AI infrastructure goes geographic

AI computing is spreading into new places and new providers, as Nvidia expands in India, SiFive raises new chip money, and CoreWeave adds Anthropic. (nvidia.com) On April 10, CoreWeave said Anthropic signed a multi-year deal to run Claude workloads on its cloud, with capacity coming online later in 2026. CoreWeave shares rose more than 13% that day, and the company did not disclose the contract’s value. (usnews.com) The Anthropic deal followed CoreWeave’s April 9 announcement that Meta had expanded its own commitment by $21 billion through December 2032, adding to a prior $14.2 billion arrangement. CoreWeave told CNBC that nine of the 10 leading foundation-model providers now use its platform. (cnbc.com) A cloud provider rents out computing power, and an artificial intelligence cloud packs those rentals with graphics processing units, the chips used to train and run models. CoreWeave’s pitch is that specialized clouds can add this capacity faster than the biggest general-purpose cloud companies can build it for themselves. (coreweave.com) (cnbc.com) On April 9, SiFive said it raised $400 million in an oversubscribed Series G round that valued the company at $3.65 billion. The round was led by Atreides Management, with Nvidia, Apollo Global Management, Point72 Turion, T. Rowe Price Investment Management, Prosperity7 Ventures, and Sutter Hill Ventures also participating. (sifive.com) SiFive sells processor designs rather than finished chips, and its designs use RISC-V, an open instruction set that works like a shared rulebook for how a processor understands software. Chief executive Patrick Little told Reuters the company wants to move from embedded devices into data-center central processors for artificial intelligence systems. (wsau.com) India is pushing the same build-out from the infrastructure side. Nvidia said on February 17 that IndiaAI Mission is backed by more than $1 billion in public funding and that it is working with Yotta, Larsen & Toubro, and E2E Networks on domestic computing clusters. (nvidia.com) Nvidia said Yotta is building Shakti Cloud with more than 20,000 Blackwell Ultra graphics processing units in Navi Mumbai and Greater Noida. Larsen & Toubro said on February 18 that it plans a gigawatt-scale artificial intelligence factory, including a Chennai expansion to 30 megawatts and a new 40-megawatt Mumbai facility. (nvidia.com) (larsentoubro.com) Those projects give governments, startups, and model companies more ways to buy compute without relying only on Amazon Web Services, Microsoft Azure, or Google Cloud. They also give Nvidia more routes to sell chips, networking gear, and software into regional data-center operators instead of only into United States hyperscalers. (nvidia.com) (cnbc.com) The immediate test is delivery. CoreWeave has to bring Anthropic capacity online later this year, SiFive has to turn new funding into data-center products, and India’s partners have to convert announced megawatts and graphics processing units into working clusters. (usnews.com) (sifive.com) (larsentoubro.com)