TeraWulf Bought Certainty This Week. The Market Sold the Sector.

A $19B Anthropic lease with a Google backstop, a Fluor GMP, and a 46% drawdown in 90 days — TeraWulf is now a spread business being valued as a momentum trade.

WULF · Earnings Call · 2026-11-06

The trade got better while the stock got worse

On Wednesday, TeraWulf announced two things that would normally be worth a re-rating: Google's credit support behind the financing of its roughly $19 billion Anthropic lease at Justified, and a definitive EPC agreement with Fluor covering all three data center buildings. Two days later the stock sits 46% lower over 90 trading days, 51% below its July 9 close. The company's own tape is a single unbroken leg down. That divergence is the story. The global tape is doing the same thing to the whole theme: AI data center names fell about 29% over 90 days with 58 decliners against 7 advancers, and frontier AI lab exposure dropped 44% in the last 30 days alone. Investors are no longer paying for contracted megawatts — they are repricing the cost of capital used to build them. TeraWulf's management spent the entire call arguing that this is precisely the environment their model was designed for.

We have always believed in raising capital when you can. Not when you have to.
Paul Prager, Chairman and Chief Executive Officer · 2026-11-06

Version 2.0 of the backstop — read the fine print

CFO Patrick Fleury drew a distinction that did not exist a year ago. At Wolfe Compute, Google's support came alongside warrants; at Justified, he said, TeraWulf is granting no incremental upfront equity, the cost of the support is borne by the tenant, and the backstop falls away once Anthropic reaches investment grade. That is a materially different contract than the one the company signed in 2025, and it explains why investment grade now sits atop the company's keyword ranking — it is no longer a preference, it is the operating constraint. “they are not public, and they are not investment grade yet. And as you know, that has been our number 1 focus” — Patrick Fleury, Chief Financial Officer · 2026-11-06 Nazar Khan framed it bluntly: “TeraWulf is not a Neo Cloud. We are an energy infrastructure company.” — Nazar Khan, Chief Technology Officer · 2026-11-06 The distinction matters because the market has started treating the two identically. It also explains the answer to Mike Grondahl's question about Riot's capital-light powered-land structure — Paul Prager called it "interesting. But it is not currently our primary focus," and Khan added that TeraWulf intends to earn a return on the full $10-12 million per critical IT megawatt rather than the first $2 million.

The numbers finally arrive

Q3 revenue was $76.2 million against $44.8 million in Q2, with HPC lease revenue up 97% sequentially to $63 million — 83% of the total. Only 102 critical megawatts were earning rent during the quarter. Exiting Q4 adds 168 megawatts at CB-4; exiting Q1 adds another 168 at CB-5; by then management expects roughly $700 million of annualized lease revenue. Wolfe Compute is tracking at approximately $8.9 million per megawatt against original guidance of $8-10 million. The reported margin of 17% looks alarming next to an 85% target, but it is contaminated by tenant fit-out revenue and pre-revenue development spend — the adjusted HPC leasing segment margin was about 83%. The fundamentals tape, which lags to the July quarter, still shows the pre-inflection shape: total revenue of $45 million with a gross margin of 19.9%, against a price-to-revenue multiple of 74.6x. That valuation is the reason dilution is a live debate on every call.

What TeraWulf stopped talking about

The most informative part of this call is what barely came up. Chesapeake/Morgantown — the Maryland gigawatt campus with 500 megawatts of battery storage that was a headline asset six months ago — drew exactly one question, and Prager's answer was notably conditional: “there are conditions precedent to close there. That have yet to be fulfilled by the seller, and those conditions are important to us.” — Paul Prager, Chairman and Chief Executive Officer · 2026-11-06 The keyword record agrees: Maryland spiked as a gainer two quarters back and has since gone quiet. The Abernathy joint venture has moved in the same direction. Rather than an operating asset, it is now described as a financing event — roughly $150 million of proceeds expected on or before December 31, 2026, and another ~$130 million by April 30, 2027, with a fair-value gain running through net income. "Unlocking equity from stabilized assets and recycling that capital back to the parent will be a priority in 2027." The JV that was announced with considerable fanfare is being monetized, not built out. With Muskie, the opposite is happening. Khan laid out the path to a full gigawatt: a 500 MW ESA under Kentucky Power's existing tariff delivering in 2028, plus a filing with the Kentucky PSC within weeks for the Big Sandy 760 MW CCGT that supports the second 500 MW, with an embedded $100 million residential credit spread over ten years. Approval is expected mid-2027, power by end-2029. That is the deliverable power argument in its most concrete form — and the reason Prager could say the site "is far more valuable today than it was at the beginning of the week."

The spread is the business

Strip away the campus names and this is a carry trade between yield on cost and weighted average cost of capital. Fleury has been prosecuting it for three quarters and has not softened it. “if your WACC is not below your yield on cost, by definition, there is no equity value” — Patrick Fleury, Chief Financial Officer (CFO) · 2026-08-05 Which is why the equity-overhang question, asked by Stephen Byrd, got the most emphatic answer of the call: “we will not sign 1 of the biggest leases that the company has ever signed and be forced to dilute our shareholders at these levels. We just will not do it.” — Patrick Fleury, Chief Financial Officer · 2026-11-06 The tool kit cited — customer prepayments, project-level equity from infrastructure funds at higher multiples, and the fact that Wolfe Compute is under-levered versus peers — is plausible but unproven. The balance sheet shows why it needs to be: liabilities at 98.2% of assets and effective net cash of negative $110 million as of the July quarter, before the Q3 capital-markets work added roughly $2.4 billion of cash and restricted cash group-wide. So the tension resolves like this: TeraWulf has spent two years converting a Bitcoin miner into a contract-financed landlord, and this week it bought three of the four things that de-risk that model — credit support, a fixed-price builder, and liquidity. The fourth, cheap project debt, is still pending. Meanwhile the market has decided that the entire category's cost of capital is going up, and it is pricing TeraWulf as a member of that category rather than as the vendor of the spread. On the tape, the cost of capital is winning. On the contracts, the project level math is winning. Until one of those two facts changes, the stock is a bet on the second — and the first half of 2027, when CB-5 rent commences and the Justified financing prints, is when we find out which one was right.