Case studies › WULF
TeraWulf is funding AI capacity with a share count, not a contract
TeraWulf closed the quarter with about US$2.4 billion of cash and restricted cash, a US$400 million expanded revolver with a further US$150 million accordion, and a stated intention to fund remaining Lake Mariner commitments and the planned Justified equity contributions without going back to the equity market.
On the printed numbers that is credible. The question the call kept returning to was not whether the liquidity exists but what happens to it when a tenant does not pay.
The structure, and who is really standing behind it
Google's backstop covers specified Anthropic obligations and can provide lease assumption or a debt-reducing termination payment on tenant default. Google's credit support costs are recoverable from the tenant, and the arrangement ends once Anthropic meets specified investment-grade rating requirements.
| Item | Detail |
|---|---|
| Cash and restricted cash | ~$2.4bn |
| Revolver commitments | $400m |
| Accordion | $150m |
| Lake Mariner, critical IT build | ~90% complete |
| Long-lead equipment procurement | over 80% |
| Development cost per critical IT MW | $10–12m |
| Annual contracting target | 250–500 MW |
Read carefully, the Google support is a credit enhancement with an expiry condition. It ends when the tenant reaches investment grade. That is designed correctly — the backstop is replaced by the tenant's own rating — but it means the terminal credit is Anthropic's, not Google's, and the market will reprice at the crossover.
The build is nearly done; the next one is not
Lake Mariner is roughly 90% complete with long-lead equipment over 80% procured. Development cost is held at US$10 to US$12 million per critical IT megawatt with contractual cost protections. Those are the good numbers and they are why the funding question is answerable this quarter.
Muskie is the next one and it is not at 90%. The Kentucky campus has a contracted path to 1 gigawatt of electrical service, with the first 500 megawatts scheduled for delivery from 2028 and the second phase needing additional generation infrastructure plus regulatory approval, targeted 2029.
That is a two-year gap between the current cash position and the next revenue line, bridged by a contracting target of 250 to 500 megawatts a year. The gap is where dilution lives.
The dilution channel, named
Management listed the alternatives: customer prepayments, project-level equity, and capital recycling from stabilised infrastructure. The emphasis is on preserving the spread between financing cost and lease yield, and there is a stated preference for investment-grade tenants or equivalent credit support.
The word "without equity capital markets" is doing a lot of work across a two-year build. Customer prepayment and project equity are both third-party money, and both are priced against the same spread. If the spread widens, the cheapest marginal dollar becomes the share count again.
Note also what is absent from the plan: TeraWulf intends to stay an infrastructure owner and not take GPU ownership or technology risk. That caps the upside per megawatt and removes the obsolescence risk. It is a landlord model and should be valued as one.
Where the case breaks
The first break is the credit crossover. When the Google backstop falls away at Anthropic's investment grade, the tenant is the credit. A downgrade after that point lands directly on the lease yield the whole model is built on.
The second is the 2028 gap. If Muskie's first 500 megawatts slip past 2028, the cash position carries a longer bridge with no incremental revenue, and the funding mix moves toward equity by arithmetic rather than choice.
The third is the spread. Management said it will select customers on risk-adjusted returns rather than to accelerate announcements. Every quarter where the target is missed is a quarter where the market applies pressure to do the opposite.
What the arithmetic says
US$2.4 billion of liquidity, a 90% complete build with cost protection, and a next phase that delivers in 2028 against a 2029 second tranche, funded by prepayment and project equity at an undisclosed spread. The landlord is solvent through this build. The equity is not the funding source this quarter; the count is what the next one is measured against.
A 2028 first delivery against a 2026 cash balance. The bridge is prepayment and project equity at a spread nobody has disclosed, and the share count is the cheapest marginal dollar if it widens.
Sources
Liquidity, revolver, accordion, construction progress, procurement, per-megawatt cost, contracting targets, backstop terms and delivery dates are the company's own third-quarter 2026 results and call disclosures. The two-year gap and the credit crossover point are derived from the stated build dates and backstop expiry condition and can be re-run.
Published for information only. Nothing here is investment advice or a recommendation. This note is not prepared in accordance with the legal requirements designed to promote the independence of investment research. No position is held in the security covered.