Riot had a lackluster report with no new surprises to speak of. In terms of timing, the initial 112MW build-to-suit data centers will not be complete until early 2027 despite being energized much earlier. For Riot, the timing issue is centered around design and construction rather than available power.
We’ve done some back-of-the-napkin math in our previous analysis to put a value around $792 million in average annual revenue for the 600MW for a total of $8 billion over ten years, with the understanding the beginning stages will begin at a much smaller size as capacity ramps and end much larger than the averaged figure of $792M.
Analyst estimates are currently for $805M next year, and Bitcoin operations are producing $160.8 million this quarter (this is lumpy of course). There’s certainly a path to where Riot sees strong upward revenue revisions when the 600MW is energized and delivered. Analysts have estimates for Q1 FY27 at $181M, which would theoretically create upside when the AI data center story materializes if we assume Bitcoin continues to offers the low end with a $200M run rate.
The upside in analyst estimates is more apparent on the bottom line with our back-of-the-napkin math suggesting net operating income of $673 million annually. Last year, Riot reported net operating income of $92.3 and has reported $5 million in NOI year-to-date (currently this is largely dependent on Bitcoin). Analyst estimates call for a decline in EPS of (43%) for FY2026 and net losses in FY2027. Therefore, if/when management truly delivers the 600MW in the next 1-2 years, it should drive a significant rebound on the bottom line.
What we must grapple with is the timing for this rebound as other Miners are executing more quickly. The AI market is moving like a freight train, and Q1 FY2027 seems awfully far away. The opportunity cost with stocks is tricky, as Riot could announce a deal with a hyperscaler at any time, yet so could its many peers – in fact, they’ve been doing so at a fast clip such as Applied Digital, TeraWulf, and more.
Below, we look at Riot and if the opportunity cost of remaining in the stock makes sense or not.
First 112MW to Enter Construction in Q1 2026
While Riot announced this quarter that it has initiated its data center strategy with development of the first 112MW of IT capacity at Corsicana, construction on these first two 56MW buildings is not expected to commence until Q1 2026.
Riot says that completion of the core and shell will enable it “to deliver full build-to-suit data centers in 2027,” hinting that any deals secured with tenants may not contribute meaningful lease revenue until then. Compare this to other miners such as TeraWulf, Applied Digital or Core Scientific, who will already be delivering hundreds of MW of capacity by the end of 2026 and who have deals in place with visible revenue ramps.
The reason this is important is because miners’ value lies within timing – this is a race for power and how quickly they can deliver data center power to customers. Currently, Riot is slower to execute on the construction side as the language here suggests that the power will be available without a need for further regulatory approvals, rather the build and design phase is the delay: “We have also further progressed on the ongoing infrastructure development at Corsicana, including the 600-megawatt substation expansion, where the first 400-megawatt auto transformer of this expansion development is now on site being installed and remains on track for energization in Q1 next year and the Core & Shell development of the first 2 buildings of our Phase 1 development plan, which will allow us to deliver full build-to-suit data centers in 2027.”
Later it was stated the Core & Shell development was the more time intensive aspect: “The first phase of construction of the Core & Shell is the most time-intensive but capital-light portion of the build-out with total expected development cost of $214 million, representing approximately $1.9 million per IT megawatt for the first 2 buildings.”
In total, Riot is aiming to develop 672MW of IT capacity at Corsicana across two separate phases, with this 112MW build part of its first 504MW phase across eight separate buildings. Riot did not provide an update on when it expects to complete this in full, rather stated the pace of development would be dictated by tenant demand.
For the first 112MW, Riot expects to spend ~$214 million on capex across the next six quarters, translating to ~$1.9 million per MW for the core and shell. Note that this does not include substation capex or land acquisitions, which are expected to be $18 million in Q4, taking total spend on both to ~$138.6 million for the full year.
Riot’s 1.86GW Offers Rebound on Bottom Line – But When?
Riot has 1.86GW of power permitted and readily available for use, with the company aiming to transition this entirely over to data center capacity when economically feasible. Riot lags peers with a smaller power pipeline than peers in the Miner Universe, with Applied Digital recently disclosing its active power pipeline tripling in two quarters to 4.3GW. Additionally, Galaxy has up to 3.5GW available, IREN has 2.9GW and Cipher has 2.4GW in its pipeline.
Similar to the numbers above, we had done some calculations on what the roughly 2GW is worth for Riot and came up with the following: “For Riot’s combined Corsicana and Rockdale facilities offering 1.7GW of available power, and assuming both are fully converted to AI with a 1.3 PUE for ~1.3GW of critical IT load, the two could be worth more than $23 billion for a 10-year deal structured at similar terms, or average annual revenue of $2.34 billion.”
When we look at 2030 analyst estimates, it seems to be pricing in most of the 2GW with estimates for $2.5B, up from $660M today. Therefore, I’m asserting that most of this has been priced in and there are better deals on the market – with the information I have today.
Brief Financial Update
Riot beat on the top and bottom line in Q3 with revenue up 112.5% YoY to $180.2 million, driven by 138% growth in Bitcoin mining revenue to $160.8 million. Engineering revenue was $19.1 million, up 51.1% YoY, while Engineering backlog rose 135.4% YoY and 34.5% QoQ to $159.6 million, with 90% originating from the data center sector. GAAP EPS was $0.26, ahead of estimates for $0.13.
For our perspective, the most important aspect for Riot boils down to cash and capex, as the company has not yet secured a data center tenant deal and for the moment will be fronting the capex for the powered shell itself.
Riot reported $330.8 million in cash and equivalents in Q3, while holding 19,287 Bitcoin worth nearly $2.1 billion (including 3,300 BTC held as collateral). Debt was $839.7 million, with $253.2 million current.
Capex is projected to be $153 million in Q4, with $131.6 million going to miner purchases and miner infrastructure, and the remaining going towards Corsicana’s substation, land and initial capex for the first 112MW. Riot says its key capex needs through year-end are fully-funded with cash on hand, yet current debt suggests the company will need to raise more cash sooner rather than later to progress with more phases at Corsicana.
Conclusion:
There is a disconnect between analyst estimates for Riot on the bottom line once Corsicana’s 600MW is delivered as the net operating income will provide a significant boost to the bottom line. We’ve seen this across the board where Miners rebound from being deep in the red from their Bitcoin mining operations to seeing healthy margin expansion.
However, this one is hard to time – it is a complete guess if Riot can deliver sooner than Q1 2027. In the meantime, we’ve identified some strong trends in play right now that we prefer to re-allocate to.
The Discovery tier offers fast-paced research on new stock ideas the I/O Fund is interested in, with technical setups and comprehensive deep-dive analysis. Be the first to know what exciting new tech, AI and energy stocks the I/O Fund is tracking.
To subscribe to Discovery with 30% off, please click here to email us or email premium@io-fund.com and mention code DISCOVERY30.click here to email usclick here to email us or email premium@io-fund.com and mention code DISCOVERY30.
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.
Galaxy’s Q3 was the best quarter in company history, per management, with the company seeing record crypto trading volumes, facilitating the sale of 80,000 Bitcoin worth more than $9 billion for a major client. Initial Data Center revenue was recorded in Q3 of just $2.7 million, though management explained that Data Center would be immaterial to results until the first half of 2026 when the first phase at Helios is delivered to CoreWeave.
Management had some interesting comments about being tied to CoreWeave when it comes to future financing and the economic opportunities around diversifying to possible hyperscale customers, as well as why they will not pursue a GPUaaS model like IREN. Management also teased a “big update” in Q2 next year for the newly-launched consumer-facing platform GalaxyOne.
Adjusted Gross Profit Surges in Q3
Galaxy reported revenue of $29.2 billion in Q3, though as we had explained in our initial analysis, Galaxy Bridges Crypto and Asset Management with AI Data Centers, this reflects gross principal trading (Digital Asset sales) that are subsequently offset by transaction expenses.
As such, adjusted gross profit is a better metric for investors to track, as it strips out the inflated appearance of the top line from Digital Asset sales, which were $28.8 billion in Q3 with transaction fees of $28.3 billion. The reported revenue figure does provide clues into the health of Galaxy’s crypto business, as Digital Asset sales surged nearly 231% QoQ, including the 80K Bitcoin trade, but even backing that out, this still would represent north of 100% QoQ growth.
Galaxy reported record adjusted gross profit in Q3, driven by the sharp uptick in Digital Assets and an improvement in adjusted gross margin in the segment, as well as strong performance in Treasury & Corporate. Data Center did contribute ~$2.7 million in adjusted gross profit in the quarter, though it will not become a more material driver until 2026 when revenue ramps as the first phase of Helios is delivered to CoreWeave.
Adjusted gross profit was $728.4 million this quarter, up more than 143% QoQ and 976% YoY. Adjusted gross margin was 2.5%, down from 3.3% in Q2 due to the much higher revenue base ($29.2 billion versus $9.1 billion).
Brief Updates on Digital Assets, Data Center
Galaxy provided some insights into its Digital Assets business, which saw its strongest monthly performance in July, bolstered by that 80K Bitcoin sale. Management said that this momentum carried through the remainder of the quarter, with a meaningful 140% QoQ increase in trading volumes, growth in its client base and increased institutional engagement.
Assets under management and under stake also saw strong QoQ growth, reaching a combined $15.4 billion, with $8.8 billion AUM and $6.6 billion AUS, up from $8.9 billion at the end of Q2. Sequentially, this represented growth of ~$3 billion in AUM and ~$3.5 billion in AUS. Galaxy explained that this level of AUM/AUS would represent more than $40 million in annual recurring revenue from fees, as management expects a majority of the new growth in AUM to be long-lasting, high-fee paying assets.
Galaxy also launched its first consumer-facing platform GalaxyOne in the quarter, with four initial product offerings: Premium Yield, offering 8% yield on cash for accredited investors; GalaxyOne Cash, an FDIC-insured high-yield cash account; GalaxyOne Crypto, support trading and transfers of select crypto products, and GalaxyOne Brokerage, support commission-free equities trading. This boosts Galaxy’s competitiveness with Coinbase and Robinhood in the realm of consumer-facing equity and crypto platforms.
Data Center Progress & Funding
Galaxy’s delivery of 800MW of contracted power to CoreWeave at Helios remains on track and on time, with the first 133MW phase set to be delivered in the first half of 2026, the second 260MW in 2027 and the third and final 133MW in 2028. Galaxy owns a 345kV substation at the campus, capable of supporting 900MW of capacity, with an adjacent 345kV substation owned by Wind Energy Transmission of Texas (WETT) under development to help meet future power needs.
Galaxy noted that it is also in queue with a handful of other interested parties for more power in Texas, and hinted that 800MW will not be their total powered footprint, with an additional 2.7GW under study at Helios, though the future pipeline may remain limited beyond that.
In Q3, Galaxy secured a $1.4 billion, three-year project financing facility to fund the $1.7 billion build-out of the first phase. Management stated that they have been drawing on the facility pretty regularly and on a straight-line basis, and by the end of the quarter the company had drawn ~$430 million so far, taking its total debt outstanding to $1.15 billion. Additionally, Galaxy closed a $460 million PIPE deal for the data center side, with the $325 million in net proceeds going to further fund the expansion of Helios beyond the first phase.
The project financing draw provides a rough idea of how much capital Galaxy will need through 2028 to support the entire 800MW capacity for CoreWeave. Assuming further financing needs come at similar terms and construction at similar costs, Galaxy may need to finance $4.2 billion in total, with $2.8 billion for the second phase. Galaxy now has $1.9 billion in cash and stablecoins on its balance sheet, up ~$700 million sequentially, providing ample room for self-funding the equity to secure financing for the upcoming phases.
Net Income and EPS
Driven by the surge in adjusted gross profit, Galaxy reported a very strong $505.1 million in net income in Q3, up a whopping 1,546% QoQ. Net margin on the reported revenue figure was 1.7%, up from 0.3% in Q2 and up from (0.4%) in the year ago quarter.
This translated to $1.01 in GAAP EPS, or $1.12 in adjusted EPS, or ~5x higher than consensus estimates for $0.21 in the quarter. However, it is important to note that this figure should not be annualized as it is likely to be reliant on the health of the crypto industry and trading volumes, which can vary wildly from quarter to quarter.
Adjusted EBITDA was $629.4 million, nearly 3x the $211.4 million generated in Q2. Adjusted EBITDA margin was 2.2%, down slightly from 2.3% in the prior quarter.
Earnings Q&A:
More Context and Clarity on Helios Build-out
Management provided an important update on the construction progress for the first phase of Helios, confirming that it remains on schedule, with the first data hall to be powered on in December of this year before reaching full operational status in the first half of 2026.
CIO Chris Ferraro explained that 70% of Galaxy’s concrete and civil work is complete and that equipment deliveries and installation is underway. He added that Galaxy is “now placing chillers and putting together the piping system that will form the backbone of our advanced liquid cooling design, an essential component to support next-gen GPUs at industry-leading cabinet densities. Our e-houses, which contain the critical electrical infrastructure have started to ship from the integrators and medium-voltage switchgear and transformers are already being set on their pads.”
With this progress, Galaxy expects the building to be sealed from weather within the next few weeks, allowing work on the interior to proceed regardless of outdoor weather. The next major milestone on deck is powering on the first data hall in early December, then shifting to preparing the hall for service.
For the second and third phases, management provided a quick update, stating that they are “proactively securing long lead time items like backup diesel generators and medium-voltage switchgear early, locking in cost certainty and delivery time lines.” This is important as it will help de-risk the build-out from an infrastructural viewpoint and mitigate potential impacts from equipment delays.
Update on Power
Considering that Galaxy has already contracted out all 800MW of its approved power, management provided a deeper look into power delays and connection requests in ERCOT’s grid.
Ferraro explained that “ERCOT's interim process and the level of scrutiny applied to large loads requesting to interconnect to the system has led to delays in additional capacity approvals” across Texas, as the state’s grid has been overwhelmed over the last 12 months and ERCOT does not want to take on projects that would risk destabilizing the grid.
Piper Sandler’s Patrick Moley questioned about approval timelines, considering Galaxy has a substantial 2.7GW awaiting approval, to which management responded that they will not have an exact timeline for approval until it happens and that “predicting the date is probably a fool's game.” However, they believe that signs from ERCOT, WETT and their utility partner AEP approving and finalizing studies at a faster pace give them a higher degree of confidence in approval for capacity coming online in late 2028 to 2029.
Management had mentioned that WETT’s Pitchfork 345kV substation is expected to “deliver an additional 3 gigawatts of power capacity with 2 synchronous condensers adjacent to the Helios campus starting in 2028,” which would support the entirety of the campus’ remaining buildout with some buffer room. Galaxy stated that there are “increasing proactive reach outs to us from very large customers in addition to our current partner, CoreWeave, who all want to know when are we getting approval for how much and over what time period,” so getting approval for the entire 2.7GW of additional capacity will unlock a tremendous opportunity to meet future demand.
Looking beyond Helios, management implied that there may not be much more in the pipeline until prices come down: “Markets for some of these companies without contracts, without customers, the market is pricing in a tremendous amount of optimism. And so that feeds through to the price of projects. And so in the short run, I don't think you're going to see us reaching out and buying a whole lot more power at these prices.” Considering the capital intensity of simultaneous multi-GW build-outs, this is a smart move as it would prevent Galaxy from overstretching its cash with too many builds.
CoreWeave’s Credit Quality as Anchor Tenant
Another interesting discussion circled around to CoreWeave as Galaxy’s primary anchor tenant, the economics of being this closely linked to CoreWeave and the transition to a multi-tenant site at full capacity.
When questioned about the potentiality of refinancing the project financing facility to unlock some capital, CIO Chris Ferraro explained that their ability to refinance this debt and raise future debt at attractive rates is linked rather directly to CoreWeave’s credit profile:
“The views of CoreWeave's credit profile, which lenders are very focused on, in addition to Galaxy's credit profile are changing and getting better by the day on both fronts. And so the ultimate outcome is really going to be a function of where we and CoreWeave and the markets are then.” He added that if rates were in the high single-digits and if Galaxy and CoreWeave remain successful, financing rates may more lower.
However, CEO Mike Novogratz seemed to contradict this, stating that the market is in a period of “trying to understand what CoreWeave's credit quality is today and what it should be on the forward. And that's going to be a big, big determinant of their ability to get better lease rates, their ability to get financing, [and the] ability for us as the landlord to finance our projects. And so where that goes, how the market evolves is thinking on CoreWeave is something we're very focused on and something that's a little unknown today.”
This is quite an important quote, as Galaxy’s financials and ability to secure future financing at attractive rates is inextricably linked to CoreWeave’s financial health. As we discussed in our free newsletter on CoreWeave, the neocloud is facing a wall of capex in Q4 and may continue to turn to the debt market to fund its expansion, likely weakening its credit profile. This could then hamper Galaxy’s ability to find attractive financing for its upcoming projects.
Considering CoreWeave is Galaxy’s only AI tenant for the time being, it raised a crucial point about diversifying the customer base in the future to a multi-tenant structure as Helios expands beyond 800MW. Novogratz stated that there is a “real decision to be made as to whether on a net economic basis, whether a lower-yielding lease from a higher credit quality tenant, net balances out to a better economic equation for us as we think about broadening the portfolio.”
What he is implying here is that an investment-grade hyperscaler will be able to secure a lower-yield lease, carrying a lower annualized opportunity to Galaxy, but because the creditworthiness of the hyperscaler exceeds that of CoreWeaves, Galaxy’s associated costs and funding will be more attractive and translate to a better return on investment (long-term revenue minus funding costs and interest expenses on debt).
Shying Away from GPU-as-a-Service
Unlike IREN, Galaxy is not going to prioritize or even attempt to shift towards renting GPUs out as a service, opting instead to take deals like the one it signed with CoreWeave. Galaxy said this is because they do not have the same software + hardware advantage that CoreWeave and others may have where they have a strong value-add proposition on top of the bare metal, and it is not something they have invested in.
Management also chalked this decision up to a lack of understanding over the useful life of GPUs, preferring to invest instead in long-lived infrastructure where their expertise is more suited:
“We’re not confident in what useful life of GPUs are ultimately going to be, and the cycles of GPU efficiency are pretty nascent still. And so we like very much investing in long-lived infrastructure that we understand useful life of, and we don't quite yet understand what the useful life of GPUs are. And so the business model around return on capital on GPUs, particularly if you haven't added real expertise in real value add, I think is a really challenging thing to decide to do. So we're not thinking about it.”
GalaxyOne: Expect “Big Update” in Q2 2026
Considering how recent the launch of GalaxyOne was, the impact on growth is likely to be minimal in the near-term as Galaxy works to build out the product suite. Management expects it to take a handful of quarters to get momentum on the platform, and tentatively expects to have a “big update” for investors in Q2 2026.
Galaxy teased that it has a “really ambitious” road map over the next six to 18 months to transform GalaxyOne into a “one-stop serves all wallet,” with the aim to add products and offerings that help high net worth consumers invest and store wealth, and reallocate across equities, bonds, digital assets and more. The platform also would open up an ability for cross-platform collaboration between trading, asset management and staking, and serve as a new outlet for Galaxy to continue growing its AUM/AUS and increase its fee-based ARR rather substantially.
Conclusion
There is a lot to like from Galaxy’s Q3 report, with adjusted gross profit surging from a strong crypto backdrop while data center updates show continued progress towards start of service for CoreWeave in 2026. Management remains focused on building out Helios and is encouraged by signs from utility partners and regulators about receiving approval for additional power to expand the campus later this decade.
Damien Robbins, Equity Analyst at I/O Fund contributed to this analysis.
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.
The saying that “good things come to those who wait,” is what Riot Platforms is banking on. The company is taking its time in preparing its two major sites for AI data center customers. Meanwhile, Bitcoin mining peers seem to be rushing toward deals, and their stocks have been popping off over the past few weeks as a result.
When asked in the earnings call what the hold up is, Riot explained they are looking to maximize the deal terms as they have one of the most attractive sites in the country (an analyst’s words regarding the Corsicana site, not theirs). Corsicana is especially attractive as it’s located near the Tier 1 market of Dallas and has ample power and water for liquid cooling. The Rockdale site is the second priority for securing a lease and will be used for Bitcoin mining in the interim.
Overall, Riot has stated they have “over 1.7 gigawatts of power immediately available near major markets.” Riot’s balance sheet also stands out with $2.4 billion in liquidity due to having $330 million in cash and 19,000 Bitcoin.
However, what investors are up against is navigating near-term volatility while a deal announcement could be another quarter (or two) away. This is because Riot is going through a master site design this quarter (ending now in Q3), which helps to start the discussions for a deal. It was stated the site design is what would kick off discussions with customers: “And if you're talking about engaging with serious counterparties, this is the type of information that they [want] you to come to the table with in order to advance discussions substantially.”
While Riot is currently in the final stages of designing the data center and working to secure a lease, it may be Q4 or later before a firm deal is announced. On the revenue side, if a deal is done in Q4, it will likely be more geared toward mid-year 2026 or the second half based on current timing for substation development at Corsicana.
In the opening remarks, Riot explained they are in no rush: “To be clear, we are not pursuing a so-called pivot into AI HPC initiatives with a view of doing a "deal.” Rather, we have added a new data center development capability, which we will apply to as much of our power portfolio as possible and which will transform our company in the years to come.”
Riot’s 8-Step Process for AI Data Center Pivot
Riot has outlined its eight step process behind its AI data center pivot, with the company now focusing fully on the sixth, seventh and eight steps:
Engage consultants: Riot tapped data center industry consultant Altman Solon in January to explore feasibility of AI/HPC pivot at Corsicana
Add experience to board: Riot added three new board members in February with relevant data center experience
Engage financial advisors: Riot tapped Evercore as a financial advisors and Northland as a co-placement agent to lead discussions with possible AI customers
Continue infrastructure development: Pertains to the development of the 600MW substation at Corsicana
Build internal expertise: Riot stated it is making key hires to build out its internal data center expertise
Complete basis of design: Riot is leveraging new hires and external consultants to best design data centers that fit its available infrastructure
Secure lease with tenant: Riot says it is “actively progressing” towards securing a lease with a “high-quality” tenant and collaborating on final design iterations
Build future pipeline: Pertains to further expanding its AI data center footprint in the future with powered land acquisitions and development
While Riot is currently in the final stages of designing the data center and working to secure a lease, it may be Q4 before a firm deal is announced. On the revenue side, it will likely be more geared toward mid-year 2026 or the second half based on current timing for substation development at Corsicana.
Riot’s Corsicana Data Center Can Offer 600MW for AI Data Centers
Riot is working to expand its Corsicana data center, purchasing 355 acres of land in May and an additional 238 acres in July, taking its total acreage to 858 and allowing it to “accommodate various data center designs and development plans.”
Corsicana has 1 GW of power capacity, with 400 MW currently operational and earmarked for Bitcoin mining, with the additional 600MW of power capacity for AI/HPC hosting is expected to be ready in early 2026. The facility also has two fiber providers on-site, offering ample connectivity and low latency to major hubs Dallas and Austin. Riot also disclosed that its all-in power costs at Corsicana (for BTC mining) were ~$0.035/kWh, matching IREN with industry-low power prices.
The entire 1GW is expected to be available sometime over the next 4-5 quarters: “So first, for some of the critical infrastructure that's needed to build this capacity, we have already secured. I'm referring to the 600-megawatt substation that's being built that's expanding the site to 1 gigawatt. We have already procured that equipment. That equipment is already arriving, and that is going to take our Corsicana site to 1 gigawatt in 2026.”
What’s unique about Riot is they already have approval for power to both sites, whereas newer sites are subject to legislation like Texas Senate Bill 6, which requires Bitcoin miners to cover some of the costs of expanding the grid. There are also curtailments to where miners/HPC operators have to reduce usage when the grid is maxed out and are required to have backup power. The company stated they will be able to power the pivot to AI without any further steps: “This power isn't pending certain steps happening. This power is coming in the next 6 months and scaling up from there.”
In Q4 FY24, Riot estimated it would spend ~$65 million in capex on developing the 600MW substation and acquiring other long-lead items required to energize this phase. Through the first half of FY25, Riot spent $51.3 million in capex related to the substation, and projected an additional $17.9 million in spending through Q4, or ~$69.2 million in total capex, a 6.5% overrun. Riot’s Q4 acquisition of E4A likely plays a critical role as the company is engaged in substation equipment, transformer assembly and power plant servicing. Additionally, the mid-year land purchases, water lines and other site costs are expected to total approximately $77 million this year.
The company is evaluating more expansion opportunities for AI/HPC, and assessing feasibility for developing further power capacity, but has not made any strides outside of Corsicana yet. Combined with its 700MW Rockdale facility, Riot can offer 1.7GW of power in close proximity to Tier 1 Dallas and also Austin, though Corsicana’s first 400MW phase and Rockdale are presently fully outfitted for Bitcoin mining and will be retrofitted in phases when management believes the timing is right.
How Much Could Corsicana be Worth?
Riot is well aware of the opportunities in pivoting to AI data center hosting, more so now considering the flurry of AI deals that have been signed with miners recently. As we pointed out in our Advanced analysis in January, Marathon Digital and Riot Platforms: Leveraged Bitcoin Bets, Riot’s management had said in Q3 2024 that there is “notable sense of urgency for power access in 2025 with AI HPC companies willing to pay a premium for timely access at attractive sites.” Riot also stated at the time that it was in preliminary discussions with AI HPC firms over some level of capacity, and it would see if there are deals to monetize capacity at a better rate than mining.
As of now, Riot has yet to sign an AI hosting deal, though analysts from Needham expect the company to be in advanced discussions by Q4 and sign a lease agreement as early as Q1 2026. This would be a few quarters behind some of the earlier movers in the industry, but this hinges on its Corsicana facility, with its AI side still under development. Notably, Riot likely will not transition its entire footprint over to AI hosting, stating in its 10-Q that is in developing a “scalable data center platform designed to allocate a portion of our power capacity toward supporting AI/HPC workloads.”
Despite the lack of a deal, allocating 600 MW of power capacity to AI (assuming ~440 MW of critical IT load as Needham estimates) allows a reasonable projection for what Corsicana could be worth under similar terms as other miners.
Assuming a deal structured around $1.8 million in average annual revenue per MW of critical IT load, Corsicana’s AI data center side could fetch a deal worth ~$8 billion over ten years, or ~$792 million in average annual revenue. Other miners have outlined targeted net operating income or adjusted EBITDA margins on these deals of >85%, suggesting that Corsicana could deliver adjusted EBITDA or net operating income of ~$673 million annually on average. As a reminder, these are average annual figures, meaning that revenue in the initial stages will begin at a much smaller size as capacity ramps and end much larger than these stated figures.
For Riot’s combined Corsicana and Rockdale facilities offering 1.7GW of available power, and assuming both are fully converted to AI with a 1.3 PUE for ~1.3GW of critical IT load, the two could be worth more than $23 billion for a 10-year deal structured at similar terms, or average annual revenue of $2.34 billion.
Bitcoin Mining Expansion Remains on Track
On the flip side, Riot is still committed to expanding its Bitcoin mining fleet and hash rate, even though the Corsicana pivot reduced its 2025 deployed hash rate target by 8 EH/s.
In Q2, Riot updated its year-end hash rate target to 40 EH/s, up from 38.4 EH/s, likely driven by its recent addition of 125 MW of capacity for Rockdale by acquiring additional assets from Rhodium and closing pending litigations. This represents ~26% YoY growth in Riot’s deployed hash rate, aligning with Riot’s target to maintain ~4% of the global network, currently at ~984 EH/s. By Q1 2026, Riot is aiming to bring its hash rate up to 45 EH/s, what would represent its largest sequential expansion in hash rate since Q3 2024 as it works to double its power in Kentucky to 127 MW.
Rockdale will remain a Bitcoin mining operation until Corsicana finds a customer. Riot positioned it as a strength to keep Rockdale as Bitcoin mining operations for now, stating: “So you can think of our strategy as using Bitcoin mining at sites like Rockdale to monetize that power to ensure that no power stranded and wasted, turning that into meaningful cash flows for the company and then ultimately looking to transition that capacity to data center leases when the time is right.”
Financials
Bitcoin Mining Remains RIOT’s Core Revenue Driver
Riot’s Q2’25 results reaffirm its position as one of the largest self-mining operators in the world. The Company reported total revenue of $153 million, more than doubling from $70 million in the year ago quarter. Most of this revenue came from Bitcoin Mining segment, which contributed $140.9 million, up 152% year-over-year. The drivers here are straightforward: Riot produced significantly more Bitcoin, at higher prices. On top of that, Bitcoin itself averaged roughly $98.8k during the quarter compared to $66k last year. The combination of higher volume and higher pricing powered a sharp rise in top-line contribution.
Average operating hash rate increased from 11.3 EH/s last quarter to 31.7 EH/s this quarter, while deployed hash rate reached 35.4 EH/s by quarter-end. Think of hash rate as Riot’s “factory horsepower”. The more horsepower it has online, the larger its share of Bitcoin production – but it also burns more energy which makes power contracts and efficiency just as critical.
Engineering revenue, though a fraction of mining, also provided a small boost. The segment generated $10.6 million in revenues, up modestly from $9.6 million last year. This now includes contributions from December 2024 acquisition of E4A Solutions, which expanded Riot’s capabilities in custom electrical and fabrication work. Engineering also carries $22.7 million of contract liabilities, suggesting a healthy pipeline of projects yet to be recognized.
“Other’” Revenue totaled just $1.5 million, largely residual income from legacy hosting contracts that have since been terminated. Importantly, Riot no longer reports hosting a standalone segment, underscoring the Company’s shift to pure self-mining plus a smaller engineering team.
Looking forward, consensus expects continued topline growth through FY25 and into FY26, albeit at a decelerating pace. Seen in the chart above, analysts expect 99.6% and 30.47% YoY growth in Q3 and Q4 to round out FY25. Estimates moderate looking further into FY26 with growth drifting off to mid-teens (17.6%) by Q4’26.
For investors looking at AI upside, there are no contracts signed and no AI revenue recognized in Q2. The AI narrative remains aspirational, not actual. Compared to peers like Applied Digital and Terawulf, both having already signed multi-year, multi-billion dollar hosting agreements with hyperscalers, Riot is behind on the diversification curve. At this point, AI should be viewed as upside optionality embedded in Riot’s power infrastructure, rather than a near-term driver of financial results.
Key Metrics:
Total Revenue: $153.0M (+119% YoY)
Bitcoin Mining: $140.90 (+152% YoY)
Engineering: $10.6M (+10% YoY)
Other: $1.5M (hosting residual)
Avg. BTC price: $98.8k (vs ~$66.6k YoY)
Gross Margin Expansion Driven by Healthy Mining Economics, Operating and Net Margins Inflated by Crypto Gains
Mining economics remained healthy. Cost of revenue in the Bitcoin Mining segment of $78.2 million implying mining gross profit of $62.7M and margin of 44.5%. Power accounted for the majority of costs at $62.2M, 79% of total segment costs. Other items included $5.7M in taxes and fees, $4.7M in compensation, $4.1M in miscellaneous operating expenses, and $1.5M in insurance.
All-in power costs rose from 2.1 cents /kWh last year to 3.5 this quarter as Riot added Corsicana and Kentucky to its portfolio. Even with this increase, Riot remains one of the lowest-cost operators in the sector. Rockdale’s 345 MW PPA remains a key differentiator, locking in fixed-price power through 2030 with the ability to resell curtailed load back into ERCOT during peak demand.
On a GAAP basis, Riot’s margins were eye-popping. Operating income was $216.1 million, translating to an operating margin of 141% compared to negative 167% last year. Net income of $219.5 million equated to a net margin of 143%, compared to negative 121% a year ago. As always, context matters. Riot’s operating and net margins reflect the heavy influence of new Bitcoin accounting rules.
Prior to 2025, Bitcoin and other digital assets were classified as “indefinite-lived intangible assets” under US GAAP, meaning that Company’s had to carry them at cost on the balance sheet and test annually for impairment. If BTC prices dropped below carrying value, an impairment charge was booked. If BTC prices recovered later, you could not mark the assets back up, leading to asymmetric accounting – lots of impairments, but no mark-ups.
Effective January 2025, the Financial Accounting Standards Board (FASB) implemented a new rule which requires digital assets like Bitcoin to be carried at fair value, with changes flowing through net income. This means that each quarter, Company ‘s must revalue their BTC to market prices as of balance sheet data, causing non-cash gains / losses that distort operating income and GAAP figures.
In Q2, this resulted in a $262.8 million “Change in fair value of Bitcoin” gain flowing directly through the income statement. This single item more than explains the swing to profitability. Without it, Riot’s core operating performance looks much thinner. Further, operating profitability was also pressured by Depreciation and amortization ($83.2M), stock-based compensation ($59.7M), and unrealized losses on securities reached ($57.1M).
GAAP EPS came in at $0.65 basic and $0.58 diluted, versus ($0.32) loss in Q2’24. Looking forward, analysts do not expect annual improvement toward profitability in FY25 or FY26, as EPS estimates are ($0.35) and ($0.48) in FY25 and FY26 respectively. GAAP EPS is not a reliable measure of Riot’s underlying profitability. Investors should instead focus on cash operating margins and Adjusted EBITDA to understand the true economics of the mining business.
Key Metrics:
Mining gross profit: $62.7M
Mining gross margin: 44.5%
Operating income: $216.1M
Operating margin: 141% (vs – 167 YoY)
Net Income: $219.5M
Net Margin: 143% (vs –121 YoY)
Basic EPS $0.65
Diluted EPS: $0.58
Cash Flow and Balance Sheet
Despite reporting net income, Riot’s operating cash flow was sharply negative at ($231.3 million), compared to ($122.1 million) in the prior quarter and ($42.5 million) in the prior year. That equates to an OCF margin of (151%). That disconnect again stems from the way Bitcoin revenue is recognized. Riot books revenue at the fair value of BTC when mined, but no cash is received unless those coins are sold. In Q2, Riot mined about $283.7 million worth of BTC, inflating revenue and net income, but none of that translated into cash inflow.
Working capital shifts added further drag. Accounts Payable fell by ~$78M, deferred revenue declined ~$34M, and accrued liabilities decreased ~$12M, all of which consumed cash, resulting in negative OCF even as GAAP profits surged.
On the balance sheet, Riot remains well capitalized but increasingly levered. As of June 30, 2025, the Company held $255.4 million in cash, plus $74.9 million in restricted cash. Bitcoin holdings amounted to $1.71 billion unrestricted and $353.7 million restricted, giving Riot one of the largest corporate Bitcoin treasuries globally. Total assets stood at $4.29 billion, offset by $838.4 million in debt and equity of $3.30 billion.
While liquidity is ample, leverage has crept higher as Riot continues to finance its large-scale buildouts in Corsicana and Kentucky. ATM equity issuances also remain part of the funding toolkit, with share count rising in the quarter form new issuances and an acquisition-related issuance. The trade-off here is clear: Riot has the liquidity to scale, but at the costs of higher leverage and dilution.
Capital intensity remains high. Riot has $83.5 million in remaining commitments to MicroBT for miner purchases, expected to be paid in 2025. During the first half of the year, the Company placed $86.4 million in deposits on new equipment and added $93.2 million of property and equipment. These figures highlight the scale of ongoing investment required to maintain Riot’s leadership in hash rate capacity:
Litigation is a non-trivial overhang. Riot is facing a lawsuit seeking nearly $500 million from GMO, a Japanese IT and mining company, related to termination colocation agreement. GMO alleges that Riot wrongfully terminated the agreement and is seeking damages. Riot disclosed the case it’s in legal proceeding footnote but stated it cannot reasonably estimate the outcome or probability of loss at this time.
This claim is significant: ~$500M represents almost 15% of total assets ($4.29B) and is more than double the Company’s quarterly income. While Riot has a strong equity base and deep liquidity, an adverse ruling would be material. Notably, peers like APLD or IREN don’t face comparable litigation exposure, making Riot’s risk profile a bit heavier by comparison. Investors should treat this as a tail risk that could swing sentiment if unfavorable developments occur.
Bulls vs Bears
The Bull Case:
Scale and Cost Advantage: Riot operates one of the largest fleets in North America with over 35 EH/s deployed and maintains a competitive power cost of ~$3.5 cents per kWh, keeping it at the low end of the industry cost curve.
Balance Sheet Optionality: With $255 million in cash and $1.7 bitcoin of Bitcoin holdings, Riot has liquidity unmatched by most peers, providing downside cushion and upside leverage to BTC prices.
AI / HPC Upside: While not yet monetized, Riot is positioning its multi-gigawatt power platform for future AI/HPC hosting, offering optionality that could re-rate the stock if contracts materialize.
The Bear Case:
Accounting Noise: GAAP earnings are dominated by Bitcoin fair-value revaluations, making reported profitability a less useful metric that is volatile and disconnected from underlying cash economics.
Litigation and Leverage: The $496 GMO lawsuit and rising debt levels ($838M) create potential balance sheet and legal overhangs, especially if BTC retraces.
Lagging AI Execution: Unlike peers APLD and WULF, RIOT has yet to sign meaningful AI/HPC deals, leaving it more exposed to Bitcoin cycles with less diversification.
Conclusion:
Analysts have called Riot’s Corsicana data center site one of the most attractive in the country, as it’s located near the Tier 1 market of Dallas with ample power and water for liquid cooling. Riot is taking its time in preparing the 600 MW site for AI data center customers, and it may be Q4 or later before a firm deal is announced. This timing dynamic, possibly months behind other mining peers who have already secured deals, opens the door for potential near-term volatility.
Damien Robbins, Equity Analyst at I/O Fund contributed to this analysis.
Every Thursday at 4:30 pm Eastern, the I/O Fund team holds a webinar for premium members to discuss how to navigate the broad market, as well as various stock and crypto entries and exits. Beth Kindig offers weekly deep dives including lesser-known cryptocurrencies and AI stocks, plus the team offers trade alerts. The I/O Fund team is one of the only audited portfolios available to individual investors. If you’d like to subscribe to the Advanced Market Signals plan, email us at premium@io-fund.compremium@io-fund.com.
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.
Galaxy is a crypto infrastructure company that is branching out to power AI data centers. Unlike the other Bitcoin miners we have covered recently, Galaxy is deeply entrenched in the crypto industry and will continue its crypto operations while expanding to build an AI campus in Texas called “Helios.”
On the crypto side, Galaxy offers trading, crypto wallet, custody staking and tokenization. They also offer investment products including a treasury for Solana and recently handled one of the largest trades in Bitcoin’s history at $9 billion. They provide infrastructure for popular Layer-1 crypto company Solana, including being a node validator and helping both Solana and Ethereum to offer liquid staking to institutions. In total across all of the funds and investment products, Galaxy has $9 billion AuM.
When it comes to data center capacity among the Miners, Galaxy is pursuing one of the largest with a planned 3.5GW of data center capacity. However, management stated they will not see a big impact on revenue until 12-24 months from now: “I do think, to highlight what Chris said, this next 12 months is really important because we go from putting out a lot of money to having an asset that starts spitting out cash. And once 12 months from now, 24 months from now, Helios becomes a big cash generator for us, it allows a lot more flexibility in how we deploy capital.”
There will be an initial impact on revenue come Q1 2026. CoreWeave is the only customer for Galaxy at this time with the company beginning to recognize revenue at the start of next year: “As a reminder, in the Data Center segment, we do not expect to report financial results until Q1 of 2026 when we begin recognizing revenue from CoreWeave under Phase 1 of our lease agreement.”
There are puts and takes to Galaxy having a successful, core operation in crypto while also expanding into AI energy. The company only recently began trading on the Nasdaq after being on the Toronto Exchange (TSX) for about five years. The operating margin of 3.4% is better than other Miners due to crypto infrastructure and investment services that offset any losses from mining. However, this may be inconsequential over time given data center deals are high margin. One could argue the lack of focus may make it difficult to compete in two highly competitive markets.
As discussed further in the call, Galaxy will become more attractive once it diversifies beyond CoreWeave and/or once we are nearer to more substantive delivery of the Helios campus.
Below, we look at what separates Galaxy and a few items we’d like to see before initiating a position.
Helios Campus is Massive, and is a Top 5 or Top 10 Data Center Site
Galaxy’s Helios offers 3.5GW of power with management stating this likely qualifies them for a Top 5 data center site: “We executed and bought another 1 gigawatt of potential capacity around the Helios site. That would make that a 3.5-gigawatt site, got to be one of the top 5 data centers in the world if we get that fully built out.”
Here is the timing for how that’s expected to come online:
CoreWeave is committed to the 800MW currently approved at the site. This is considered Phase 1, Phase 2 and Phase 3.
Of this Phase 1 is expected to deliver 133MW “throughout the first half of 2026.”
The company is moving toward breaking ground on Phase 2
Timeline is not clear on the additional GWs yet but it was stated there would be an acceleration of sorts on ERCOT approval to where the additional 1GW recently added would likely be approved on the timeline of the 1.7GW that has been under approval. Here is what was stated: “So I would think about the additional 1 gigawatt of interconnect request at the land we are acquiring to be somewhat on par with our existing 1.7 gigawatts in the backlog. And so think about 1.7 gigawatts going to 2.7 gigawatts in terms of timeline.”
When you combine the fact that management has stated Helios will become a big generator of cash in 12 to 24 months from now, it’s likely they are referring to the 1.7GW to 2.7GW being built out and energized.
One thing that stood out in Galaxy’s commentary was (at this time) they do not have other customers in the pipeline beyond CoreWeave, stating “I think our partnership with CoreWeave is going to take up the vast majority of our attention over the next few years.”
Funding Remains the Hurdle
By now, you may be noticing a theme which is that real estate is the easy part, and even access to GPUs is easing, yet powering the data centers and funding these massive projects remain the bigger hurdles.
Galaxy ended Q2 with $2.6 billion in equity capital, up more than $700 million sequentially. This increase was fueled by the $480 million primary capital raise in May, appreciation in digital assets and balance sheet investments, and a $292 million one-time accounting adjustment from the corporate reorganization. This equity raise is allocated for Phase 1 Helios (133 MW) and project debt financing is nearly complete.
While Phase 1 is essentially de-risked from a funding standpoint, it represents just a fraction of Helios’ full 3.5GW potential. At management’s estimated capex of $11-13 million per MW, Phase 1 requires roughly $1.5 billion of total spend. Scaling Phases 2 and 3 to reach the multi-gigawatt buildout will demand tens of billions in additional capital, well beyond the cash and balance sheet equity Galaxy has today. Management has flagged private equity partnerships and refinancing strategies as potential levers, but the implied cost of capital (~10-11%) underscores the financing challenges ahead.
The takeaway is that the Phase 1 financing is locked in, but the enormity of capital required for subsequent phases remains a gating factor for realizing Helios’ 3.5GW vision.
Galaxy has secured funding only for Phase 1 thus far, leaving the far majority of the 3.5GW open in terms of how to raise the capital to expand the Helios site.
Here is what was shared on the earnings call: “As a reminder, the equity portion of Phase 1 has already been funded through our existing equity capital. Once we have secured the project level debt financing, we will have the capital necessary to fund the anticipated CapEx for Phase 1 of approximately $11 million to $13 million per megawatt. For Phase 2, we are still finalizing the design and engineering specifications, but expect the total project CapEx to be slightly higher than the Phase 1 on a per-megawatt basis. We have already commenced work on project level debt financing for the Phase 2 project. Throughout the Phase 1 financing processes, we've established strong relationships with a wide range of banks and private credit managers who are active in the space, and I have confidence in our ability to secure debt financing for Phase 2 in the coming months.”
The company called out potential private equity deals as one source of capital, with the goal of refinancing mid-project as terms become more attractive. With that said, the rates the company is seeing now were stated to be around 10%, which is quite steep.
“Yes. James, so I think our expectation on where we land on Phase 1 is in line with what we've articulated in the past. It will come out at a sub-10% stream rate. But when you take into account upfront fees and potential breakage, depending on when you assume we'd have a refinancing event or not, we'll likely end up in the 10% to 11% expected yield in terms of cost of capital, even as credit spreads are tightening in real time.”
Brief Overview of the Crypto Operations
Galaxy’s crypto offerings are multi-faceted as the company is deeply entrenched in the crypto industry. Primarily, of the many different crypto products and asset management solutions the company offers, the tokenization of assets could be the most lucrative. Galaxy is one of the largest infrastructure companies by staking weight on the Solana platform, where the tokenization of assets is expected to do quite well. Solana accounts for 46% of on-chain trading volumes compared to Ethereum’s 22%.
According to the earnings call, Galaxy saw the strongest month in the company's history in July for its digital assets business. Part of this was representing a large transaction worth 80,000 Bitcoin for a value of $9 billion.
Crypto trading desk that offers both spot trading and derivatives. The company is said to have over 1400 institutional partners that trade on its platform.
It’s OTC derivatives platform saw $20 billion in volume last year.
The company is also a large crypto lender with $1.1B in loans in Q2 2025.
The company recently integrated with Fireblocks, a digital asset security platform with 2,000 institutions to help broaden access to Galaxy’s staking solutions.
Galaxy asset management oversees $4.7B across ETFs and other alternative/venture funds. In the case of asset management, it was stated that Galaxy makes a 1% fee.
Six days ago, it was announced that Galaxy will provide the asset management and treasury service for a $1.65B Solana treasury. A primary goal of the treasury is to help integrate Solana into traditional finance markets. It’s expected that Galaxy will provide treasury services for additional assets, as well.
The company was an early investor in Solana and recently tokenized Galaxy’s stock on the Solana blockchain.
Galaxy is also partnered with Invesco as they launched a joint spot Bitcoin ETF in 2023-2024 and have filed for a Solana spot ETF.
Financials Overview:
Galaxy Digital is evolving into a hybrid model: part crypto financial services platform, part AI / HPC infrastructure developer. Unlike some of the other miners (APLD, WULF, IREN) that have pivoted hard into AI data center hosting, Galaxy is choosing to maintain its crypto roots while simultaneously building one of the largest AI campuses in the world. This dual-track approach creates both unique optionality and added complexity. This shows up most clearly in the financials, where Galaxy’s reported revenue is inflated by crypto trading inflows, unlike peers whose top lines are directly tied to contracted hosting revenue.
Revenue Optics vs. Economic Reality
Galaxy’s revenue optics differ sharply from peers, and it is critical to distinguish between reported revenue from its assets under management (AuM). Reported Q2’25 revenue was $9.1 billion, (-30% QoQ, flat YoY), but this figure primarily reflects grossed-up digital asset sales that are subsequently offset by transaction expenses. By contrast, Reported AuM of roughly $9 billion reflects capital managed or staked on behalf of clients across asset management products, staking platforms, and funds.
While fee income from this pool was just $17 million in Q2, AuM remains an important context metric because it shows the scale of Galaxy’s crypto franchise and potential fee base as markets expand. For peer context, APLD, WULF and IREN have no comparable AuM base, as their reported revenue reflects directly contracted hosting or mining economics.
As Management cautioned, “top line is overshadowed by gross principal trading, investors should focus on Adjusted Gross Profit (AGP) instead. On that basis, Q2 AGP of $299 million marked a swing from ($204) million in Q1’25. Given the hybrid model, Galaxy further bifurcates AGP at the segment level:
Digital Assets: $71 million, up 10% QoQ
Treasury & Corporate: $228 million, driven by market-to-market gains on crypto holdings and balance sheet investments.
Data Centers: No revenue yet: Phase 1 lease with Coreweave begins in Q1’26
Where Galaxy diverges from peers is that once Helios begins contributing, the step-change in recurring revenue could be massive:
Company guidance: For the first 393 MW (Phase I + II), Galaxy disclosed ~$900 million in average annual revenue over the 15-year term, with >$700 million expected in the first full year of energization.
Implied economics: $900M / 393 MW = ~$2.3 million revenue per MW per year.
Scaling up:If the 526 MW of contracted / committed IT load (Phases I – III) is built, this would imply ~$1.2 billion in annual revenue at similar economics. If the full 800 MW of gross approved power were contracted on similar terms, that could reach ~$1.8 billion per year.
Contrast this with:
APLD: Q4’25 revenue of $38 million was entirely from hosting, backed by $11 billion in contracted CoreWeave HPC lease (~15 years of ~$730 million per year, which scales to ~49x FY25 revenue)
WULF: Began recognizing AI lease revenue is Q3’25 from its Core42 partnership, with early revenue flow already diversifying away from BTC mining.
The takeaway here is that Galaxy’s reported revenue is inflated by trading flows, but its economic revenue base ($299 million) is far smaller and more volatile. Peers like APLD, WULF, and IREN are already stacking up contracted, recurring HPC revenue, whereas Galaxy’s step-change won’t show up until 2026. When it does, back-of-the-envelope math suggests $900 million to $1.2 billion annually from CoreWeave alone, making it one of the largest HPC lease arrangements in the space.
Margins Inflect but Volatility Remains
Operating income improved to $166 million vs. a ($392) million loss in Q1. By comparison, APLD and WULF are still reporting large operating losses as they front-load HPC build outs. These peers are loss-making today but will structurally expand margins as long-term leases begin flowing. Galaxy is already GAAP profitable, but primarily due to market-sensitive mark-to-market gains.
Net income landed at $30.7 million, compared to losses of ($295 million) QoQ and ($126 million). Its worth noting that these margins remain razor thin: GAAP net margin of 0.3% versus APLD’s recurring hosting margins in the 20 – 30% range. Drivers of the swing:
$135 million digital asset gains and $195 million investment gains, with Bitcoin up 30% QoQ and Ethereum up 36% QoQ
Offset by $127 million impairments and a $125 million loss on derivative notes.
While peers will see margin expansion as AI leases ramp, Galaxy’s profitability remains tied to crypto market cycles until Helios revenue turns on.
EPS Turns Positive on Market Gains
GAAP EPS was positive in Q2’25. Management introduced Adjusted EBITDA as a clearer profitability lens, arguing this better reflects operating trends in Digital Asset and Treasury, removing derivatives noise. Like APLD, Galaxy’s non-GAAP metrics show progress. However, investors should be aware – Galaxy’s adjusted figures are still market sensitive and not contracted.
Balance Sheet and Cash Flow
Galaxy’s balance sheet is both its strength and its complexity, as it is large in scale but more volatile than peers due to crypto exposure. As of Q2’25, Galaxy held $1.18 billion in cash / stablecoins and $2.0 billion in net digital assets, giving it more self-funding capacity than miners reliant on continuous external financing.
Q2’25 Balance Sheet Highlights:
Total Assets: $9.08B, up 43% QoQ from $6.34B
Total Equity: $2.62B, up 38% QoQ from $1.90B
Cash & Stablecoins: $1.20B, flat QoQ
Balance Sheet Net Digital Assets: $1.28B, up 40% QoQ from $908M
Balance Sheet ventures, Fund, and Other Investments: $718M, up 15% QoQ from $623M.
Net Income of $30.7M vs. ($295M) in Q1’25.
Adjusted EBITDA of $211M vs. ($290M) in Q1’25.
These figures underscore Galaxy’s liquidity advantage today, though the reliance on digital assets and investments make them inherently more sensitive than contracted revenue streams. The crux here is that crypto assets create valuation swings that peers with cleaner hosting-focused balance sheets don’t face to the same degree.
WULF: Similar leverage profile to APLD, funded via staged equity and project-level partnerships for its Core42 build-outs.
IREN: Cleaner balance sheet with low net leverage and industry-low power costs, though at a smaller absolute scale.
Conclusion:
Galaxy Digital’s Q2’25 is a case study of complexity vs. visibility. Bulls would say that GLXY provides unique optionality, allowing investors to gain exposure to both crypto financial services (trading, custody, asset management) and one of the largest AI campuses globally (Helios at 3.5GW). Compared to peers, Galaxy also has superior balance sheet flexibility (vs. APLD and WULF) and is already showing GAAP profitability.
Bears would argue that earnings quality is lower due to market sensitivity, recurring fee revenue is declining, and Coreweave remains the sole data center tenant. Execution risk associated with ERCOT approvals and multi-phase financing is material. Ultimately, Galaxy is a barbell play consisting of crypto upside coupled with AI optionality. While APLD, WULF, and IREN are purer HPC pivots, investors must decide whether Galaxy’s dual exposure should be considered a healthy form of diversification, or distraction.
In the near-term, upward price action could be driven by Solana’s price, tokenization of assets, or stablecoins. As we go along into 2026, additional positive price action could occur if Galaxy secures a larger hyperscaler or offers visibility on how they will fund the 3.3GW that are left to fund beyond Phase 1. However, there are an equal number of reasons Galaxy and other miners could see volatility as they all greatly depend on high beta being in favor in the broad market. For now, the bulk of a decision on entries and exits will be made with technicals.
Note, this is a momentum stock and if we were to enter the positions, we plan to adhere strictly to risk management.
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.
