.
In 2024, Tesla’s AI infrastructure assets rose $3.6 billion YoY to $5.15 billion, and it is still working on developing FSD. If you have to scale the compute factor by 10x, while ensuring plants can handle manufacturing millions of Optimus robots, millions of Cybercabs, existing models and a new affordable model, maintaining capex at or above $11 billion annually does not seem sufficient given that Tesla is just now approaching a 2 million vehicle scale annually after having spent close to $50 billion over the past decade.
Conclusion
Musk has been quite vocal about 2025 being Tesla’s “most pivotal year” and possibly the “most important” in its history. Despite strumming up optimism for Optimus and autonomous driving advancements this year, the growth story looks challenged with data for January and February showing substantial YoY declines across Europe and China.
Management’s commentary saw a subtle change from 20% to 30% growth in Q3 for deliveries to now only a return to growth mentioned in Q4, suggesting Tesla is also tempering expectations for deliveries in 2025.
2025’s EPS estimates are already dropping, falling more than 12% since the beginning of the year to $2.85, while revenue estimates have already been revised $4.3 billion lower to $112 billion. New products such as robotaxis and Optimus are unlikely to be growth drivers in 2025, with initial use cases likely limited to small scale and internal operations.
Margins came under more intense pressure in Q4, and are facing even more headwinds from idle capacity in Q1 as Tesla paused production in order to ramp up the refreshed Model Y, while Energy storage provided no relief despite a jump to record deployments in Q4. As Tesla is aggressively pushing for better affordability, it’s putting automotive gross margin at risk of falling into the single-digit range.
Price often bottoms before fundamentals, which means Tesla may be in a buy zone soon. Find out what potential entries the I/O Fund is watching for Tesla and other AI stocks in our upcoming webinar with Portfolio Manager Knox Ridley on Thursday, March 13 at 4:30pm EST. Learn more here.
I/O Fund Equity Analyst Damien Robbins contributed to this report.
Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.
Broadcom reported a strong Q1, with revenue topping estimates by $330 million as management noted that AI revenue rose 77% YoY in the quarter. AI revenue reached $4.1 billion this quarter, beating previous guidance of $3.8 billion. The beat was driven by strength in networking with management mentioning last quarter that custom silicon would see stronger growth in H2. It was also stated networking was 40% of AI revenue while it’s typically 30% of AI revenue (confirming the higher product mix this quarter).
The stock was up double-digits after hours specifically when Broadcom stated next quarter AI revenue would be $4.4 billion and when management reiterated their serviceable addressable market forecast of $60 billion to $90 billion, stating “these R&D investments are very aligned with the roadmap of our three hyperscale customers as they each race towards 1 million XPU clusters by the end of 2027. And accordingly, we do reaffirm what we said last quarter, that we expect these three hyperscale customers will generate a Serviceable Addressable Market or SAM in the range of $60 billion to $90 billion in fiscal 2027.”
Managment also mentioned there are two more hyperscalers that plan “to tape out their XPUs” this year, referencing future expansion of the quoted SAM.
We’ve seen many AI stocks with similarly strong reports sell off this quarter, whereas the market loves Broadcom’s strong margins, its cash and the clarity management provides in terms of exact AI revenue per quarter.
Revenue
Broadcom’s revenue increased 24.7% YoY to $14.92 billion in its fiscal Q1, ahead of estimates for $14.59 billion due to strong AI momentum. Management guided for Q2 revenue to be flat sequentially at $14.9 billion, ahead of estimates for $14.73 billion.
Looking ahead, growth is expected to decelerate to 19.3% in Q2, a more than 5 point deceleration from 24.7% growth in Q1. Broadcom is lapping comps that include VMWare’s acquisition, hence why there is a large QoQ deceleration in Q1.
Most importantly, AI revenue rose 77% YoY and ~11% QoQ to $4.1 billion. For Q2, Broadcom said that it expects AI momentum to continue in Q2 “as hyperscale partners continue to invest in AI XPUs and connectivity solutions for AI data centers,” forecasting AI revenue to be $4.4 billion, up just over 7% QoQ and 44% YoY. This would put AI revenue at $8.5 billion for the first half of the fiscal year, up more than 57% from $5.3 billion in the same period last year.
Key Segments
Semiconductor Solutions revenue rose 11% YoY and was approximately flat QoQ at $8.22 billion, decelerating 1 point from 12% YoY growth in Q4.
AI revenue was $4.1 billion, ahead of management’s guidance for $3.8 billion on “stronger shipments of networking solutions to hyperscalers on AI.” Management added that they “see a steady ramp in deployment of our XPUs and networking products,” supporting its AI revenue guide of $4.4 billion in Q2.
Non-AI semiconductor revenue was $4.1 billion, down 9% QoQ due to a seasonal decline in wireless revenue.
Broadband “showed a double-digit sequential recovery in Q1 and is expected to be up similarly in Q2 as service providers and telcos step up spending.”
Server storage “was down single-digits sequentially in Q1, but is expected to be up high-single digits sequentially in Q2.”
Enterprise networking is expected to remain flattish in Q1 and Q2 as customers work through inventory.
Wireless was flat in Q1 and expected to be flat YoY in Q2, and resales in industrial were “down double-digits in Q1 and are expected to be down in Q2.”
Infrastructure Solutions revenue rose 47% YoY and 15% QoQ to $6.70 billion. Management said the QoQ performance was “exaggerated though by deals which slipped” from Q4 to Q1, though the strong YoY performance was driven by two factors – converting largely perpetual licenses to one full subscription, and “upselling customers to a full stack VCF, which enables the entire data center to be virtualized.” At the end of Q1, approximately 70% of Broadcom’s 10,000 largest customers have adopted VCF.
For Q2, Infrastructure Software revenue is expected to be $6.5 billion, up 23% YoY.
Margins
Margins expanded significantly across the board in Q1, with GAAP operating margin seeing the strongest expansion on a QoQ basis of 9.1 points.
Gross margin was 68.0% in Q1, up from 64.1% in the prior quarter as gross profit topped $10 billion for the first time. Adjusted gross margin was 79.1%, expanding from 76.9% in the prior quarter.
Operating margin witnessed the strongest QoQ expansion, with Broadcom reporting a 42.0% margin in Q1, up from 32.9% in Q4; this is marking a return to pre-VMWare acquisition levels. Adjusted operating margin was 65.9%, up from 62.7% in the prior quarter.
Net margin was 36.9% in Q1, a strong 6.1 point expansion from 30.8% in the prior quarter. Adjusted net margin was 52.4%, up from 49.6% in the prior quarter.
Adjusted EBITDA margin also expanded nearly 3 points sequentially to 67.6%, with adjusted EBITDA surpassing $10 billion for the first time in Q1 at $10.08 billion. This also marked a strong, consistent expansion from the high-59% range at the start of FY24 when VMWare’s integration was impacting margins. Q2’s adjusted EBITDA margin was guided at 66%, a slight sequential contraction.
EPS
Given the strong margin expansion in Q1, Broadcom delivered a strong GAAP EPS beat of 35.7%, while its adjusted EPS beat was smaller in nature at ~6%.
GAAP EPS of $1.14 beat estimates for $0.84.
Adjusted EPS of $1.60 beat estimates for $1.51, for growth of 45.5% YoY. This is expected to be peak growth for adjusted EPS in fiscal 2025, with estimates pointing to 37% growth in Q2 below ending the fiscal year at almost 23% growth.
Cash and Balance Sheet
Operating and free cash flow were both strong in Q1 due to the margin strength and revenue outperformance, with both recording a margin of >40%.
Operating cash flow was $6.11 billion, up 9% QoQ. OCF margin was 41%, improving from 39.9% last quarter.
Free cash flow was $6.01 billion, up almost 10% QoQ. FCF margin was 40.3%, improving from 39.0% last quarter.
Inventories were $1.91 billion, increasing more than 8% QoQ.
Cash and equivalents totaled $9.31 billion, while debt decreased $1 billion QoQ to $66.58 billion.
Earnings Call:
Market Opportunity:
The market opportunity for Broadcom is quite large – represented by SAM of $75B at the midpoint by 2028, up from a $16B run rate right now. An analyst asked a similar question in terms of chips and management reiterated their current forecast is with only three customers right now whereas two more are likely to go into volume production in the coming years.
Timothy Arcuri:
Thanks a lot. Hock, in the past, you have mentioned XPU units growing from about 2 million last year to about 7 million you said in the 2027, 2028 timeframe. My question is, do these four new customers, do they add to that 7 million unit number? I know in the past, you've sort of talked about an ASP of 20 grand by then. So those — the first three customers are clearly a subset of that 7 million units. So do these new four engagements drive that 7 higher, or do they just fill in to get to that 7 million? Thanks.
Hock Tan:
And thanks, Tim for asking that. To clarify, as I made — I thought I made it clear in my comments. No, the market we are talking about, including — when you translate the unit is only among the three customers we have today. The other four, we talk about engagement partners. We don't consider that as customers yet and, therefore, are not in a served available market.
China Exposure:
Management was asked how many of the hyperscalers were from China, but they declined to comment. This is a notable question given any hyperscaler from that region has a low chance of being unscathed over the next four years. There were reports earlier this week that Broadcom may be losing major customer ByteDance, for example.
Management stated “no comment” yet it’s important to note the concern.
Conclusion:
Broadcom is quality and will help us hedge any anti-Nvidia narratives – which are bound to come up from time to time. As far as quality goes, you cannot find a better pair in AI (NVDA-AVGO). These are the juggernauts and we want exposure to AVGO as its empire slowly expands. Don’t forget AVGO’s AI software opportunity, which we’ve expanded on in the past.
That’s officially a wrap for the I/O Fund earnings season!
Regarding the AI trade falling out of favor this quarter, don’t let the market fool you for one minute – investors have never had it so good as to be able to track and verify there are hundreds of billionshundreds of billions pouring into one trend. It’ll be volatile, and it’ll be scary – and then ultimately thrilling, but this trend is bursting at the seams with demand. Supply is not merely bottlenecked; it’s dammed at the flood gates. The supply issues will work themselves out, and the I/O Fund and our Members will be well-positioned when those flood gates open.
Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.
Bloom Energy delivered solid Q4 2024 results, culminating in a top-line beat of $64.85 million and a bottom-line beat of $0.12 per share. Revenue surged 60.4% YoY in Q4 to close out the year with positive free cash flow for the first time since 2019. Adjusted operating margin tripled from last year. 2024 laid out the foundation for deals and revenue to accelerate in 2025. Notably, the game-changer deal with AEP was a large driver for Q4.
With that said, energy-related stocks have lumpy revenue, lumpy margins and lumpy cash. Admittedly, fundamental analysis is challenging in this sector as there are often government subsidies to also consider, deal sizes can be enormous yet infrequent, and things change suddenly depending on macroeconomics. Like crypto, due to being absent of reliable fundamentals, technicals must be in the driver's seat for energy-related stocks, and investors should be prepared to have an active management stance.
The “Game Changer” AEP Deal up to 1 GW of Bloom Fuels Cells
According to Bloom, the company has key advantages by offering solid oxide fuel cells compared to other power solutions such as quick time to deployment, reliability, not needing backups, electrochemical (non-combustion) electricity generation, low to no carbon emissions and easy to obtain air permits.
The company also offers modular fault tolerant architecture, which we covered in the past stating: “BES uses core 325 kW base blocks customized to work in parallel with the local electric grid in standby or backup mode to kick in when the local power becomes unavailable. The 325 kW base blocks can be duplicated and scaled up to multiple MWs for any project. They can also be used as the primary power source. They can be used off-grid or parallel as a microgrid. BES has a high density compared to solar or wind of 100 MW per acre with features such as stackable servers and combined heat and power solutions. However, nuclear is in the kilowatts per acre, and therefore, is by far the highest density power solution.” Notably, Bloom Energy does not foresee nuclear being a true competitor until 2030: “I really don't think you're going to move the needle, between now and another eight years with nuclear.”
Due to these benefits, utility companies like AEP are contracting Bloom Energy to help serve the outsized power demands that AI data centers requires. We pointed out in our earlier coverage of Bloom Energy, the significant game-changer deal with American Electric Power (AEP) to “secure up to 1 GW of Bloom Energy SOFC for their data center customers and other larger energy users. AEP placed an order for the installation of 100 MW of fuel cells at customer sites, with further expansion orders expected in 2025.
AEP expects commercial load to grow 20% annually over the next three years, driven by data center development. The company is in the process of finalizing the first customer project agreements, and discussions are ongoing with several other customers. AEP’s hyperscaler customers include Google, Amazon, Microsoft and Meta Platforms.
AEP will purchase Solid Oxide Fuel Cells (SOFCs) from Bloom Energy with an initial order of 100 MW, with more expected in 2025, and integrate them into their customer energy systems, prioritizing AI data centers. Large customers will cover all costs for the fuel cell projects under a special contract. AEP will oversee deployment and installation at customer sites.
Bloom will supply the SOFCs to AEP, providing the core technology for on-site power generation. Bloom will likely offer ongoing support and maintenance services. “
AEP’s initial order of 100 MW at $3 per watt would equate to $300 million for the product hardware and $600 million of maintenance services over the following 20 years. The potential product sales for 1GW would equate to $3 billion with potential service revenues of $6 billion over 20 years. The potential total contract value of AEP’s 1 GW deal would be $9 billion over 20 years including products and services.
Bloom is in Talks with Other Utilities for Similar Deals to AEP
The company is using the AEP deal as a template for more deals with utility companies, who can provide Bloom’s solution for their customers. During the Q4 2024 conference call, Morgan Stanley analyst Andrew Percoco asked if they expect more deals like AEP’s with a large utility serving data center customers rather than directly to the data center.
CEO Shrider answered that Bloom Energy is not competing with the utilities but providing them with the ability to service their hyperscaler customers. He said this.
“This is KR and again on your first question, yes the answer is we are talking to several utilities who are interested in some kind of arrangements along the lines of what we announced with AEP. And it is A, they are realizing that no matter how fast they augment their transmission distribution system, no matter where generation happens or not, getting the power to the end customer, between now and 2030, is going to be a big issue unless you produce power where you need it.”
Revenue Rises to Record Levels in Q4 to Close Out a Record Full Year 2024
On Jan 12, 2025, the I/O Fund pointed out that Bloom Energy would need to have a big Q4 with revenue growing 67.8% YoY towards $598.5 million to achieve management’s mid-point guidance, which would imply a large chuck of product revenue being pushed out into Q4. Bloom Energy delivered close enough with 60.4% YoY revenue growth to $572.39 million, firmly beating the $507.54 million by $64.85M. The revenue spike was driven by collecting large receivables from their related party, SK ecoplant.
Q4 revenue grew by 60.4% YoY and 73.2% QoQ to $572.39 million, compared to the Q3 revenue drop of (17.5%) YoY and (1.6%) QoQ to $330.4 million. Nearly 40% of 2024 total revenue came in Q4. The Company received its large related party, SK ecoplant, receivable in the quarter. Bloom Energy also assisted them in selling a majority of the 73MW of energy servers they held as part of a delayed project.
Q4 revenue of $572.39507.54 million beat consensus estimates of $507.54 million by 12.78%, compared to Q3 revenue of $330.4 million, missing consensus estimates for $383.19 million by 13.78%.
Q4 Product and Service revenue surged 67.2% YoY to $525.5 million, compared to Q4 2023 Product and Service revenue of $314.4 million.
Full-year 2024 revenue rose 10.5% YoY to $1.474 billion. Product and Service revenues rose 12.1% YoY to $1.158 billion. Management provided full-year 2025 revenue guidance of $1.65 billion to $1.85 billion, midpoint is $1.75 billion for 19.1% YoY growth.
Key Metrics and Backlog
Bloom Energy closed out 2024 with a total backlog of $11.5B, down (4.96%) YoY, comprised of Product backlog of $2.5B, down (16.67%) YoY and Service backlog of $11.5B, down (4.96%) YoY. CFO Berenbaum pointed out that due to the supply agreement with SK ecoplant, the Product backlog would have been 30% higher, by $900 million, to $3.9 billion. This indicates the Company is seeing more demand for its products.
“We have $2.5 billion of product backlog. Excluding dynamics around our supply agreement with SK ecoplant, our product backlog would have been up roughly 30% year-over-year. As a reminder, at the end of 2023, we had extended the term of our SK ecoplant distribution agreement to the end of 2027, and increased their purchase commitment to 500 megawatts, all of which was included in our year ending 2023 backlog.”
Dropping Product and MW Accepted Metrics as of Q4
CFO Berenbaum noted on the Q4 2024 conference call that the Company is shifting focus away from earlier used metrics that measure kW shipped (IE: Products Accepted and MW Accepted).
Business has “evolved” as Product revenue is now recognized upon shipments, not acceptance as in the past based on grid-connected baseload power or timing of revenue recognition. The Company is now offering a broader range of solutions ranging from microgrids, carbon capture, AI data center solutions to combined heat and power systems, which provide more complex value to customers. Solutions go beyond just energy production and there the focus has shifted to traditional metrics like revenue growth, non-GAAP gross and operating margins, and cash flow from operations. Berenbaum said this.
“As we've discussed over the past few quarters, management is primarily focused on overall revenue, product revenue growth, non-GAAP gross and operating margin and cash flow from operations. In the past when we were primarily shipping grid connected, baseload power, and timing of product revenue recognition was somewhat divorced from timing of product shipments.”
Margins Consistently Expand in 2024
Q4 gross margin was 38.3%, up 47.8% YoY and 60.9% QoQ. This was an improvement over Q3 2024 gross margin of 23.8% versus gross margin of (1.3%) in Q3 2023. Gross margin improved on a sequential basis for Q3 gross margin improving 16.7% QoQ and Q2 gross margin improving 25.9% QoQ. Q1 gross margin fell by -37.4% QoQ from Q4 2023 gross margin of 25.9%. Management guided the full year 2025 non-GAAP gross margin around 29%.
Non-GAAP gross margin rose to 39.3% in Q4, up 43.4% YoY and 14.1% QoQ. The operating margin turned positive to 18.3% in Q4, up from 3.6% YoY, and from (2.9%) in the previous quarter. Non-GAAP operating margin closed at a yearly high of 23.3%, compared to 7.7% in the year ago period and 2.5% in the previous period.
Bloom Energy drove another record year of double-digit core energy server product cost reduction, which benefits both Products and Services.
Non-GAAP EPS Rises Consecutively for the Fourth Quarter in 2024
Q4 non-GAAP EPS was $0.43, beating consensus estimates for $0.31 by $0.12 or 38.7%, rising 514% YoY from $0.07. This was an improvement over Q3 GAAP EPS of a loss of ($0.01), which missed consensus estimates for a profit of $0.08 by ($0.09), and fell from $0.15 in the year ago period.
Management hinted that Q1 2025 EPS could be up approximately 20% to 30% YoY. For the full year 2024, EPS was calculated using the basic outstanding share count of 227 million, compared to the fully diluted share count of 294 million.
Cash Flows Surge in Q4 Driven by SK Ecoplant Receivables Collection
The collection of SK ecoplant receivables helped operating cash flow surge to $484.23 million, marking the first positive quarter of operating cash flow in 2024. However, a large portion of the cash flow surge came from collecting the SK receivable which was speculated at $325 million by BMO Capital Markets analyst Ameet Thakkar, which is a one off.
Operating cash flow was down ($69.5 million) in Q3. The operating cash flow margin in Q4 was 84.6%, a vast improvement from -21% in Q3.
Free cash flow improved to $473.3 million from down ($83.76 million) in Q3. The free cash flow margin was 82.7% in Q4, up from down (25.35%) in Q3. Bloom Energy closed out Q4 and 2024 was $960.97 million in cash and cash equivalents and debt of $1.12 billion.
Growth By Segment: Products and Services Climb Out of a Hole in Q4
Bloom Energy revenues come in four segments. The largest segment is Product revenue generated from the sale of Bloom energy servers (BES) directly to customers through partnerships and preferred distributor agreements (PDAs). This segment can be lumpy but has been trending in the right direction, starting Q1 2024 with -27% YoY growth to 80% YoY growth by Q4 2024. Bloom Energy closed out 2024 with $2.5 billion in product backlog, excluding the supply agreement with SK ecoplant.
The Services revenue is less lumpy as these are comprised of maintenance contracts and performance guarantees. Extended maintenance contracts are received at the beginning of each service year. Payment comes in the form of a customer deposit that gets recognized over the service period. Q4 non-GAAP gross profit was $4 million in Q4, a vast improvement over a loss of $33 million in the year-ago period. Service was profitable on a non-GAAP basis in every quarter during 2024. Bloom Energy closed out 2024 with $9 billion of service backlog. The Company has a 100% attach rate of service with their product sales. Service contracts can range anywhere from five to 20 years, which is how the large long-term service backlog forms.
**Electricity is a $10 million segment and has been omitted due to being a small fraction of total revenue.
Valuation
Bloom Energy trades at a forward price/earnings (P/E) ratio of 51.72.
The price/sales (P/S) ratio is 3.41, forward P/S is 2.94.
The price-to-book value ratio is 9
The debt-to-equity ratio is 2.616.
Earnings Call:
AI’s $500 Billion Data Center Boom Is Stranded: Power Shortages Demand Urgent Action
The bottleneck with constructing AI data centers lies in the ability to secure the vast amount of power needed to operate them. Power availability is the chokepoint that dictates the entire ecosystem’s viability. Meanwhile, AI components like GPUs have a short shelf life that risks rapid depreciation if power isn’t secured quickly.
I/O Fund pointed out, “Nvidia’s upcoming Blackwell generation boosts power consumption even further, with the B200 consuming up to 1,200W, and the GB200 (which combines two B200 GPUs and one Grace CPU) expected to consume 2,700W. This represents up to a 300% increase in power consumption across one generation of GPUs with AI systems increasing power consumption at a higher rate.“
CEO Shrider gave an example with NVIDIA’s GPUS, “The earnings call from NVIDIA yesterday, okay. Whatever they shipped in that quarter, 90 days, if it were fully facilitized in a data center, will consume anywhere between 2 and 2.5 gigawatts of new power. That's the capacity that's needed. With the growth guidance that they gave you. You fast forward that for the next 12 months. That's just the chips coming from that one company. Fully facilitized will be somewhere in the range of 10 to 13 gigawatts. More than 50% of that is going to stay in the United States. That's more than 6 gigawatts. You're talking about $500 billion worth of infrastructure outside of power. That has to happen even at $5,000. Sorry, $5 billion in terms of facilitating that per gigawatt, that's less than 10%.”
Sridhar underscored the very real dilemma of rapid depreciation for the chips, while trying to secure power. Bloom’s customers are frantic about being able to secure power for this reason.
“Let me make this very clear. You're spending more than $500 billion building a data infrastructure that needs power. If power is not available, and the chips you're installing there have a very short shelf life, because they become old, every year the value of that chip drops like crazy. So, the time to power premium is so high, and the cost of bringing that power, even if you pay the premium is worth every penny of it.”
This is where Bloom Energy Servers come into play offering “AlwaysON” continuous 24/7 power that’s independent to the electrical grid or used as a backup power source. A Bloom Microgrid can be installed to take primary control over critical loads and customize power delivery. Energy Server blocks are repeatable and scalable with the 325 kW base block, which can be duplicated and scaled to multiple MW for any project.
Time to Power is a Competitive Advantage and Purchasing Criteria
CEO KR Shridar emphasized the competitive advantages that Bloom brings to the table, one of which is the ability to bring clients online quickly. He stated this.
“Four years ago, most of our bookings took two to three years to deploy and convert to revenue. The majority of 2024 revenue came from deals that were both signed and recognized in the same year. I expect us to continue to deploy orders quickly, just as we did in 2024. Because for many customers today, the most important purchasing criteria, is time to power. They need reliable power and they need it now. Our Bloom solution, is purpose built to meet that need.”
The fear of data centers getting priority over commercial and industrial (C&I) companies with the utilities is causing them to seek out off-the-grid solutions as Bloom Energy provides. This just adds to the demand for Bloom’s solutions.
The Need for Less Infrastructure May be Beneficial in this Administration
President Trump has declared a national energy emergency. However, he favors fossil fuels and has vowed to end delays for federal drilling permits for oil and gas production. Trump has and is repealing many of the Biden era environmental agenda including modifying and cutting back the Inflation Reduction Act. However, Trump is a big supporter of AI infrastructure as evidenced by his promotion of the $500 billion Stargate project which is a joint venture between OpenAI, SoftBank, Oracle and MGX. The project will build AI data centers across the country to create jobs and enable AI innovation. Both of these factors play into Bloom’s wheelhouse. By promoting more natural gas production and infrastructure, it enables Bloom Servers to more readily access its most used fuel source, natural gas, as management stated.
“Gas is available, the infrastructure is there. And our solution, without needing to add additional transmission, distribution, and making the average ratepayer incur that cost, is going to be politically very attractive. For all those reasons, we think that that's a great market for”
ITC Credits to End in 2028 for Section 48 Projects Before 2025
Shrider also noted that Bloom’s customers, financiers and other commercial ecosystem partners have collectively secured the option to receive full investment tax credit (ITC) for future purchases.
He said this.
“They are entitled to 40% credits nationwide in light of our U.S. manufacturing and 50% credits in predefined energy communities. They can enjoy the tax benefits for systems placed in operation in the United States between now and the year 2028. This Safe Harbor, if fully exercised, has the potential to yield between $12 billion and $15 billion of gross product revenue to Bloom.”
Under commercial ITC section 48 which was extended to Dec 31, 2024, it allows for a 30% ITC base. There is a 10% when using U.S. made components and an addition 10% bonus if installed in “energy communities” referring to fossil fuel towns with high unemployment (IE: coal mining towns). These result in a maximum 50% ITC credit for projects that started pre-2025, which get a 4 year continuity window which is the Dec 31, 2028, deadline.
This means any of Bloom’s installations started or contracted in 2024 can claim the 40% to 50% ITC is the project is completed by 2028. New projects shift to Section 48E. The $12 billion to $15 billion in potential projects refers to the 1.2 GW installed base and new orders like the AEP deal. The $2.5 billion product backlog is the floor and the $12 billion to $15 billion is the potential ceiling.
Could this have possibly triggered a pull-forward effect on orders and the backlog for Bloom Energy? Very possible. It doesn’t impact Q4 revenues because Bloom only recognizes the revenue when shipped, not on orders placed. However, projects starting after Dec 31, 2024, fall under section 48E, which is the Clean Electricity Investment Tax Credit (CEITC).
This is a technology-neutral ITC introduced by the IRA to replace section 48. However, section 48E requires electricity generation or storage with zero emissions, which favors solar, wind, nuclear and batteries. It excludes natural gas, biogas, CHOP and non-electrical technology. Hydrogen fuel would qualify since there is no carbon emissions from that, but its not very cost effective.
CFO Dan Berenbaum covered the $2.5 billion in products and $9 billion in service backlog. He spoke about the shifting priorities now with the business model.
“As we've discussed over the past few quarters, management is primarily focused on overall revenue, product revenue growth, non-GAAP gross and operating margin and cash flow from operations. In the past when we were primarily shipping grid connected, baseload power, and timing of product revenue recognition was somewhat divorced from timing of product shipments.”
Question and Answering Session: Reading Between the Lines
Bloom’s Business Relies on Third-Party Financing
Bloom Energy’s business is very cash intensive and its ability to deploy its backlog is directly tied to its ability to secure project financing. Its pointed out in the 10-K filing, “We arrange financing for our customers’ purchases of our products based on certain conditions, such as their credit quality and the expected minimum internal rate of return on the customer engagement. If these conditions are not met, we may not be able to find financing for their purchases of our products, which would have a negative impact on our revenue in a particular period. If we are unable to arrange financing for our products, our business could be harmed. Additionally, certain financing options, as with all leases, are also limited by the customer’s willingness to commit to making fixed payments, regardless of the products’ performance or our performance of our obligations under the customer agreement. If we are unable to arrange future financing for any of our current projects, it could negatively impact our business.”
CEO Shridar pointed out how capital efficiency and their ability to reduce costs has helped, “So for 2025, so we are being pretty tight about how we manage working capital. We're being very tight around how we manage expenses. We are investing prudently in the right things for the business. And to echo KR's comments, as we've said before, we're quickly approaching about 1 gigawatt worth of manufacturing. We've talked about being able to triple that capacity for roughly $150 million. So as KR said, we're able to grow our capacity in a very capital efficient manner. And to be clear, we will do that when we see the growth.”
Revenue is Now Recognized on Shipments
Wolfe Research analyst Chris Senyek tried to get an idea of the shipments in Q4 from the AEP deal, trying to get a clue if all 100 fuel cells for AEP were shipped in Q4. CFO Dan Berenbaum didn't take the bait and stated they don't talk about the specific timing of shipments for specific customers. However, an interesting point was that their revenue recognition policy has changed from being recognized upon transfer of control to the customer when products are delivered, installed and accepted by the buyer, but now Berenbaum said this.
“Let me just get that out upfront. As I said, in general, we recognize product revenue on shipments that, you know, way back the company used to be divorced a little bit; we used to recognize more of our product revenue on customer acceptance. That shifted a while ago. So now in general, our product revenue is recognized on shipments.“
He added, “And using Bloom to do that is very advantageous for them, from a time to deployment permitting ease, reliability, not needing backups, are being air pollution free, therefore being able to get air permits. For all those reasons, they like our technology. And more importantly, if that growth is being driven by data centers, the reliability of our modular fault-tolerant architecture is unbeatable.”
RBC analyst Chris Dendrinos asked about a breakdown of the backlog between AI and C&I. While CFO Berenbaum bluntly said they were "not going to breakdown the components of the backlog specifically", CEO Shrider did provide some clues.
“I think, we've already given you that kind of numbers, is that roughly one-third of our deployed backlog of greater than 1 gigawatt is towards data centers. And what is happening is that sector obviously is growing a lot faster, than everybody else.”
CEO Shrider also added that Bloom Energy is not dependent on China for the supply chain when asked about tariff implications.
Natural Gas Infrastructure Favors Bloom Energy Servers
Colin Rusch asked an important question about natural gas infrastructure since most Bloom servers use natural gas (not hydrogen) as the primary fuel source. Rusch inquired how much of the backlog was dependent on the incremental execution of gas infrastructure put in place in 2025.
Management didn’t really have much of an answer, “Again, it is a big mix. When we look at quarter-to-quarter, what we implement it is about it like depends on how quickly the projects are ready. That's why we gave it, we give a range of numbers. It's not just whether we can build and ship, it is whether those projects are ready. In many places, it's available. In some places it takes months, and in other places it may be more than a year. So not a single answer to your question. It is across the board.”
Since most Bloom servers operate on natural gas, the infrastructure is important. Schrider talked about regions in the United States and access to natural gas. Very large installations are difficult to do in the Northeast (New York and Massachusetts) due to the lack of gas pipeline infrastructure. West Virginia and Pennsylvania have plenty of gas. He expects Virginia to be attractive and the Great Lakes in Michigan is a “sleeping giant”.
“Gas is available, the infrastructure is there. And our solution, without needing to add additional transmission, distribution, and making the average ratepayer incur that cost, is going to be politically very attractive. For all those reasons, we think that that's a great market for us going forward.”
While Gas Turbines Are Competitors, Nuclear is Not a Concern Until At Least 2030
The demand for onsite power has caused many data centers to turn to gas turbines, like the GE Vernova H-class, to power data centers near term, which is an alternative to Bloom. Natural gas turbines mix compressed air with natural gas ignited at temperatures exceeding 2000 degree Fahrenheit. The hot gas expands through rotating blades that spin the turbine to produce electricity. However, gas turbines emit a lot of carbon dioxide due to their combustion process, whereas Bloom energy servers produce an electrochemical reaction to generate electricity, with no combustion resulting in significantly less carbon emissions.
Management feels secure that they have a six to eight year before small modular reactors come to market. By then, hydrogen could also be more readily available as a economically efficient fuel source.
“We can't afford to wait for nuclear to come before we do AI. So that's what creates this opportunity for the next four to six to eight years. And for us, look, when nuclear comes, the hydrogen play becomes really interesting. Other things become really interesting. So we have pathways – where we can play with those dynamics,” said Berenbaum.
Mitigating Tariffs Headwind Impacts in 2025
When asked about input prices, supply chain and tariff impacts in 2025, Shridar acknowledges that tariffs are potential challenges in their attempts to reduce costs. Tariffs are an unpredictable factor that can impact their operations, but cost reduction is deeply embedded into its business model. This is exemplified by the consistent double-digit cost reductions for the past 12 years, excluding a year of the COVID-19 pandemic, through various methods. The Company recognizes the potential headwinds, but they can be mitigated with cost-reduction efforts relying on their supply chain.
“They come from various different mechanisms, from diversity of supply base, the geographies of where we procure our materials from the efficiency with, which we manufacture, the yield that we get, the engineering advances that we make, all those lead to a cost reduction. So while definitely tariff related issues can be a potential headwind for us, it's one of many factors and we as a company are committed to finding ways around, and still getting to cost reduction.” Bloom is also not overly dependent on China for its components, which offers resilience against potential geopolitical conflicts.
Conclusion: A Strong Growth Pipeline Can Bloom in 2025
Bloom Energy has a compelling story and after two decades, it appears like the stars are aligning for them. Their energy server costs continue to decline, leading to margin expansion. The AI data center boom is power hungry and the game changer deal with AEP enables them to partner with the utilities that are selling the electricity to the data centers. Time to power is definitely a competitive advantage for Bloom. They have demonstrated how they’ve cut time to power down from years to months. Natural gas infrastructure is also a key for them since most of their energy servers run on natural gas, and the United States is the world's largest producer of natural gas and liquefied natural gas (LNG).
Heat capture also boosts the efficiency of Bloom energy servers. We previously wrote.
“BES (Bloom Energy Server) is designed to work with existing carbon capture utilization and storage (CCUS) and combined heat and power (CHP) technologies. CCUS mitigates emissions from natural gas as BES generates a pure stream of CO2 that can be used or sequestered. CHP allows the exhaust heat generated by BES (operating at a core temperature of 1,500 degrees Fahrenheit or 800 degrees Celsius) to be channeled and made available for use, further increasing the efficiency of the system. The high-temperature exhaust stream can produce steam in addition to electricity, resulting in 90% lifetime total system efficiency by adding Heat Capture.”
It's also important to note the slight improvement in customer concentration expanding from their related party, SK ecoplant and SK Eternix. Customer concentration has improved. At the end of 2024, three customers accounted for 53% of total revenues, of which the related party SK ecoplant was 28%. This is a vast improvement from the end of 2023, where two customers accounted for 63% of total revenue, of which SK ecoplant accounted for 37%.
At the end of 2024, three customers accounted for 76% of total accounts receivables, of which its related party was 23%. This was a vast improvement from the end of 2023, where a single customer, related party SK ecoplant, accounted for 74% of total accounts receivables.
The potential 50% ITC for projects that started before 2025 are also a value proposition that provides a $2.5 million Product backlog floor and a $12 billion to $15 billion ceiling if fully exercised. Bloom didn’t mention much about the AEP deal and what amount of the 100 fuel cells were shipped in Q4. The dramatic revenue and cash flow surge was driven by collecting on SK ecoplant's account receivables rather than AEP orders. If this is the case, then the AEP deal provides a lot of upside in 2025, and management’s guidance may actually be a lowball.
Like the Chinese bamboo tree, 2025 may be the year Bloom Energy breaks ground and growth surges. The story gets more compelling.
Welcome to the I/O Fund’s new Discovery Tier, where we cover a new stock idea on a weekly or bi-monthly basis. We are excited to bring you more coverage from the I/O Fund team geared toward new idea generation only.
Jea Yu, Equity Analyst at the I/O Fund, contributed to this article.
Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.