Jensen Huang told an audience at GTC Taipei on Sunday that Nvidia’s Vera data-centre CPU is in full production, with Anthropic, OpenAI, SpaceX, Oracle and CoreWeave among its first customers and millions of units already being manufactured for fall shipment. Yahoo Finance reported ByteDance and Dell are also evaluating the part.

The framing matters. Nvidia spent a decade defining the accelerator layer above the CPU; Vera is the move into the silicon underneath. Huang called the roughly $200 billion CPU segment “a new major growth driver” and described what Nvidia is building as “a market that never existed before.” Ian Buck, the company’s VP of hyperscale and high-performance computing, said Vera produces tokens 1.8 times faster than current x86 processors, the metric that translates directly into inference economics for the named customers.

Hours earlier, Huang unveiled RTX Spark, an Arm-based PC superchip co-developed with Microsoft and MediaTek and fabricated on TSMC’s 3-nanometer node. Six PC makers, Microsoft, Dell, HP, ASUS, Lenovo and MSI, will ship laptops and compact desktops using it this fall. An Nvidia spokesperson said the collaboration had been running for “many, many years.” The company also said supply won’t be constrained by the global memory shortage.

Markets read it as encirclement. In U.S. pre-market trading, AMD and Intel fell nearly 4 percent, Qualcomm almost 7 percent, and Apple 0.6 percent, while Microsoft rose 3.1 percent. Huang also told reporters to expect “a surprise new product” in the second half of 2026.

Nvidia is now pricing the CPU, the GPU and the laptop on the same roadmap. The competitive question isn’t whether incumbents respond. It’s whether their architectures were ever the product customers were actually buying.

Sources