Anthropic's $19 Billion TeraWulf Lease Turns AI Power Into an Investor Story

TeraWulf said Monday that Anthropic signed a 20-year lease for a purpose-built AI infrastructure campus at its Justified Data site in Hawesville, Kentucky, a transaction the company expects to generate about $19 billion of contracted revenue over the initial term.

The deal matters beyond one stock because it shows how the artificial-intelligence buildout is moving deeper into physical infrastructure. For investors, the AI trade is no longer only about chips, model companies or cloud software. It is also about who controls power, land, cooling, interconnection and enough capital to turn those inputs into usable computing capacity.

TeraWulf's announcement also came with a capital reset. The company said it will sell its 50.1% stake in the Abernathy Joint Venture in Texas to a Fluidstack-led investor group, monetizing roughly $450 million of invested capital and freeing money for wholly owned AI infrastructure opportunities.

What TeraWulf and Anthropic announced

TeraWulf said the Anthropic lease covers a purpose-built AI infrastructure campus at the Justified Data site in Hawesville. The company said the campus will be developed in phases, with initial capacity expected to come online during the second half of 2027 and a full ramp to about 401 megawatts by early 2028.

The revenue figure is large but forward-looking. TeraWulf described the lease as expected to generate approximately $19 billion of contracted lease revenue over the initial 20-year term and said it is expected to be supported by investment-grade credit. Investors should read that as a long-term contracted-revenue claim tied to future delivery, not as revenue that has already been earned.

The company also said the Abernathy sale will monetize its roughly $450 million investment at a premium to invested capital. The Abernathy project was established in 2025 to develop a 168-megawatt critical IT load AI data-center campus in Texas with Fluidstack.

Why investors care about power, not just chips

The second-layer story is that AI capacity is becoming an infrastructure-finance market. Model developers need chips, but chips need electricity, cooling, real estate, transformers, grid access and operators willing to sign construction and lease commitments years before a campus is fully online.

That changes the investor question. Instead of asking only which semiconductor supplier wins the next order cycle, investors now have to ask which data-center developers can secure large customers, fund long construction schedules, manage power constraints and deliver capacity on time. In that context, TeraWulf's former bitcoin-mining footprint becomes relevant because miners already built businesses around power-intensive computing sites.

Reuters described the transaction as helping TeraWulf shift away from relying on bitcoin mining, while MarketWatch said the deal validated the company's pivot toward supporting the AI buildout. The point is not that every crypto-mining asset will become an AI campus. It is that power-connected sites have become more valuable when AI tenants are willing to sign long-term commitments.

Who is affected

TeraWulf shareholders are affected first because the deal gives the company a clearer AI infrastructure narrative and a large contracted-revenue target. MarketWatch reported that TeraWulf shares ended Monday up 4.9% after rising more sharply earlier in the day, and that the stock has gained more than 80% in 2026.

Other AI infrastructure and former crypto-mining names are also affected because the transaction gives investors a fresh comparable deal. MarketWatch reported gains in several peer stocks, including CoreWeave, Iren and Hut 8, as the market reassessed the demand for dedicated AI capacity.

Anthropic is affected because the lease adds a dedicated future capacity path for its AI workloads. Business Insider described the project as a new Kentucky AI data center intended to power Anthropic under a 20-year agreement. For AI labs, access to compute has become a strategic constraint as much as a technology expense.

Local utilities, grid planners and communities are affected too, although the financial disclosures do not answer every local question. A 401-megawatt critical IT load campus is a major power user, and the broader AI infrastructure buildout is likely to keep pressure on power procurement, permitting, cooling systems and regional transmission capacity.

The caveat: delivery risk is still real

The main limitation is timing. TeraWulf said initial capacity is expected in the second half of 2027 and full capacity by early 2028, meaning investors are valuing a project that still depends on development execution. Interconnection, equipment procurement, construction, cooling, financing and customer requirements can all affect the eventual economics.

There is also concentration risk. A long lease with a major AI customer can improve revenue visibility, but it also ties a large project to one tenant's compute needs and credit profile. TeraWulf said the lease is expected to be supported by investment-grade credit, but the market will still want more detail on financing, margins, service-level obligations and any expansion terms.

Finally, investors should separate the headline contract value from profit. A $19 billion revenue expectation does not tell readers the cost of building and operating the campus, the financing structure, the margin profile or the return on invested capital. Those details will determine whether the lease is merely impressive in size or genuinely accretive over time.

What To Watch Next

Watch TeraWulf's next investor updates for project financing, capital spending, customer-credit support and margin guidance. Those details will show how much of the headline revenue can translate into durable cash flow.

Watch permitting, interconnection and construction milestones at Hawesville. The first major test is whether the company can keep the second-half 2027 initial-service target on track and provide evidence that the full 401-megawatt ramp remains realistic for early 2028.

Watch comparable AI infrastructure leases across data-center developers, utilities and former crypto miners. If more AI labs sign long-duration capacity deals, investors may start valuing power-ready sites as a scarcer asset class within the AI supply chain.