Anthropic has signed a $35 billion cloud-computing deal with Nvidia-backed Lambda for capacity at a Texas data center, according to Reuters, citing a person familiar with the matter.
The project is being developed by Hut 8 in Nueces County and is expected to provide about 350 megawatts of capacity, the source said. The capacity will use Nvidia systems and support growing demand for Anthropic’s Claude AI. The source spoke anonymously because the information was not public.
The Wall Street Journal, which first reported the agreement, said Nvidia itself would hold the lease on the data center. Reuters said Anthropic, Lambda, Nvidia and Hut 8 did not immediately respond to requests for comment outside regular business hours. The reports did not provide a timetable for bringing the Anthropic capacity online.
The Lambda agreement is separate from Anthropic’s recently announced deal with Nscale. Last week, Anthropic said it would spend $45 billion to rent AI cloud capacity from Nscale’s West Virginia data center campus. Taken together, the reported Lambda contract and the Nscale commitment amount to $80 billion in planned cloud spending, but they involve different providers, locations and capacity arrangements.
The latest deal also places Anthropic in a broader race among AI developers to secure electricity, data-center space and advanced chips before demand arrives. For Anthropic, the immediate target is more capacity for Claude, including its Claude Code product, as usage grows. Lambda specializes in Nvidia-based cloud infrastructure and is itself backed by the chipmaker, while Hut 8 has shifted from cryptocurrency mining toward large-scale AI data-center development.
The 350-megawatt figure describes the capacity covered by the Texas project, not Anthropic’s total computing footprint. The value and project details come from an unnamed source rather than a company announcement, and the companies had not publicly confirmed the terms when Reuters published its report. That makes the agreement a reported infrastructure commitment alongside, rather than a replacement for, the Nscale expansion.