Nuvacore, a six-month-old semiconductor startup based in San Jose, California, is developing a central processing unit for servers that power AI workloads. Rather than committing to an established instruction set architecture such as x86 or Arm, the company plans to build its processor's core functionality first and optimize it for computations most common in AI infrastructure. The startup was founded by Gerard Williams, John Bruno, and Ram Srinivasan, chip design veterans; Williams previously sold his startup Nuvia to Qualcomm for $1.4 billion in 2021.
Led by Sequoia Capital, which led the seed round; the current round has not closed and investors have not been fully disclosed.
Nuvacore's reported $2.5 billion valuation, reached before shipping a single product, illustrates how intensely investors are betting on alternatives to the x86 CPU duopoly of Intel and AMD in AI data centers. CPUs have become a critical bottleneck in AI infrastructure, acting as traffic controllers for Nvidia accelerators and running autonomous software agents, which has made the market attractive enough to support pre-revenue valuations at this scale. Broader context reinforces the trend: semiconductor startups raised roughly $10.7 billion in the first five months of 2026 alone, already approaching the $12.2 billion raised globally in all of 2025.