Nvidia details its Vera CPU for data centers, its first CPU with a custom core design, featuring 88 cores and 176 threads, set for general release in H2 2026
Nvidia's Vera CPU is its first bid to become a key player in the data center CPU market. Although Grace has seen some success …
Nvidia now has a defined 88-core, 176-thread CPU offer for data-center buyers planning deployments for the second half of 2026, moving Vera from roadmap architecture toward a purchasable platform.
The custom-core design gives Nvidia more control over the CPU component of its data-center systems rather than relying solely on its prior Arm-based CPU approach.
Second-order effects
Intel and AMD face a more direct challenge in AI-oriented server deployments where customers can evaluate Nvidia's CPU alongside its broader compute platform; early Vera benchmark reporting has already framed that comparison.
Server buyers and system partners will need to weigh Vera's CPU characteristics against the operational advantages of adopting a more tightly integrated Nvidia rack design.
Third-order effects
If customers adopt CPU-plus-accelerator systems from the same vendor, AI infrastructure competition could shift further from individual chips toward integrated, rack-scale platforms.
The strategic value of owning the CPU layer rises as vendors seek to control performance, memory design, and system integration across heterogeneous AI compute.
The trend: Vera is part of the shift toward vertically integrated AI infrastructure, in which chip vendors increasingly design CPUs, accelerators, and rack-scale systems as a single platform.
Most forecasts have total $NVDA racks between 70-80K for 2026. That is up 2X over 2025, so if we assume 2X growth into 2027, then that's 150-160k in 2027. Whether or not the spare GW exists to support that many is another question. One we are tracking @DiligenceStack
Interesting report from @theinformation that Nvidia will be able to make 1000 Vera Rubin racks per day, which is $630b per quarter. Actually a little hard for me to believe and haven't checked the math, but wow if true. And Vera CPU racks and Groq LPU racks would be incremental
👀Nvidia says it will be able to produce up to 1,000 Vera Rubin racks ~PER DAY~ If it does, that would generate >$630B in revenue per quarter for Nvidia & manufacturing partners [image]
Gavin Baker on the 1,000 Vera Rubin racks a day report: “a little hard for me to believe and haven't checked the math, but wow if true” The math checks. 1,000 x $7M x 90 days is $630B a quarter, system level Consensus doesn't. Last July the street's five-years-out NVDA estimate […
NVIDIA included me in an embargoed technical review of the Vera CPU architecture ahead of today's whitepaper release on Vera The following are my key takeaways from the information they shared: The headline specs were mostly on the roadmap already - 336B transistors, 288GB of [im…
The first-ever measured silicon numbers for @NVIDIA Vera Rubin NVL72 are in 😲 First measured performance shows 10x more tokens per megawatt than Blackwell. No projections. Real results from live hardware. [video]
@nvidia dropped the full Vera CPU architecture disclosure this morning, one day before @AMD takes the stage at Advancing AI. I wrote up what was revealed, why the pendulum is swinging back to per-core performance after a decade of core-count scaling, and which benchmark baselines
10x more tokens per megawatt. CoreWeave has the first measured performance of NVIDIA Vera Rubin NVL72, showing 10x improvement in tokens per second per megawatt on DeepSeek-R1 compared to Blackwell.
💡 NVIDIA Vera is the CPU for agents and benchmark results from @DeepInfra show that it's more than 2x as fast compared with other CPUs. Read the results ⤵️
$NVDA says its next-generation Vera Rubin platform remains on schedule with major customers already testing the hardware. The company also claims its new Vera CPU is faster than $AMD Turin processor [image]
AI agents are only as fast as the CPU powering them — so we redesigned the CPU. NVIDIA Vera with Olympus Cores delivers up to 1.8x higher performance on agentic workloads, with the memory bandwidth and per-thread speed to keep AI factories running at full tilt. Read more: [image]
🚀 The NVIDIA Vera Rubin platform is here, with 10x better performance per watt. ➡️ The NVIDIA ecosystem, including @CoreWeave, @GoogleCloud, @Microsoft, and @Oracle Cloud, are standing up NVIDIA Vera Rubin NVL72 to deliver the lowest token cost for the agentic era. ➡️ NVIDIA [ima…
We run agents in production. So when NVIDIA built Vera CPU for agents, we benchmarked it ourselves. @nvidia Vera CPU hit 2.2× the best x86 on agentic orchestration and was the fastest of all four architectures we tested. Methodology locked before the hardware arrived 👇 The