Anthropic and Nscale reportedly discuss another $45 billion compute deal
Anthropic and Nscale are reportedly discussing another $45 billion AI computing deal, but the talks are not yet a confirmed agreement.
Anthropic and AI infrastructure provider Nscale may be preparing another huge computing agreement, according to reporting cited by TechCrunch. The reported discussions would involve a further $45 billion deal, although the talks are still under way and the details could change before anything is announced.1
That uncertainty matters. This is not a signed contract or a confirmed expansion: TechCrunch says Bloomberg reported the talks, citing people familiar with the matter, while also noting that it contacted Anthropic and Nscale for comment. Neither company had confirmed a new agreement in the source material available for this report.1
Why the reported deal matters
Anthropic has been securing enormous amounts of computing capacity as it expands its Claude AI business. A second agreement on the same headline scale as the $45 billion arrangement reported in August would underline how quickly the cost of training and serving frontier AI models is moving from a technical concern into a major infrastructure market.1
The earlier arrangement was reported as a plan for Anthropic to rent around 460 megawatts of capacity at an Nscale data-centre development in West Virginia. That context helps explain why a fresh deal would be significant: it would not simply add a few racks of servers, but potentially commit the companies to another large, long-term build-out of power, facilities and specialised hardware.2
Nscale and Anthropic are also both pursuing public offerings, according to the reporting. TechCrunch says Anthropic is expected to go public this year, while Bloomberg has reported investor pushback around Nscale's proposed valuation. A new infrastructure commitment would therefore be watched not only as an AI-industry story, but also as a test of how public markets view the spending needed to support rapidly growing model companies.1
The wider question is whether demand for AI compute will keep justifying deals measured in tens of billions of dollars. The companies involved are betting that more capacity will translate into more capable products and more paying customers. Investors, meanwhile, are increasingly asking how quickly that spending can become durable revenue rather than an arms race between labs and cloud providers.
Our opinion
The most important fact here is the uncertainty, not the eye-catching dollar figure. If the talks produce a signed agreement, it will be a meaningful development in the race to build AI infrastructure. Until then, readers should treat the $45 billion figure as a reported possibility rather than money already committed. The story is worth watching because even negotiations at this scale show how much compute the industry believes it may need — and how much financial risk companies are willing to take to secure it.
We will update this story if Anthropic or Nscale confirms the discussions or provides material detail about the proposed capacity, timetable or commercial terms.