Nuvacore Eyes $2.5B Valuation for AI-Agent Data Center CPU

A six-month-old chip startup that has not shipped a product is now seeking hundreds of millions of dollars at a valuation of roughly $2.5 billion, according to Reuters. The exclusive report, published October 9 and citing two people familiar with the fundraising, positions Nuvacore as the latest bet that the AI infrastructure boom has moved well past graphics processors and into the silicon that orchestrates them.

The San Jose-based Nuvacore is designing a new central processing unit for data centers — the chips that route work to accelerators like Nvidia's GPUs and, critically, run the software behind autonomous AI agents. The Information had previously reported that Nuvacore was raising $200 million or more, but the $2.5 billion valuation figure had not been publicly reported before the Reuters exclusive. Both the size of the raise and the valuation could still change, the sources cautioned, and the round has not yet closed.

The humble CPU is suddenly the hot chip

Graphics processors get the glory in the AI buildout, but central processors are the traffic controllers: they direct data flows, manage workloads, and run the always-on software — including AI agents — that sits on top of the accelerators. As reported by Reuters, appetite for CPUs has swollen because of the thirst for data-crunching capacity to power chatbots such as Anthropic's Claude, and the trend is accelerating as agents scale from demos to always-on systems that run for hours or days.

The market dynamics favor newcomers with a sharp thesis. Intel and Advanced Micro Devices have long dominated the data center CPU market with x86 technology, and companies designing chips based on Arm's technology — including Amazon and Microsoft's custom silicon — have fought to make inroads. But none of the incumbents designed their flagship CPUs from scratch for an era where autonomous agents run persistently across data centers. That gap is exactly where Nuvacore is aiming.

The bet is not without precedent. Venture investors have shown a growing appetite for pre-product semiconductor companies tied to AI workloads, and agent-related infrastructure has been one of the hottest funding themes of the year, from startups building agent runtime software to the data centers that will host it. A $2.5 billion valuation before first silicon suggests investors believe the agent era needs a different kind of CPU — and that whoever builds it could sit at the center of the AI economy's next phase.

Building the chip before choosing its language

What sets Nuvacore apart, according to the company's own description of its strategy last month, is its decision to build the chip's core functionality before committing to an architecture. Instead of starting from x86, the technology behind Intel and AMD chips, or Arm, the dominant design in mobile devices, Nuvacore's engineers plan to design the core first and select the instruction set architecture afterward.

The approach is meant to avoid the limitations and requirements of any specific architecture, letting the team optimize directly for the computations most common in data centers and AI infrastructure. In theory, that freedom could yield a processor tuned precisely for the long-running, compute-intensive workloads that agents generate — workloads that combine orchestration, memory management, and coordination with GPU accelerators.

Industry veterans will note the risk: chip architecture ecosystems carry decades of tooling, software support, and developer familiarity, and sidestepping them is a bold call for a company founded this spring. But the history of computing has a few examples of clean-sheet designs breaking through when the workload shifted decisively, and autonomous agents represent one of the clearest workload shifts the data center has seen in a decade.

The money is moving faster than the silicon

For now, Nuvacore has not developed a product, and the company declined to comment on the fundraising, according to Reuters. The round's structure and final valuation remain in flux, and the two sources spoke on condition of anonymity because they were not authorized to discuss the talks.

Even so, the numbers tell a story about where investor conviction sits. A $2.5 billion price tag for a six-month-old company values the thesis — purpose-built CPUs for the agent economy — ahead of any shipped hardware. With Sequoia backing the company and agents emerging as the defining enterprise software category of the decade, Nuvacore is betting that the software's home hardware will need to be built as deliberately as the agents themselves.

Sources: Reuters via a syndicated exclusive by Max A. Cherney and Krystal Hu; prior reporting on the raise by The Information; summary coverage by newzshort.