There’s a strange thing happening in the fight to dethrone Nvidia. The company that already owns the AI hardware game just invested in an 11-year-old startup building CPUs on completely different technology.
SiFive, founded by UC Berkeley engineers who created the open-source RISC-V chip design, just closed a $400 million funding round that values it at $3.65 billion. Nvidia was in the round. So were a bunch of serious money players: Atreides Management (led by former Fidelity bigwig Gavin Baker), Apollo Global Management, D1 Capital Partners, and others. This is the company’s first major funding since March 2022.
On the surface, this looks like Nvidia hedging. But it’s actually way more calculated than that.
The Architecture That Doesn’t Play By Intel or ARM’s Rules
Here’s the thing about RISC-V: it’s not x86 (Intel’s territory) and it’s not ARM (basically everything else). It’s open source and neutral, which means any company can use it, modify it, and build on it without paying licensing fees to a central authority or worrying about proprietary lock-in.
SiFive operates like ARM used to. It doesn’t manufacture chips. It licenses designs that customers can customize for their own needs. The company stayed quiet for four years after that last funding round, which suggests they were actually building something worth backing.
Until recently, RISC-V was relegated to smaller applications: embedded systems, IoT devices, the kind of hardware that doesn’t grab headlines. But technology moves fast when billions of dollars are on the table.
Why Nvidia Cares About An Open-Source Chip
Nvidia’s play here isn’t about replacing GPUs. It’s about control and ecosystem building. SiFive’s new designs will work with Nvidia’s CUDA software and its NVLink Fusion, a rack server system that lets different CPUs plug directly into what Nvidia calls its “AI factory.”
In other words, Nvidia is creating a world where different CPU architectures can plug into its infrastructure. That’s brilliant defensive strategy. Intel and AMD are still trying to build their own AI accelerators to compete with Nvidia’s GPUs. Meanwhile, Nvidia is building a platform flexible enough to absorb whatever CPU architecture comes next.
The irony is sharp: as Nvidia’s traditional rivals scramble to compete, Nvidia is backing an alternative technology stack entirely. It’s like owning the stadium while others are arguing over which team should play there.
The Business Model Matters More Than The Chip
What makes SiFive genuinely interesting isn’t the architecture itself. It’s that the company exists in an unusual position. Open designs plus neutral backing plus proper funding means SiFive could theoretically serve customers that would never trust a single vendor like Intel or ARM.
That has business implications beyond just chip design. In a world increasingly paranoid about supply chain vulnerabilities, open-source alternatives to proprietary monopolies suddenly look less like hobbyist projects and more like strategic assets.
The funding round attracted investors who clearly think long-term about infrastructure. Atreides also backed Cerebras Systems’ $1 billion round. These aren’t people gambling on quick exits. They’re betting on foundational shifts in how computing gets built.
What This Says About The AI Hardware Race
The fact that this round happened at all signals something important: the fight over AI hardware isn’t just about who builds the fastest GPUs anymore. It’s about architecture, ecosystem lock-in, and who controls the standards that future systems will run on.
Nvidia winning isn’t surprising. Nvidia investing in its potential competition and making sure that competition runs on compatible infrastructure? That’s the move of a company that knows the real game is about platform dominance, not just performance.
Whether SiFive actually challenges anything meaningful is still an open question. But Nvidia clearly believes the company’s going somewhere worth watching, which might matter more than any technical benchmark at this point.
