Summary TeraWulf's pivot to HPC hosting is proving successful, securing a $3.7B contract with Fluidstack, backed by Google, and building a strong backlog. TeraWulf's asset-light strategy focuses on infrastructure rather than GPU ownership, minimizing depreciation risk and ensuring predictable, high-margin revenue streams. Signed contracts, clear unit economics, and secured power capacity provide strong visibility into future cash flows and support a Buy rating upgrade. Despite recent price surges, I see a 2.5x upside potential for WULF as HPC revenue ramps, with manageable execution risks and a positive risk-reward profile. TeraWulf Overview TeraWulf Inc. (NASDAQ: WULF ) might well be considered a late mover among the Bitcoin ( BTC-USD ) miners in high performance computing [HPC]. They only began any serious HPC pivot last year when they announced (in their September 2024 monthly production update ) the successful completion of their 2MW proof of concept project called “Wolf Den” to showcase their HPC hosting capabilities. Despite the late pivot, TeraWulf is now targeting the right opportunities and building a strong HPC backlog. The latest $3.7 billion contract with Google ( GOOG )-backed Fluidstack is a prime example of well-timed execution resulting in successful contracts. There are several things TearWulf has done right to get here. One of them being that their HPC pivot strategy has been less exposed to GPU hardware lifecycles and depreciation because CaPex allocated to HPC has been focused more on power and infrastructure, which retain value longer than GPUs. TeraWulf does not maintain a large, depreciating fleet of GPUs for revenue generation. Spendings on GPU for the company’s Compute business line recorded so far was when the proof-of-concept Wolf Den was being built, which involved operating a compact Nvidia A100 GPU system at the Lake Mariner Facility. And this strategic execution was the highlight for me when I last covered WULF in May . Looking back, I think the more cautious Hold rating was overly conservative, and a Buy would have been a better call. I believe TeraWulf is now at the "right place at the right time" in HPC hosting. Among the miners, TeraWulf can be considered a late bird in HPC hosting. I believe this has allowed TeraWulf to craft a better strategy for the WULF Compute business line. Sometimes it’s the ones who come late to the party that leave with the best gains, after watching the early guests spill their drinks. In its HPC approach, TeraWulf is avoiding the operational complexity that purchasing and maintaining GPUs for HPC hosting presents, and will be focusing on providing infrastructure to HPC clients – power, cooling, rack, and space – as the management has iterated. TeraWulf’s HPC hosting is basically: bring your GPUs and we’ll host them. - Excerpt from my WULF coverage in May. Bitcoin miners who pivoted to HPC earlier took the GPU risk and will likely incur the depreciation. I think it is risky enough for a Bitcoin miner to be exposed to crypto price cycles, then to compound that with further exposure to hardware cycles doubles the risk. Seeking Alpha WULF is up 180% since visiting it on the release of the Q1 earnings in May, and is up ~59% since the Fluidstack-Google announcement yesterday. For investors who may have been on the sidelines, I believe the main question now is whether conditions support for more upside (from a fundamentals perspective), or if entry at this stage would be a case of an overhang of missed buying opportunity that presents the risk of catching the top, since WULF is already up a lot. WULF Still Presents a Compelling Buy Despite Its Price Surge Without beating around the bush, I strongly believe that the outlook for TeraWulf has changed greatly for the better, and I am upgrading my stance to Buy at this point. The pivot to HPC hosting has moved from talk to signed contracts, and revenue recognition from HPC is already underway, beginning Q3 2025, according to management’s guidance in the Q2 earnings call. The outlook for strong HPC revenue is now highly visible. The Fluidstack agreements anchor about $3.7 billion of contracted revenue over ten year period, with two five-year options that lift the total to about $8.7 billion, under a modified gross lease (meaning TeraWulf is paid a fixed fee for infrastructure services while most variable costs are passed through to the client, keeping revenue predictable and limiting operational risk). Google will backstop $1.8 billion of those obligations and receive warrants equal to roughly 8 percent pro forma WULF ownership. The deployment will start with about 40 MW online in the first half of 2026 (just months away) and the full 200+ MW by year end 2026. Unit economics also look attractive and visible per management’s guidance. Management disclosed an expected site net operating income [NOI] margin of 85% on the Fluidstack deployment, which implies about $315 million of annual site NOI at full run rate ($3.7B contract / 10 years = $370M/yr in revenue; $370M x 0.85 NOI margin = ~$315M annual NOI). The revenue from the Fluidstack contract is highly predictable and grows over time due to built-in price escalations (starting around $370 million per year and rising over the 10-year contract). Under the modified gross lease, TeraWulf collects a fixed fee for providing power, cooling, and rack space, while most variable costs are passed through to Fluidstack, keeping revenue predictable and margins protected - an example of TeraWulf’s asset-light HPC hosting approach, which caught my attention in May. The earlier Core42 lease also validates the above approach. Terms of the Core42 deal call for 60 MW of critical IT load at $125 per kilowatt per month, equal to $1.5 million per MW per year, with a 3% annual escalator and a twelve-month prepayment. Management framed EBITDA margin at ~70% and an unlevered return of 17 to 18% on a $6 million per MW build, which is compelling in today’s data center market. Scale optionality remains, with an expansion option up to 108 MW of additional critical load. On the infrastructure side, TeraWulf’s power and sites provide strong capacity optionality for lessees. Lake Mariner has interconnection approval for 500 MW with applications pending to reach up to 750 MW, and the Cayuga site adds an 80-year ground lease with rights to develop up to 400 MW, with 138 MW expected ready online in 2026. These are zero-carbon, low-cost power markets with existing transmission and water, which shortens timelines. This infrastructure and power base is TeraWulf’s potential moat in AI infrastructure. I believe TeraWulf’s financial base is on the verge of improving, while HPC ramps. Q2 2025 revenue was $47.6 million, up 34% YoY. Bitcoin mining capacity reached 12.8 EH/s also up around 45% YoY. Though net loss widened, TeraWulf returned to positive adjusted EBITDA of $14.5 million in Q2 - a notable improvement over Q1's negative adjusted EBITDA of -$4.7 million. Management guided to recognizing HPC hosting revenue beginning in Q3 2025, which marks an inflection point for TeraWulf's financial profile. And at $3.88 billion EV and TTM revenue of $144.1 million, WULF currently trades around 27x EV/Sales. Now, with the $3.7 billion Fluidstack contract potentially adding roughly $370 million a year in revenue at full deployment, sales could potentially rise to over $500 million annually (if TeraWulf maintains around the current annual run rate from its current revenue streams). If the market continues to value WULF at the same EV/Sales multiple of 27x, the enterprise value could increase proportionally to about $13.83 billion (or more conservatively above $10 billion). Assuming net debt remains stable (which is also plausible given early cash inflows from Fluidstack, projected high NOI margins, and phased capital spending as the rollout of HPC infrastructure will be in phases), this would imply a share price around 2.5x above current price (if we stick to our more conservative ~$10 billion EV). A ~2.5x or ~250% upside from the current $8.85 share price would make WULF trade around $22. Any expansion of the multiple due to factors like higher margins, as execution of the Fluidstack and Core42 contracts scale, would potentially push the upside further. Risks and Takeaway Execution risk remains real in the HPC hosting business, so I model in some cushion (a conservative 10–15% adjustment to projected revenue and phased deployment assumptions). The critical variables here are managing the construction schedule and coordinating with clients on the on-time arrival of their hardware. The Google backstop reduces counterparty risk on the Fluidstack deal. The prepayment and escalators on Core42 improve returns while the team funds and builds. I believe WULF's risk-reward skews positive at this point of the build phase. TeraWulf's entry into HPC is based on an approach that minimizes capital and market risks. This approach makes WULF a Bitcoin miner worth being on the watch list. In a sector where timing and strategy are everything, TeraWulf may have just found its sweet spot in HPC. - Except for my WULF coverage in May. In May, I initiated coverage with a takeaway that WULF was watchlist-worthy because of its HPC business approach. My current takeaway is simple: TeraWulf now has contracts, credible partners, clear unit economics, and secured power (Lake Mariner and Haynesville sites, totaling ~40 MW to start with full 200+ MW capacity by end of 2026). If you were sidelined earlier, I believe the setup has improved. I see enough line of sight to future cash flows to upgrade to Buy .
