Nvidia backs SiFive; cloud shift
SiFive raised $400 million at a $3.65 billion valuation with backing from Nvidia to push RISC‑V CPU designs toward AI datacentres, signaling interest in open CPU architectures. Separately, reports say Nvidia plans to exit operating its own cloud and focus on supplying chips and software platforms to partners. (parameter.io) (sg.finance.yahoo.com)
Nvidia is backing SiFive’s push into artificial intelligence data centers while stepping back from running its own cloud service. (sifive.com) (finance.yahoo.com) SiFive said on April 9 that it raised $400 million in an oversubscribed Series G round that valued the chip-design company at $3.65 billion. Atreides Management led the round, and Nvidia, Apollo Global Management, Point72 Turion, T. Rowe Price Investment Management, Prosperity7 Ventures, and Sutter Hill Ventures also participated. (sifive.com) (datacenterdynamics.com) SiFive said it will use the money to speed up its high-performance data-center roadmap and expand engineering teams for “agentic artificial intelligence workloads.” Reuters reported the company is targeting central processing unit designs for data-center systems used alongside artificial intelligence accelerators. (sifive.com) (msn.com) A central processing unit is the general-purpose manager inside a server, while Nvidia’s graphics processing units handle the heavy math for artificial intelligence. SiFive sells processor blueprints, not finished chips, so customers can license its designs and build their own silicon around them. (sifive.com) (riscv.org) Those blueprints use RISC-V, pronounced “risk-five,” an open instruction set architecture that works like a shared language for chip designers. RISC-V International says the standard is open, and SiFive says that openness lets customers customize processors without depending on a single proprietary architecture. (riscv.org) (sifive.com) Nvidia’s check matters because artificial intelligence servers need more than graphics chips; they also need central processors to move data, schedule work, and manage memory. TechCrunch reported that Nvidia’s investment points to growing interest in open central processing unit designs as companies build more specialized artificial intelligence infrastructure. (techcrunch.com) (riscv.org) At the same time, Nvidia is leaning harder into selling the platform around those systems instead of operating a cloud business itself. Nvidia’s cloud-partner page says its network of cloud providers delivers hosted computing on Nvidia infrastructure, and its DGX Cloud Lepton service is described as a way to connect developers to that partner network. (nvidia.com 1) (nvidia.com 2) Yahoo Finance reported on April 13 that Nvidia plans to exit operating its own cloud business and focus instead on supplying chips and software platforms to partners. Earlier reporting from The Information, cited by Data Center Dynamics and Investing.com in September 2025, said Nvidia had already been shifting DGX Cloud toward internal use rather than direct competition with Amazon Web Services and Google Cloud. (finance.yahoo.com) (datacenterdynamics.com) (investing.com) That leaves Nvidia in a familiar position: sell the picks, the software, and the reference designs while partners run the rented machines. SiFive’s new funding and Nvidia’s cloud retreat both point to the same target in April 2026 — owning more of the artificial intelligence stack without owning every server. (nvidia.com) (sifive.com)
