Nscale and Anthropic are reportedly in talks for yet another $45B compute deal

Anthropic and AI infrastructure provider Nscale are reportedly nearing a second $45 billion compute agreement, roughly a month after the pair announced a first deal of the same size. Bloomberg reported the talks, citing people familiar with the matter, who cautioned that negotiations are continuing and that the details could still change. TechCrunch said it had contacted both companies for comment.
The reported figure mirrors the earlier arrangement exactly: a little over a month ago, the two companies publicly announced a $45 billion compute deal. If a second agreement of that scale is confirmed, it would mean two consecutive contracts of the same headline value between the same AI lab and the same infrastructure provider within a short window. The sources characterizing the talks did not attach a timeline, site, capacity figure or delivery schedule to the reported deal, and none of those details have been disclosed.
For developers, the practical relevance of a compute contract lies less in the dollar figure than in what it secures: training capacity for future models and inference capacity to serve existing ones. The dollar amount alone says nothing about how many accelerators are involved, which generation of hardware, where the data centers sit, or when capacity comes online. A second large commitment from a model developer to a single infrastructure provider could indicate sustained demand growth rather than a one-off expansion, but that is an inference from the size and repetition of the reported numbers, not something the companies have stated.
The backdrop is a broader pattern. Anthropic has signed a string of large compute agreements over the past year, and this would be the latest in that sequence. That spending has unsettled a portion of investors, who worry that companies are committing too much capital for too little return. Anthropic has continued signing such partnerships anyway, reportedly driven by the belief that the investment will pay off. Readers weighing the news should note that the concern described is about the economics of the whole sector, not a specific criticism of either company's finances.
Corporate context adds another layer. Both Anthropic and Nscale are pursuing IPOs, with Anthropic still expected to go public this year. Nscale is also seeking a listing, though Bloomberg reported on Monday that the company has encountered investor pushback over its proposed $30 billion valuation. This suggests the timing of the reported deal may not be neutral: a large, named customer commitment is the kind of evidence a private infrastructure company can point to when arguing for its valuation, and a model developer can point to when arguing it has secured supply. That is my reading of the incentives, not a stated motive of either party. A likely trade-off is that announcing headline dollar figures invites scrutiny of exactly how those sums translate into capacity and revenue.
For product teams, the near-term effect is more likely indirect than immediate. GPU supply secured through long-term contracts shapes which model tiers a developer can rely on, how quickly new capabilities reach an API, and whether rate limits loosen or tighten during demand spikes. None of that is specified here. What is clear is the direction of travel: a model developer is reportedly willing to commit tens of billions of dollars to third-party infrastructure twice within a month, and both counterparties are heading toward public markets where such commitments will be examined closely.
Why it matters: If confirmed, the reported deal would show Anthropic continuing to lock in very large compute commitments even as some investors question AI spending levels, and it would give Nscale a second marquee contract ahead of its IPO. Developers and product teams should treat the figure as a signal about expected capacity growth rather than a concrete change in pricing, availability or model access, none of which has been announced. Inference: the IPO plans of both companies make it likely that more detail on capacity and terms will surface during the listing process, which is when the practical implications for API users would become clearer.
Based on reporting from the original publisher. Visit the source for full context and later updates.
Publisher excerpt
If the deal goes through, it would be the latest in a long string of giant compute deals Anthropic has struck over the past year.