SiFive, the company trying to make RISC-V the default architecture for AI data centers, just raised $400 million in a Series G round. The deal values the company at $3.65 billion. Standard startup funding story, right?

Except one of the investors is Nvidia.

Read that again. The company that dominates AI computing so thoroughly that entire nations are competing for its GPUs just put money into a startup whose entire thesis is that the future of data centers will not be built exclusively on Nvidia hardware. SiFive is designing RISC-V central processor units for the data center, the exact type of chip that could reduce dependence on Nvidia accelerators for AI inference workloads.

This is Jensen Huang playing chess while everyone else plays checkers.

The round was led by Atreides Management, with Apollo Global Management, Point72, Prosperity7 Ventures, and Sutter Hill Ventures also participating alongside Nvidia. Reuters first reported the deal on Thursday.

Here is why this matters. The AI infrastructure stack is shifting. Training models still requires massive GPU clusters, and Nvidia owns that market. But running trained models (inference) is increasingly a CPU-heavy workload. As the industry moves from building AI to deploying AI, CPUs are becoming the bottleneck. There is a worldwide shortage. Arm just built its first in-house chip in its 35-year history to address it. Google just expanded its Intel Xeon partnership for the same reason.

RISC-V is the wild card. Unlike x86 (Intel and AMD) or Arm, RISC-V is an open-source instruction set architecture. Anyone can build chips using it without paying licensing fees. That has made it incredibly attractive to companies that want to escape the Intel-Arm duopoly on CPUs, particularly in AI data centers where margins are thin and volume is everything.

SiFive is the biggest pure-play RISC-V company in the world. With this $400 million, it plans to build production-grade data center CPUs and expand its AI IP portfolio. The company has been supplying RISC-V processor cores to chipmakers for years, but this funding signals a leap from component supplier to direct competitor in the data center CPU market.

So why would Nvidia invest? Three reasons.

First, Nvidia already uses RISC-V cores inside its own chips. Better RISC-V technology benefits Nvidia directly. Second, if RISC-V CPUs become the standard companion to GPUs in AI data centers, Nvidia wants to be on the inside of that ecosystem, not watching from outside. Third, and most cynically, investing in a potential disruptor gives you intelligence on the disruption. You learn how fast it is coming and how to adapt.

The bigger picture is this: the AI chip market is entering its second phase. Phase one was about who builds the best GPU for training. Nvidia won that. Phase two is about who builds the best infrastructure for running AI at scale. That is a CPU game, a networking game, and a systems game. Nvidia just placed a bet that RISC-V will matter in phase two.

For Intel and AMD, this should be alarming. Their x86 architecture has dominated data center CPUs for decades. Arm is already eating into that from one side. Now RISC-V, with $400 million in fresh capital and Nvidia is blessing, is attacking from the other. The CPU duopoly just got a well-funded third option.

Sometimes the smartest move is funding your own competition before someone else does.