This evening, IREN reported record revenue of $240.3 million, up 355% YoY, and net income of $384.6 million, impacted by a $665 million gain on financial instruments. The company has experienced a significant rebound after retrofitting its Bitcoin mining operations for AI data centers, though its AI Cloud has not yet shown substantial growth, with just $7.3 million in revenue the quarter, up 4% QoQ with only 2,067 GPUs operational. However, cloud revenue growth is expected to accelerate rapidly to $500 million by Q1 and $3.4 billion by the end of 2026 – what remains is the financing and execution.

While the I/O Fund has participated in the AI-energy momentum with successful Bitcoin Miner entries, we want to be clear that these are currently momentum trades for us. In the most recent report, IREN provided a more detailed breakdown on how it will fund $5.8 billion in GPUs, using a $1.9 billion pre-payment, estimating $2.5 billion in financing secured against the GPUs and contract, and the remaining $1.4 billion from cash/cash flows, debt, equity or convertible notes. Payments for the GPUs will be phased in alignment with deliveries through 2026.

However, until financing for the full data center buildout is secured and ARR visibility materializes, we will continue to treat names like IREN as high-risk trades and adhere strictly to our stop levels.

IREN Signs $9.7 Billion Deal with Microsoft

Earlier this week, IREN announced a five-year, $9.7 billion data center with Microsoft, providing the cloud giant with 200MW of capacity of Nvidia’s GB300 GPUs. As is the case with miners, capacity will roll out in phases through 2026, with IREN aiming to deliver the full capacity by year-end. Delivery of the GPUs is not expected to commence until March 2026, hinting that the first tranche will likely not be deployed until calendar Q2.

IREN has provided a detailed breakdown of the deal value, capex costs and pre-payments:

• $9.7 billion total deal value, translating to annualized run rate revenue of $1.94 billion per year, or $9.7 million per MW.
• 85% estimated EBITDA margin, or ~$1.65 billion per year on average.
• $1.94 billion pre-payment (20%) from Microsoft, credited to the third, fourth and fifth year of the deal, covering some of the upfront GPU costs.

Moving to capex and GPUs:

• All-in GPU cost of $5.8 billion including InfiniBand, cabling, servers, etc; while IREN did not disclose the total contracted GPUs, prior GB300 purchases imply that this would represent approx. 72,500 GPUs.
• Data center capex costs of $14-16 million per MW, including $9-11 million for infrastructure, $3 million for 100MW supercluster architecture and flexible rack densities, and $2 million to accelerate deployment of the full capacity by the end of next year.

As stated above, IREN provided a more detailed breakdown on how it will fund the $5.8 billion in GPUs, using the $1.9 billion pre-payment, estimating $2.5 billion in financing secured against the GPUs and contract, and the remaining $1.4 billion from cash/cash flows, debt, equity or convertible notes. Payments for the GPUs will be phased in alignment with deliveries through 2026.

IREN Raises ARR Projection to $3.4B by End of 2026 with ~140K GPUs

With the Microsoft deal now under its belt, IREN had updated its annualized run-rate revenue (ARR) projection to $2.5 billion, which reflects no change to its prior target of $500 million by Q1 2026 excluding the deal.

However, IREN now guided for $3.4 billion in ARR by the end of 2026, which includes its original $500 million target, plus an additional $1 billion ARR target for its remaining 110MW capacity at Mackenzie and Canal Flats. Again, IREN warns that this figure “is not fully contracted, there can be no assurance that it will be achieved, and actual revenue may differ materially.”

Source: IREN

IREN says this forecast assumes ~63K GPUs deployed at its British Columbia sites, which would require the company to procure, receive and install an additional ~40K GPUs before the end of next year. Considering that its ~23K GPU fleet cost upwards of $1.2 billion, IREN may need to find another >$2 billion to scale to 63K in British Columbia, or potentially even $3 billion if it goes primarily for GB300s. With the company already looking for several billions in funding for Microsoft will little to show for AI Cloud revenue, this could require more debt or creative financing methods such as GPU-collateralized loans.

IREN’s 2.9GW Places Third Among the Miners

In terms of overall power capacity, IREN would rank third in the miners with its 2.9GW, behind Applied Digital’s 4.3GW active development pipeline and Galaxy’s 3.5GW. IREN’s Sweetwater campus accounts for a majority of its capacity at 2GW, with substation energization for the first 1.4GW coming in the first half of 2026 and the second substation energization for the remaining 0.6GW in the second half of 2027. Considering the high costs of fully outfitting this entire 2.9GW of capacity with next-gen GPUs, it’s unlikely that IREN’s power pipeline will expand substantially in the near future.

The more important question for IREN is two-fold: how long after substation energization will Sweetwater be ready for service, and how can IREN fund a full 2GW build-out quickly? Big Tech and semis continue to harp on power being the primary constraint, and the differentiating factor for miners is who can deliver the most power the fastest.

Current timelines for Applied Digital, Galaxy and TeraWulf project each will have less than 1GW online by 2027 for key customers, but if IREN can bring the first 1.4GW of Sweetwater online by 2027 (or at least a portion of it), it could be in a better position to secure more lucrative cloud deals. However, self-funding the full buildout will be a challenge as current GPU prices suggest 2GW could cost nearly $60 billion.

Rental Pricing and Payback Periods

IREN’s October 7th announcement about securing multi-year AI cloud contracts included an important but potentially overlooked phrase: “New NVIDIA Blackwell GPUs continue to be contracted ahead of delivery on an average term of 2 years, at pricing that supports a ~2-year revenue payback.”

What this means is that customers are currently willing to contract Blackwell GPUs for two years, as Rubin GPUs will be released late next year and will likely be highly sought after, and IREN is expecting its Blackwell GPUs to be paid back in the same two year contracts.

However, considering how quickly Nvidia (and AMD) are upgrading GPUs and the performance gains each generation brings, these older generation GPUs quickly get priced out of the market. For example, Nvidia’s H100 GPUs were renting for approximately $3.00 per hour in January 2025, prior to Blackwell’s ramp, yet now are renting for <$2 per hour, a (33%) decline that is only likely to exacerbate as the Blackwell Ultras ramp up.

Source: Bloomberg via X

Thus, relying on a two-year payback period under a two-year contract suggests that residual revenue and cash flows from these Blackwell GPUs come 2027 could be significantly lower than current contractual terms. This is because rental rates are likely to follow a similar trajectory of the H100 and decline substantially for two primary reasons: Blackwell availability will be much larger as new systems ramp through the end of the year and 2026, and early Rubin availability will likely draw a significant amount of demand for the more-performant chip.

While it may seem to be a significant positive for IREN that it can realize a two-year payback for Blackwell GPUs, the dynamics of demand and the pace of GPU upgrades imply future revenue opportunities from Blackwell may be more limited in scope.

IREN has a Cash Flow Issue

Stocks like Oracle have recently come under pressure for the enormous debt load the company is expected to inherit to fund the company’s ambitious data center plans. For example, there is rumored to be a $38 billion debt offering as soon as next week, with Morgan Stanley stating the figure could be as high as $55 billion to $75 billion.

IREN is in a similar boat as its ambitions to scale into a competitive AI data center and cloud provider are not congruent with current cash resources, meaning capital intensity and how the company will expand the balance sheet will be a key focus for investors going forward.

Last quarter, we discussed that despite operating cash flow being positive for the fiscal year at $245.9 million, the overall picture of how GPUs and data center expansion would be funded was murky. Due to outsized capex, the free cash flow was ($1.13B) or a FCF margin of –226%. In other words, every dollar of operating cash flow generated was spent 5:1 on the buildout. This improved slightly in fiscal Q1, with every dollar of operating cash flow spent only 2:1.

Financing flows filled this gap with $1.30 billion raised via converts, equity, and leases. Net-net, IREN ended FY25 with $160 million more cash than it started with, despite billion-dollar capex outlays. After year-end, the Company raised another $253.5 million via ATM equity sales and finalized a lease program that funds GPUs entirely, with fixed monthly payments of ~$2.8M and a buyout option at 18% of cost after 36 months. This strategy shifts capital intensity away from cash up front, preserving liquidity while enabling AI Cloud scaling.

However, the issue lies with the more aggressive buildout that is in front of IREN. In the previous analysis, we pointed toward IREN needing $6 billion for a 112K GPU fleet:

At Horizon 1, IREN says that it can host ~19K GB300 GPUs at 50 MW IT load, which, based on prices calculated above, would cost the company upwards of $1.5 billion. Funding this and fully outfitting its British Columbia sites for 100% AI cloud capacity would likely cost $5 billion or more, or ~10x IREN’s most-recently reported cash holdings. This also does not account for its 2GW Sweetwater campus, which IREN says can support >600K GB300s.

Analysts are expecting IREN to quickly scale its fleet through 2026, with Roth Capital projecting IREN to reach a ~112K GPU fleet by year-end 2026. This ~90K increase in GPU fleet could require IREN to take on more than $6 billion in debt, per Roth’s calculations. This would represent nearly 60% its current valuation and likely cost ~$600 million quarterly, which would not be covered by revenue nor cash flows.”

$1 Billion Convertible Raise to Support Expansion

As we discussed in our prior analysis, IREN: GPU Fleet Doubled to 23K, AI Cloud ARR Guide Raised to $500M, outfitting its pipeline with tens of thousands of GPUs will not be cheap, with the company likely to pull out several billions in debt to scale its fleet towards 100K GPUs.

In mid-October, IREN closed a $1 billion convertible note raise, giving the company approximately $922 million in capital for general corporate purposes, which will likely go towards additional GPU purchases. Considering 200MW of GB300 GPUs will cost $5.8 billion, this raise will likely only fund another 20-30MW of capacity.

Microsoft’s Nebius Deal Highlights Miner Shortfalls

Microsoft had signed a similar five-year deal with Nebius in early September worth $17.4 billion for capacity at the neocloud’s upcoming data center in New Jersey, which is expected to have 300MW capacity.

Under such terms, the deal would be worth $3.48 billion on average per year, but on a per-MW basis, $11.6 million per year, or approximately a 20% premium to IREN’s deal. Considering timing is very similar with deployment occurring throughout 2026, Nebius’ ability to command a more valuable deal may stem from its full-stack, proprietary AI cloud purpose-built for AI workloads. Nebius can offer both the powered shell and its platform with MLops services, low downtime and high cost efficiency, offering up to 3x token savings with low latency, and up to 4.5x faster time to token versus other competitors.

This further reinforces that miners’ main value proposition is simply delivering the powered shell in a timely manner, leading to potentially lower deal economics versus neoclouds like Nebius who can combine power with AI-optimized software. It’s also likely why IREN is making a deeper push to bridge the gap with its own AI-optimized cloud offering, as it could drive more valuable deals with hyperscalers, versus a deal such as Cipher’s with AWS worth $1.22 million per MW per year on average.

Financials:

Revenue

IREN reported a record $240.3 million in revenue in Q1, up 355% YoY and more than 28% QoQ, driven primarily by Bitcoin mining revenue of $232.9 million. This beat estimates for $228.5 million the quarter.

AI Cloud Revenue

IREN’s AI Cloud revenue showed minimal growth in fiscal Q1 at just 4% QoQ to $7.3 million, though this was up more than 128% YoY. Operational GPUs rose just 9% from 1,896 to 2,067, less than 10% of the ~23K the company has purchased; considering the company has announced that 11K of its fleet has been contracted under multi-year deals and are expected to be operational by the end of the year, deliveries and revenue should ramp significantly in the December quarter.

Margins

Gross margin was 66.4% in fiscal Q1, down from 71.8% in Q4, as net power prices rose 31% sequentially. Bitcoin mining gross margin shrunk from 70.9% to 65.7%, while AI Cloud gross margin shrunk from 92.9% to 90.4%.

However, operating margin was (31.8%), a stark contrast to the 11% reported in Q4, driven by a 107% sequential increase in operating expenses to $236 million, or more than 98% of revenue. This was fueled by SG&A, which rose from $53.3 million in Q4 to $138.4 million in Q1, which IREN says was driven by a “materially higher share price” that resulted in an additional $23.9 million of stock-based amortization and $32.8 million in payroll taxes related to RSUs.

Net margin was 160%, as IREN reported $384.6 million in net income, impacted by a $665 million gain on financial instruments, primarily related to prepaid forwards and capped calls on convertible notes.

Adjusted EBITDA margin also contracted from 65.1% in Q4 to 38.2% in Q1.

Cash Flows, Cash and Debt

Operating cash flow was $142.4 million, though free cash flow was ($138.2 million) as IREN spent ($180.3) million on PP&E and another ($100.3) million in GPU prepayments.

IREN reported cash and equivalents of $1.03 billion at quarter end, though noted that at the end of October, cash and equivalents totaled ~$1.8 billion following its $1 billion convertible raise. This is a sharp increase from $564.5 million as of Q4.

Property, plant and equipment was $2.12 billion, up from $1.93 billion in Q4.

Debt (convertible notes) was reported at $962.4 million, though this does not include the recent $1 billion convertible raise. IREN also added that it secured an additional $200 million in GPU financing in the quarter, bringing its total there to $400 million. IREN expects future capex needs to be met by cash/cash flows, GPU financing, Microsoft’s prepayments and additional financing methods.

Equity rose from $1.82 billion to $2.88 billion, driven by an increase in cash, financial and derivative assets. Shares outstanding rose 4% from August to October, from 272 million to 283.5 million.

Conclusion:

IREN’s AI Cloud revenue is at $7.3 million with management forecasting $500 million by Q1 and $3.4 billion by the end of next year. Of this, $1.9B is to come from the Microsoft deal over a five-year period: “When combined with the $1.9 billion expected from the Microsoft contract and $500 million from our existing 23,000 GPU deployment, this expansion provides a clear pathway to approximately $3.4 billion in total annualized run rate revenue once fully ramped.”

Estimates call for $2.5 billion at the end of fiscal year 2027 ending in June, implying the Microsoft deal and the $500 million could already be priced in. At some point, the market will want to take a rest and watch how the financing and execution pieces comes together (not IREN specific).

Also keep in mind, IREN’s main source of revenue is Bitcoin and Bitcoin prices have been dropping. As stated in our original IREN coverage, this is more of a momentum trade setup: “We are watching IREN closely and would buy on a clear breakout only. If we were to buy, we’d closely adhere to all stops.”

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. Beth Kindig and the I/O Fund own shares in IREN at the time of writing and may own stocks pictured in the charts.

Recommended Reading:

• TERAWULF BITCOIN MINER: GOOGLE TAKES 14% STAKE VIA FLUIDSTACK PARTNERSHIP
• BITCOIN MINERS ADDRESSING AI’S NEAR-TERM TIME TO POWER BOTTLENECK WITH UP TO $50 BILLION IN COMMITMENTS
• APPLIED DIGITAL: BITCOIN MINER HINTING AT RARE, HYPERSCALER DEAL
• INNODATA: FY25 REVENUE GROWTH RAISED TO >45%, YET LARGEST CUSTOMER REVENUE IS FLAT FOR TWO QUARTERS

Bloom reported a strong Q3 with revenue beating estimates by more than $90 million, strong margin expansion including GAAP operating margin shifting to positive territory. However, the reason the stock surged after hours was not found in the earnings report, rather the stock was up as much as 20% after hours from management commentary on FY2026 during the earnings call.  

Bloom Energy has seen incredibly strong price action this year of nearly 400% YTD and is nearly 600% off the April lows. Therefore, it would take something unexpected to get the stock to soar after an earnings report – yet management delivered exactly that by stating: “As we have previously announced, we are doubling our capacity to 2 gigawatts by December 2026, which will support about 4x our 2025 revenue. That expansion is all systems go. Bloom's capacity will not be a bottleneck for our customers.” 

Although management stated “about 4X our 2025 revenue,” a conservative approach would be to assume the revenue will be recognized across both 2026 and 2027. Analyst estimates are for $1.9B in 2025 and a mere uptick to 16.6% growth next year for revenue of $2.2B. However, helping the bull case further that last night’s comments offer alpha is that Bloom detailing further their primary benefit, which is quick time to power. Therefore, I imagine some of the “4X from 2025 revenue” from doubling the capacity will occur in 2026 and some will occur in 2027. I’m not choosy on which exact quarter given there is a wide disconnect in analyst estimates as 2027 estimates are for $3.5B, or less than 2X 2025 revenue.  

The 2GW represents 100% growth in capacity as AEP had contracted for 1GW going into this year. My interpretation is that Bloom Energy must be able to charge more for its power given how rapid their solid oxide fuel cells are deployed with expectations of 90 days, yet they actually delivered in an astonishing 55 days for Oracle. In the past, management had hinted they were doubling GWs but did not correlate it to a quadrupling of their revenue. From Q2 earnings call: “Now our robust product has robust demand. We will double our factory capacity from 1 gigawatt a year now to 2 gigawatts a year by the end of next year. Our mission has never felt more urgent, and we are ready.” 

Even with this blockbuster statement of “about 4x our 2025 revenue,” one has to wonder if 2GW is the baseline for 2026 capacity. The company counts one massive energy partner Brookfield, two hyperscalers and one neocloud as customers (ORCL, AWS via AEP and CRWV) plus they hinted of a fourth large customer on the call today via a gas company partnership. Secondly, management explained why their product is in high demand and will likely remain that way for some time. We cover this and more below! 

Please note, we are shifting our post-earnings formatting to have the contextual information including Q&A commentary first and financials second. 

“About 4X our FY2025” – The Comment that Caused the Stock to Surge  

I want to double click on the comment that 2GW will be “about 4X our FY2025 revenue.” Surprisingly, the first few analysts on the call breezed over the comment before an Evercore analyst asked for clarity. I’m quoting this in full given the importance of the comment in the opening remarks with the CEO confirming the analyst’s understanding and stating, “we will never be the constraint in our customers growing their data center“ and that “all systems are go.” My translation is that Bloom can increase its capacity faster than many other energy options and now that “lighthouse” customers such as Oracle have taken note, that we will see what Bloom is truly capable of in the coming quarters. As Bloom’s CEO stated, commercial demand is “accelerating.” 

Here is a further breakdown on the 4X comment: 

“Nicholas Amicucci 
Evercore ISI Institutional Equities, Research Division 

I just wanted to build upon on kind of the doubling of capacity by the end of 2026 and kind of the commentary that would support 4x the fiscal '25 revenue.  

How should we think about kind of the utilization on that capacity as we kind of enter into — again, as we enter into 2027 and we have that — the 2 gigawatts kind of up and running. I mean because if we're exploring opportunities to go beyond that 2 gigawatts, it seems like 4x full year '25 revenue, that seems like a big number that we could get there relatively quickly. So I just wanted to parse that out a little bit. 

K. Sridhar 
Co-Founder, CEO & Chairman 

Yes. So here is a simple way to think about it, right? We didn't get to where we are today to deliver what I just explained, this purpose-built factory based on just meeting a market demand as we see it right now, we just prepared ourselves. What is the beauty of Bloom being able to expand its capacity and offer what we do? Is the return on investment like invested capital? So we are fiscally very disciplined, and we only make decisions based on that added cost and its absorption, will it have a great rate of return. 

So we have a very disciplined process on this. And on top of that, we have a very clear understanding right now given time to power shortages and the importance of this as a nation-state issue for AI.  

We are committing to strive and work as hard as we need to and stay ahead such that we will never be the constraint to our customer on growing their data center. That's what we are positioned for. And we will increase capacity. We will increase it in whatever steps necessary as we see fit. But as you saw, this 2-gigawatt capacity, all systems go based on that.  

Would we use it for peak capacity? When we use it, will we use it for steady capacity? All that, you'll hear from us as we talk about our backlog and other things next year. But we are now using our OpEx wisely to invest in capability and talent to think about how do we expand beyond 2 gigawatts. That's all I can say right now. Thanks for that question.” 

Regarding how fast Bloom Energy can build the additional GW, management was confident they can do so faster than anyone else: “Today, we are able to provide our power faster than most of the others who have supply chain constraints. We can expand our capacities a lot faster than anybody else.” 

Why Bloom Energy Remains in High Demand Amidst a Crowded Energy Industry 

We’ve covered Bloom’s products extensively, yet it doesn’t hurt to refresh our understanding of what makes the company stand out given the market dynamics around how data centers are scrambling to secure power is shifting nearly daily.  

Our primary message has been “time to power” for Bloom, which management emphasized stating: “We are going to strive to make sure we are able to provide power for our customers before they are ready for it. We will not be the bottleneck. And we designed our factories; we built it with that in mind.” 

Looking beyond speed, management also focused on price-to-performance, especially when they were asked how Bloom plans to compete with small-scale gas turbines with management stating a clear benefit to their solid oxide fuel cells (SOFC) is that hyperscalers can put out a lot more tokens with their fuel cells, stating: “With the same amount of gas that's available, same amount of space that's available, we can produce a lot more tokens for the hyperscaler than any other technology can today, end-to-end. And so the value for a hyperscaler is not about the cost of power. It's about that cost of the entire value chain across the board. So price-performance ratio, we can compete with anybody.” This was reiterated in the opening comments with management stating their fuel cells produce “10x more power in the same footprint than they did 10 years ago.”  

It was also discussed that mechanical combustion solutions require a lot of batteries, whereas Bloom does not require batteries, implying that gas turbines are a band-aid solution compared to SOFCs, which can provide offer steady output without the grid or batteries.

Updates on Bloom’s Deals with Brookfield, Oracle, Hints of New Hyperscaler Customer 

Earlier this month, Bloom shares surged more than 26% to $110 on the backs of a partnership with Brookfield, which will see the asset management firm invest up to $5 billion in Bloom’s fuel cell tech to be deployed at data centers worldwide. While timelines are rather unclear for deployment of the fuel cells, the two state that the partnership does include an AI inference focused site in Europe that will be announced before year-end. To put in perspective the potential size of the partnership, this would represent nearly 3x of Bloom’s estimated revenue for fiscal 2025. 

Under the partnership, Bloom will become Brookfield’s “preferred on-site provider for Brookfield's trillion-dollar infrastructure portfolio of AI factories, data center operators, corporate facilities and factories.” What makes this partnership important is not only the fact that Brookfield has invested $50 billion towards AI and “is tripling the size of its AI strategy over the next 3 years,” but that Brookfield is willing to finance for Bloom. 

CEO KR Sridhar explained that “if there are Bloom-sourced deals that require financing, so we can offer a customer a PPA, they are willing to step in and be the financier for that. It’s an inaugural investment.” This would remove an important layer on the equation for growth for Bloom as it would help them accelerate deployments without them or customers bearing the capital or financing risks.  

Management also hinted of another hyperscaler customer in the works, but declined to provide any further details: “We signed our first deal with a major gas provider who will convert its gas to electricity with Bloom fuel cells and sell that on-site power to a third hyperscaler. The hyperscaler will announce details of this installation when it is ready.”  

Bloom also struck a deal with Oracle back in July to deploy its fuel cells for onsite power at select Oracle Cloud Infrastructure (OCI) data centers over the next 90 days. While terms of the deal such as size were not disclosed, Bloom completed shipments in just 55 days, highlighting its ability to deliver power to data centers rapidly.  

Bloom had signed a partnership with CoreWeave in July 2024, with the first fuel cells expected to be commissioned in Q3 2025 for a data center in Illinois. Bloom briefly updated on this, saying that its fuel cells are generating power at the facility. However, it is rumored that the data center is just 14MW, essentially making it a pilot/validation deployment rather than a full-scale commercial deployment.  

Nvidia’s Rubin is Coming; Bloom Energy is Ready 

There was discussion around how Bloom Energy’s solutions could become more attractive with the Rubin generation of GPUs with an analyst asserting DC/DC power would be more efficient than DC/AC power.  

Our team has covered quite closely the power requirements for Rubin Kyber racks, which could draw 600kw or 5X that of the NVL72 systems that are shipping now. You can read more here on this topic (an important read if you are invested in AI energy stocks). 

What was discussed on the earnings call is the power requirements will put immense pressure on voltage, stating: “the laws of physics dictate that you have to go to an 800-volt DC architecture if you want AI chips that have more power density, which is the only way you can improve upon AI in the next generation. This is not an if, this is not a nice-to-have. This is a must-have.” 

Bloom is asserting they are prepared for the 800-volt DC architecture (whereas most energy solutions are not such as 50MW turbines) as they are adaptable when moving beyond the 48-volt DC architectures of today. 

“That is the 48-volt DC because the small wire through which a small amount of water comes into the straw, that water was sufficient to satiate the thirst. That was when CPU racks were 13 kilowatts. We have put a lot of Band-Aids on it to make sure Blackwell chips that come somewhere near the 130 kilowatts can handle it through the straw […] We saw this coming one day. We didn't know what day in 2000 when we initially created architecture. We built an architecture where we can feed these straws appropriately right at that 800 volts, and we decided every unit we have shipped for the last 15 years has that.” 

Bloom Q3 Revenue Beat of 21% 

Bloom smashed analysts' revenue estimates by 21.3%. The company reported record revenue of $519.05 million, versus estimates of $428.07 million. Revenue grew by a solid 57.1% YoY and 29.4% sequential growth, accelerating 37.6 percentage points from the previous quarter’s YoY growth of 19.5%. 

The company’s fourth consecutive record revenue was driven by the strong demand for its fuel cell technology, driving AI data centers. We have discussed the fuel cells opportunity as a key catalyst in our article here. The company’s fuel cells are very efficient and are currently producing 10x power within the same footprint than produced previously a decade ago. 

The management highlighted three major tailwinds that are positioning the company to become a global standard for on-site power generation; a market expected to reach a trillion dollars. First, AI buildouts have increasingly made on-site power generation a core necessity. Secondly, since AI is a national priority, government policy on on-site power generation is now more liberal. Third, the company’s fuel cells are highly efficient and are witnessing double-digit YoY cost reductions.  

Revenue growth decelerates in Q4 due to tough comps, as last year’s Q4 revenue grew by 60.4%. Analysts expect Q4 revenue to grow 6.4% YoY to $608.7 million. Revenue growth will accelerate to 20% in Q1 2026 and to 23.6% growth in Q2 2026. 

Looking ahead, analysts expect 2026 revenue to grow 24% YoY and accelerate to 62% growth in 2027. Management sounded optimistic on the future growth as the company’s co-founder and CEO, K. Sridhar said in the Q3 earnings call, “This seminal year for Bloom positions us for an even stronger 2026 and beyond with higher growth and more profitability”.  

The analysts' estimates would trend higher after management's comments during the recent earnings call that doubling capacity to 2 gigawatts would support 4x the company’s 2025 revenue. Using a conservative approach, we believe revenue may be recognized over the next two years.   

Key Segments 

Products, installation, and service revenue growth showed acceleration from the previous quarter. While Electricity segment declined sequentially. 

• Products revenue grew by 64% YoY to $384.3 million, accelerating from the 31% growth in Q2. 
• Installation revenue growth spiked 105% YoY to $65.78 million, accelerating from a (13%) decline in Q2. 
• Service revenue grew by 16% YoY to $58.6 million, accelerating from the 4% growth reported in the previous quarter. 
• Electricity revenue was down (25%) YoY to $10.35 million, decelerating from a decline of (10%) in Q2. 

Margins Continue to Expand 

Bloom’s margins are improving, primarily driven by operational efficiency, product cost improvements, and operating leverage. Bloom is fundamentally transforming into a stronger company, as its GAAP operating margins were previously deep in the red, in double digits. 

• Q3 gross profits grew by 92.7% YoY to $151.68 million or a gross margin of 29.2%, up 5.4 percentage points YoY and 2.5 percentage points sequentially. Similarly, adjusted gross margins showed strong YoY and sequential improvement, primarily driven by product cost improvements and manufacturing efficiencies. 
• Operating margins improved 4.4 percentage points YoY and 2.4 percentage points sequentially to 1.5%, primarily driven by strong operational efficiencies. Adjusted operating profits grew by 470% YoY to $46.2 million or an adjusted operating margin of 8.9% compared to 2.5% in the same period last year and 7.1% in the previous quarter. 
• Net margin was (4.4%) compared to (4.5%) in the same period last year and (10.6%) in the previous quarter. Adjusted net income was $35.45 million compared to ($1.5 million) in the same period last year. Adjusted net margin improved 7.2 percentage points YoY and 1.3 percentage points sequentially to 6.8%. 

Adjusted EPS beat of 47% 

GAAP EPS came at ($0.10) in Q3 compared to ($0.06) in the same period last year. GAAP EPS was negatively impacted by a one-time loss related to unconsolidated affiliates of ($19.6 million) or a ($0.08) per share. The company reported adjusted EPS of $0.15, beating estimates by 47%, and was up from ($0.01) in the same period last year and $0.10 in the previous quarter. Bloom reported strong profits growth driven by operational efficiency, product cost improvements, and operating leverage. 

Analysts expect adjusted EPS of $0.31 in Q4 and $0.04 in Q1. Looking forward, adjusted EPS is expected to grow strongly by 84.7% YoY to $0.93 in 2026 and 122.4% to $2.07 in 2027. 

Cash Flow and Balance Sheet 

The company reported positive operating cash flows and free cash flows in the recent quarter after negative cash flows in the first two quarters of the year. 

• Q3 operating cash flows were $19.67 million or 3.8% of revenue compared to ($69.5M) or (21%) of revenue in the same period last year. Operating cash flow improvement was primarily driven by higher profits and working capital improvements. 
• Strong operating cash flows also led to higher free cash flows. Q3 free cash flow was $7.4 million or 1.4% of revenue compared to ($83.8 million) or (25.4%) in the same period last year. 
• While management has not provided concrete 2025 guidance, it noted on the earnings call that “we expect fiscal 2025 to be better than our previously stated annual guidance on our financial metrics”. It suggests that the company’s cash flows and free cash flows would be strong in Q4, based on management's guidance during Q2 results that cash flows would be at the same level as in 2024. To give a perspective, the company reported operating cash flow of $92 million in 2024 and free cash flow of $33.2 million. Year to date, the company reported operating cash flow of ($304 million) and free cash flow of ($338 million), which means operating cash flow will be about $396 million and free cash flow will be about $371 million, respectively, in Q4. 
• Cash was $595.1 million and debt of $1.13 billion at the end of Q3 2025. While debt remained unchanged, cash improved by $20.3 million from the previous quarter. 

Conclusion:  

We are in the era of “what you see is what you get” – meaning, those offering strong earnings reports right now are setting up for a strong runway as future generations of GPUs will only be more power hungry. There is far less speculation than there was at the start of the year when we first covered Bloom in terms of which energy solutions can answer the demands of the AI data center buildout. The test for investors will be figuring out how to hold-on while this market unfolds in the coming years.  

We hope to help with all of the above from being early to the products and solutions driving forward this massive market, to carefully examining the financials for confirmation the company is delivering, and providing the technicals to help stay the course while also not getting too emotional during the highs and lows.  

Our earnings season is off to a strong start, we have dozens of reports to cover for you alongside real-time trade alerts that do what few can offer – analyze the complex AI market yet also execute.  

Regarding the flawless execution, I want to thank the team on this one: Knox, Damien and Royston. It’s been a pleasure to see the pieces come together, and we hope there are many more like it to come. 

I/O Fund Equity Analysts Damien Robbins and Royston Roche 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.

Recommended Reading:

• The I/O Fund’s Top 15 AI Stocks for Q4 2025
• Galaxy: Strong Crypto Backdrop Drives Gross Profit Surge, Data Center Progress Encouraging
• Applied Digital: Bitcoin Miner Hinting at Rare, Hyperscaler Deal
• TeraWulf Bitcoin Miner: Google Takes 14% Stake via Fluidstack partnership

Bloom Energy is retracing the 10% pop it saw yesterday from the news Bloom has agreed to supply Oracle with solid oxide fuel cell (SOFC) servers. Despite Bloom Energy being in direct contact with its first hyperscaler customer with Oracle, the drawback is that BE had already baked this into fiscal year guidance. Thus, the company was only able to reiterate fiscal year guidance of $1.65 to $1.85 billion, corresponding to YoY growth of 19.1%.  

Despite the weak price action from inflated expectations around the Oracle deal, there were many bright spots in the report. Bloom reported its third consecutive quarter of record revenue and profits and reported six consecutive quarters of profitability in the Services segment.  

Bloom also outlined a plan to double its factory footprint from 1 GW to 2 GW by the end of 2026 to meet growing demand, with this expected to cost in the “ballpark of $100 million,” funded through a recent refinance of debt notes for $113 million.  

To help compare, Data Center Frontier estimates that BE has deployed 400 MW of capacity to data centers this year,  and has delivered 1.5GW of power in total across 1,200+ global installations. Therefore, 2GW exceeds Bloom’s total history of installations and is about 5X its 2025 data center business.  

Details around its new partnership with Oracle were the main focus on the call, plus how quickly Bloom can scale capacity and how the company plans to fund any future expansion.  

Although we expect Bloom to be very volatile, the fact is that very few alternative energy companies can move as quickly as BE in what our firm has dubbed an energy crisis in getting power to data centers.  

As the CEO stated on the call, to wait 5-7 years is “untenable.”  To compare, Bloom will power Oracle with on-site power solutions in as soon as 90 days. Additional key customers for BE include American Electric Power (AEP), Quanta and Equinix. Notably, Amazon and Cologix are customers of Bloom through AEP in Ohio.  

Perhaps the most important statement on the call was when the CEO stated: “We expect new orders from other AI hardware ecosystem players soon, complementing demand we see from our more traditional commercial and industrial customers.” 

Revenue Beats by 6%, Yet FY25 Guide Maintained 

Bloom reported a nearly 6% beat to estimates in Q2, reporting $401.2 million in revenue versus estimates for $378.9 million. Revenue grew 19.5% YoY, slowing from 38.6% growth in Q1.  

According to management, this represents the highest revenue and most profitable Q2 yet: “Bloom had an excellent quarter, the highest revenue and most profitable second quarter in our 24-year history […] Over the last couple of calls, I've told you that our business is at an inflection point as demand for clean, reliable and rapidly deployable power is surging. Now there is tangible evidence.” 

Growth is expected to rebound to the high-20% level in Q3 to $424.9 million, though notably this comes against a substantially weaker comp in Q3 2024’s (17.5%) decline. On the other hand, Q4 faces a difficult 60% comp, and as a result, growth is estimated to be <7% YoY to $610.5 million. 

Despite the beat in Q2 and recent deal with Oracle to deliver fuel cells to AI data centers within 90 days, Bloom maintained its full-year revenue guidance at $1.65 to $1.85 billion. This corresponds to YoY growth of 19.1%.   

Key Segments 

Product revenue growth moderated slightly in Q2 but remained in the 30% range, while Installation and Electricity both declined YoY, reversing from strong growth last quarter.  

• Product revenue increased 31.1% YoY to $296.6 million, slowing from 38% growth in Q1.  
• Installation revenue declined (12.5%) YoY to $37.4 million, reversing sharply from 194% growth last quarter.  
• Service revenue increased 3.7% YoY to $54.4 million, rebounding from a (5%) decline in Q1. 
• Electricity revenue declined (9.8%) YoY to $12.8 million, reversing from 92% growth in Q1. 

Operating Margins Expand  

On a YoY basis, margins have improved quite substantially. Bloom is transforming into a stronger company fundamentally when considering its GAAP operating margins were previously deep in the red double-digits. 

Gross margins dipped sequentially, yet operating margins expanded on a GAAP and adjusted basis. Notably, Bloom believes their operating margins will continue to expand: “Between that combination and our cost reduction continuing, you should absolutely expect our operating income to keep getting better as we go forward.” 

• GAAP gross margin was 26.7%, down 0.5 points QoQ but up more than 6 points YoY. Adjusted gross margin was 28.2%, also down 0.5 points QoQ and up more than 6 points YoY.  
• GAAP operating margin is approaching positive territory at (0.9%) in Q2, up nearly 5 points QoQ and 6 points YoY.  
• Adjusted operating margin was 7.1%, up more than 3 points QoQ and 8 points YoY. Adjusted EBITDA was $41.2 million. 
• GAAP net margin was (10.6%), down 3.3 points QoQ and up nearly 8 points YoY. Adjusted net margin was 5.5%, up 3.5 points QoQ and 9.7 points YoY.  

EPS 

Bloom beat EPS estimates on an adjusted basis as adjusted margins expanded down the line, though GAAP EPS fell short.  

• Adjusted EPS of $0.10 beat estimates for $0.02, and represented a notable $0.16 improvement YoY.  
• GAAP EPS was ($0.18), missing estimates for ($0.10) as GAAP net margin declined sequentially. 

Looking ahead, Bloom is expected to see profits at least at this level through the end of the year, generating the bulk of its earnings in Q4 at $0.31.  

Cash and Balance Sheet 

Cash flows worsened sequentially, with operating cash flow falling by more than ($100 million) versus Q1. This weighed on unrestricted cash and equivalents, while debt was unchanged. 

• Operating cash flow was ($213.1 million) in Q2 for a (53.1%) margin, nearly double Q1’s ($110.8 million) outflow. For the first half of 2025, operating cash flow was ($323.9 million), approximately flat YoY. 
• Bloom guided for FY25 operating cash flow to be approximately flat to FY24 at $92 million, suggesting 2H operating cash flow in the range of $410 million, likely concentrated heavily in Q4. 
• Free cash flow was ($220.4 million) in Q2 for a (54.9%) margin. For the first half, FCF was ($345.3 million), just over a 1% improvement YoY. 
• Unrestricted cash and equivalents totaled $574.8 million, down from $794.8 million in Q1. This raises the risk that Bloom will turn to financing methods as Bloom likely awaits cash flows meaningfully improving in Q4.  
• Debt remained steady at $1.13 billion.. According to the opening remarks: “Finally, during the second quarter, we refinanced $113 million of our convertible note that was due in August 2025 to provide more optionality to fund future growth. It was exchanged into our existing 2029 convertible notes.” 
• Inventories were $690 million, up 12.7% QoQ from $612.5 million in Q1, supporting some near-term deployment growth for the Oracle partnership.  

Earnings Q&A: 

Bloom becoming more attractive to hyperscalers: 

When asked about the Oracle deal, the CEO stated that it was “islanded power” to where Bloom is the first source and the second source (rather than being backup power to the grid) and that “it will be one single data center that the first project will power and we are working with them on many of the projects” — hinting the partnership is expected to expand over time and is “extremely significant” for Bloom Energy. 

“So we see this as extremely significant, and we are the primary source, and it is load following. So it will prove that we can load follow at large scale. It will prove that we can operate at large scale and most importantly, AI speed. It will prove that we can install stamp sizes at that level within the 90 days that we have told you we would do in our opening remarks.” 

There was also discussions that Bloom may have an important cross-sell opportunity beyond time to power, which is a combined heat power solution that helps to increase efficiency through thermal management. It was briefly mentioned it could save up to 20% of power costs: “And you are correct to point out from a value proposition wise, right, it is the equivalent of not needing 20% of your power in a data center, right? It is the equivalent of not paying for it when you have taken care of your cooling with our waste — with the waste heat as opposed to putting more electricity. That's a big deal.” 

Overall, Bloom is quite confident their solution drives down costs compared to other solutions. When asked how they compare to natural gas turbines, the CEO stated: “On the other hand, those turbines have at least 15 to 20 percentages — like percentage points more fuel that they will consume compared to us. So on the OpEx, there's a significant win. And on top of that, we have no air pollution, whereas getting an air permit to put a lot of turbines if you live in a populated area is very difficult. You're probably reading in the press. 

So you combine all those things, A, because we are easier to permit, we remove the friction to permitting, so we are faster. Time to power is everything in this business. Secondly, operating cost is lower. CapEx is at parity. You put them all together, I think we compare more than favorably to any other alternative way of producing electricity.” 

2GW on the Way to Multi-GW 

According to management, it will cost $100 million to grow from 1GW to 2GW capacity: “So we are funded for what we — we are well funded for what we need to do in terms of going to 2 gigawatts. Round ballpark numbers, think about $100 million is how you should be thinking about this. And it will come spread over quarters. And we have enough.” 

An analyst asked what gives management the confidence to double capacity – (which is more than just doubling capacity as it also represents more capacity than the company has installed in its lifetime at 1.5GW).  

The CEO pointed toward the visibility in backlog and also the large capex spend and what that essentially means for power capital equipment: 

“So we have told you in the past that in the last 2 quarters, we have seen strong commercial activity. We have told you that it is very diverse, and it is high quality. At this point in time, when we look at that pipeline, it has gotten us to a level of confidence where we absolutely feel like this is the right thing to do. That's why we are expanding the capacity, number one, right? 

Number two, this should be fairly simple, and it should be mind-boggling for all of us, and we shouldn't get numb to this fact. The large hyperscalers put together are going to spend more than $1 billion a day on CapEx, weekday and weekend. It's more than $500 billion are going to be spent just in this calendar year by those people. So you take that number of $500 billion and you say, an order of magnitude down, at least $50 billion of power capital equipment needs to be spent to electrify that additional demand that's going to come on.” 

Tax Credit Benefits: 

Lastly, tax credits are a tailwind for Bloom Energy as there are tax credit benefits to companies who use their energy solutions. According to the CEO during the Q&A session, it’s about a 30% incentive: 

K. Sridhar   Co-Founder, CEO & Chairman 

[..]  So it is a flat 30%, okay? Whereas in the previous version of the bill that ended last year, but safe harbored now, our customers can avail of either 40% or 50%, depending on whether they are not in an energy community or in an energy community. If they're not in an energy community, it's 40%. If they're in an energy community, it's 50%. So from that perspective, yes, their subsidies go down a little bit. But given how high the price of electricity has gone up, at 30%, our attractiveness will be extremely high.” 

Conclusion: 

Bloom Energy is reserved for Advanced members because it does not have a quality fundamental profile, mainly seen in its need to raise cash to build capacity. However, there is a hypergrowth story here for many years to come. Bloom provides immediate time to power within 90 days compared to nuclear in 5-7 years’ time. The company articulates why they can compete with natural gas turbines with roughly 20% better energy costs albeit Bloom is an alternative energy source and many data center operators have less experience with hydrogen fuel cells. That is why Bloom will become more of a snowball effect, to where each deal (Oracle, AWS via AEP, and Quanta) will build on itself to prove to future hyperscalers that Bloom is not only a viable choice but a preferred method to quickly get power to data centers.  

What I’m saying — perhaps in a long-winded way — is that it’s the intraquarter deal announcements that will move this stock. I anticipate the I/O Fund will actively trade Bloom over the next few GPU generations as they push forward power requirements that only a handful of companies can (quickly) meet. Bloom is one of those companies.

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.

Recommended Reading:

• Positions Report – July 2025
• The I/O Fund’s Top 15 Stocks for Q3 2025
• Coinbase Underperforms in Q2; Near-term Volatility Likely
• Meta Reports Large Q2 Beat, Ad Impressions Growth Rebounds