Anthropic, the developer of the Claude AI model, is in talks with infrastructure company Nscale over a new $45 billion agreement to meet its computing needs. Bloomberg reports this, citing sources familiar with the discussions. If the deal goes through, it will mark a continuation of Anthropic’s large-scale strategy to expand the capacity needed to train and run AI.

Notably, just about a month ago, the companies had already announced an agreement for a similar amount. However, the new talks have not yet been finalized, and their terms may still change.

Anthropic continues to expand its computing capacity

Over the past year, Anthropic has signed a whole series of major agreements aimed at meeting growing demand for computing. Developing and operating modern AI models requires ever more accelerators, data centers, and electricity, which is why access to infrastructure is becoming one of the key competitive factors among AI companies.

The partnership with Nscale is expected to help Anthropic secure additional computing resources. At the same time, details of the potential new agreement have not yet been disclosed. TechCrunch reached out to both companies for comment but had not received a response as of publication.

Such large commitments reflect Anthropic’s confidence in continued growth in demand for its models and in the idea that future revenue will justify today’s infrastructure investments.

Investors worry that AI spending is rising too quickly

Massive deals worth tens of billions of dollars are causing concern among some investors. They doubt that revenue from AI services will grow fast enough to offset the cost of computing capacity. High spending on chips, the construction of data centers, and their operation could become a serious financial risk if demand or commercial returns fail to meet expectations.

Despite these concerns, Anthropic continues to enter into infrastructure agreements. The company is betting that investments in computing will enable it to develop more powerful models and expand its business. However, in the long term, it will need to prove that its rising expenses can generate corresponding profits.

Both companies are preparing to go public

Anthropic and Nscale are also considering initial public offerings (IPOs). According to CNN, Anthropic still expects to go public before the end of 2026.

Nscale also has plans to become a public company, although complications have arisen over its valuation. As Bloomberg reported on October 18, some investors have expressed doubts about Nscale’s proposed $30 billion valuation.

Against this backdrop, the potential $45 billion deal is especially notable: its value exceeds the discussed valuation of Nscale itself. At the same time, the contract value is not equivalent to the infrastructure provider’s revenue or profit — the agreement may cover computing services delivered over a long period.

If an agreement is reached, it will be yet another sign of how expensive the race to advance AI has become. But for now, the talks are ongoing, and the final terms — as well as whether a new deal will be signed at all — remain unconfirmed.