Anthropic Signs 20-Year Data Centre Lease, Faces $75M Book Lawsuit

Three distinct developments landed around Anthropic on Sunday and Monday: a major long-term infrastructure commitment, a fresh copyright lawsuit, and a revenue benchmark that puts the company ahead of nearly every public software firm.

Key points

• Anthropic signed a 20-year lease on a TeraWulf data centre, locking in dedicated compute capacity for the long run.
• A new $75 million lawsuit alleges Anthropic pirated books to train Claude, adding to the company’s existing legal exposure on training data.
• SaaStr analysis argues that Anthropic will out-earn every public software company except Microsoft by year-end, based on its 2025 exit run rate.
• The TeraWulf deal (NASDAQ: WULF) marks a notable shift toward owned or long-leased infrastructure rather than purely on-demand cloud capacity.

What does a 20-year data centre lease signal?

A two-decade commitment to a single data centre operator is not routine procurement. It signals that Anthropic is betting on predictable, large-scale compute demand far into the future, and that it wants cost and capacity certainty rather than the flexibility (and premium pricing) of hyperscaler cloud contracts.

TeraWulf is known for operating power-focused data centres, and the deal gives Anthropic a dedicated facility it can build around its specific hardware and networking requirements. For investors watching Anthropic’s path to profitability, long-term lease obligations are a fixed cost that weighs on margins but also reduces unit-cost volatility as the company scales. The full financial terms of the lease have not been disclosed in available sources.

This move fits a broader pattern. Anthropic has been building out infrastructure relationships aggressively, including the Samsung chip discussions reported earlier this week. The company appears to be pursuing a strategy of securing supply across the stack rather than depending on any single vendor or cloud provider.

How close is Anthropic to the top of the software revenue table?

The SaaStr analysis draws on Anthropic’s reported 2025 exit run rate to make a striking comparison: by the end of 2026, the company’s revenue trajectory would place it ahead of every major public software business except Microsoft. The piece notes that observers can lose sight of how large Anthropic already is relative to established software companies, given how much attention goes to its growth rate and valuation.

Precise revenue figures are not independently verified here, as Anthropic remains private and does not report publicly. The SaaStr figures appear to be based on previously reported run-rate data. Still, if the trajectory holds, the comparison reframes Anthropic less as a startup chasing scale and more as a business that has already reached the upper tier of the software industry by revenue, while still carrying the cost structure and risk profile of a company in build-out mode.

That tension matters for anyone thinking about valuation. High revenue at this scale is genuinely rare. But Anthropic is simultaneously signing 20-year infrastructure leases, funding frontier model research, and managing a growing litigation docket. The gap between revenue and free cash flow is not publicly quantified.

What is the new copyright lawsuit about?

Authors have filed a $75 million lawsuit claiming Anthropic used pirated books to train Claude without authorization or compensation. The suit adds to a pattern of copyright litigation targeting AI developers over training data practices.

Anthropic is not alone in facing these claims. Similar suits have been filed against other major AI labs. The legal theory, broadly, is that ingesting copyrighted text to train a commercial model constitutes infringement, and that authors are owed damages. Courts have not yet settled the core questions, and outcomes in this area remain genuinely uncertain.

For Anthropic specifically, the litigation risk is worth tracking because training data lawsuits, if they produce large judgments or force licensing agreements at scale, could alter the economics of model development across the industry. A $75 million claim is material but not existential at Anthropic’s current revenue scale. The longer-term question is how many similar suits accumulate and whether any reach trial or force settlement terms that set industry precedent.

Putting it together

Sunday and Monday delivered a concrete infrastructure move, a revenue milestone framing, and a new legal headache. None of these is entirely surprising given Anthropic’s trajectory, but together they illustrate the operating reality of a company trying to scale frontier AI: capital commitments are growing, revenues are rising fast, and legal risk is accumulating in parallel. This update is published by an independent site and contains no investment advice.