NVIDIA-Backed SiFive Raises $400M at $3.65B Valuation to Bring Open RISC-V Chips to AI Data Centers

SiFive, the UC Berkeley-founded chip IP licensing company that created the RISC-V open-source architecture, has raised a $400 million Series G round at a $3.65 billion valuation. The round was led by Atreides Management and included NVIDIA, Apollo Global Management, Point72 Turion, T. Rowe Price, and D1 Capital Partners — among others. The raise is SiFive's first since a $175 million round in March 2022, when the company was valued at $2.33 billion pre-money, and comes as agentic AI workloads are dramatically increasing demand for high-performance, customizable CPU architectures in data centers.

Why RISC-V Is Suddenly Exciting for AI

RISC-V is an open-standard chip instruction set architecture (ISA) — an alternative to Intel's proprietary x86 and ARM's proprietary architecture. Unlike those closed designs, RISC-V allows any company to freely use and customize the specification. SiFive's business model mirrors what ARM has done for decades: it designs high-performance RISC-V CPU cores and licenses them to chip makers who adapt them for specific applications — but without the proprietary lock-in.

Agentic AI is the catalyst for the company's current momentum. As AI workloads shift from simple inference to complex orchestration tasks — where AI agents coordinate across applications, files, and APIs — the demand for efficient, customizable CPU performance in data centers has surged alongside GPU demand. SiFive's RISC-V designs integrate scalar, vector, and matrix compute in one standards-based interface, giving hyperscalers the architectural freedom to build differentiated data center silicon without licensing a proprietary ISA. This reflects the broader push for AI infrastructure independence, similar to the European push for open, sovereign AI stacks.

The NVIDIA Investment Is the Strategic Tell

The most significant detail in the raise is NVIDIA's participation. SiFive's RISC-V CPU designs are compatible with CUDA and will work with NVIDIA NVLink Fusion — a rack server architecture that allows different CPU designs to plug directly into NVIDIA's AI infrastructure. Rather than competing with NVIDIA's GPUs, SiFive is positioning RISC-V CPUs as the open-standard orchestration layer that feeds into NVIDIA's compute ecosystem. NVIDIA backing an open CPU architecture that integrates with its own systems is a direct challenge to Intel and ARM's dominance in the data center CPU market — and a hedge against hyperscaler customers seeking to reduce proprietary architecture dependency. The funding will be used to expand RISC-V CPU IP R&D, build out the software ecosystem including CUDA, Red Hat, and Ubuntu ports, and deepen work with data center customers deploying RISC-V silicon through NVLink Fusion.

Frequently Asked Questions

What is RISC-V and how does it differ from ARM and x86?

RISC-V is an open-standard chip instruction set architecture that any company can freely use and customize without paying proprietary licensing fees. ARM and Intel x86 are closed, proprietary architectures that require licensing agreements and limit architectural customization. RISC-V gives chip designers full freedom to differentiate their silicon designs.

Why did NVIDIA invest in SiFive?

SiFive's RISC-V CPU designs are compatible with CUDA and NVIDIA NVLink Fusion, meaning they can function as the CPU layer in NVIDIA-powered AI data center racks. NVIDIA investing in an open CPU architecture that integrates with its ecosystem positions it to benefit from the shift away from Intel x86 and ARM in AI data centers — without building CPUs itself.

What will SiFive do with the $400M raise?

SiFive plans to expand RISC-V CPU IP research and development, accelerate its data center software ecosystem including CUDA, Red Hat, and Ubuntu support, and work closely with hyperscale data center customers deploying RISC-V silicon via NVIDIA NVLink Fusion.

The Bottom Line

SiFive's $3.65 billion raise is a signal that the CPU architecture wars are entering the data center — and that RISC-V is no longer a niche embedded-systems story. NVIDIA's participation makes the strategic logic explicit: the dominant AI compute company wants an open, customizable CPU architecture plugging into its systems rather than ceding that layer to Intel or ARM. For hyperscalers who have been pushing for open chip alternatives, SiFive just became a significantly better-capitalized option. The next test is whether RISC-V CPUs can deliver data center performance at scale — something the $400 million is specifically designed to prove.