Back to News
Market Impact: 0.42

Nvidia-backed SiFive hits $3.65 billion valuation for open AI chips

NVDAINTCARMMETANETTROWAMD
Artificial IntelligenceTechnology & InnovationPrivate Markets & VentureCompany FundamentalsProduct Launches

SiFive raised $400 million in an oversubscribed round at a $3.65 billion valuation, with Nvidia participating alongside Atreides Management, Apollo, D1 Capital, Point72 Turion, and T. Rowe Price. The funding supports SiFive’s expansion from embedded uses into AI data center CPUs, with its open RISC-V designs intended to work with Nvidia’s CUDA and NVLink Fusion ecosystem. The deal underscores growing investor interest in alternative CPU architectures for AI infrastructure.

Analysis

This is less about one startup and more about Nvidia expanding its control point from accelerators into the CPU layer that sits adjacent to them. By backing an open, non-ARM/non-x86 architecture, Nvidia is effectively subsidizing a “good enough” CPU ecosystem that reduces customers’ dependence on legacy CPU vendors while keeping the software and interconnect gravity centered on Nvidia. The second-order effect is that the value pool shifts from ISA ownership to platform integration: whoever owns the rack-level scheduling, networking, and developer tooling captures more economics than the CPU designer itself. The biggest near-term beneficiary is NVDA, because any credible alternative CPU that is CUDA- and NVLink-compatible lowers friction for AI factory adoption and widens Nvidia’s addressable system footprint. That is incrementally negative for INTC and AMD, not because SiFive will displace them immediately, but because it normalizes the idea that CPU choice is modular and interchangeable inside Nvidia-centric clusters. Over 6-18 months, that can compress bargaining power for server CPU incumbents in AI racks and push more gross profit toward the interconnect/orchestration layer rather than the socket. ARM is a more subtle casualty and competitor at the same time: the market is signaling appetite for a neutral licensing model, but the strategic risk is that neutrality plus open architecture can be cheaper and less politically encumbered than ARM’s toll-road model. META and NET benefit if they can diversify infrastructure sourcing and avoid single-vendor lock-in, while TROW’s participation is a reminder that public-market capital is still chasing private AI infrastructure optionality despite late-cycle pricing. The contrarian read is that this is not an immediate RISC-V victory; it is Nvidia preemptively shaping a future standard before Intel/AMD can use CPU breadth to reopen platform negotiations. The key risk is execution. RISC-V still has to prove software maturity, validation, and reliability at hyperscale in a 12-24 month window; if integration slips, this remains a financing event rather than a product inflection. The other tail risk is that ARM or x86 incumbents respond with aggressive pricing and co-designed rack offerings, blunting SiFive’s relevance before it becomes strategically important.