Nvidia Walks Into the PC Chip Market With RTX Spark, and Intel Drops 7%
At Computex on June 1, Jensen Huang unveiled the RTX Spark Superchip, a 20-core Arm CPU fused to a 6,144-core Blackwell GPU with 128GB of unified memory, shipping this fall in laptops from Dell, HP, Lenovo, Microsoft Surface, ASUS and MSI. Intel fell as much as 7.3% on the news; Nvidia rose about 4%. The awkward twist: Nvidia still holds a $5 billion stake in Intel from last September's partnership, and the chip is fabbed on TSMC's 3nm node with MediaTek doing the CPU integration.
Nvidia's second attempt at the PC processor market is nothing like its first. The RTX Spark is a 70-billion-transistor SoC built as two chiplets joined by NVLink-C2C at 600GB/s: a 20-core Grace-derived Arm CPU on one side, a Blackwell GPU with 6,144 CUDA cores and fifth-generation FP4 tensor cores on the other, sharing up to 128GB of unified memory at 300GB/s. Power envelope: 45W to 80W. Nvidia is rating the AI side at one petaflop of FP4. That is a workstation-class number in a laptop thermal budget, and it is the entire reason Microsoft agreed to make Windows on Arm a first-class target again.
The lineup is unusually broad for a launch SKU. Asus ProArt P14 and P15, Dell XPS 16, HP OmniBook X 14 and Ultra 16, Lenovo Yoga Pro 9n, Microsoft Surface Laptop Ultra, and MSI Prestige N16 Flip AI are all confirmed for fall. Leaked pricing for the higher-end N1X variant starts around $2,900, which tells you who this is actually for: the premium creator and developer tier where Apple's M-series has been running unopposed since 2020. Nvidia is not trying to displace the $700 Intel laptop. It is trying to take the $2,500-and-up segment that funds the entire Windows OEM business, and it is doing it by handing OEMs a part that finally answers the MacBook Pro question.
The strategic damage to Intel is severe and the Intel-Nvidia equity arrangement makes it weirder. Nine months ago Nvidia took a $5 billion stake in Intel as part of a foundry-and-x86 cooperation deal that was sold as Intel's lifeline. On Monday Nvidia announced a part that competes head-on with Intel's Panther Lake and Clearwater Forest roadmap, fabbed at TSMC instead of Intel Foundry, and ships in the exact laptop SKUs Intel needs to defend. AMD takes collateral damage, since Strix Halo was supposed to own the high-end Windows AI PC tier this year, but Intel is the one whose core business model the RTX Spark is built to dismantle. The Vera CPU for datacenters going into full production the same day, also announced Monday, is the same playbook running in parallel against AMD's EPYC: the hyperscalers spending $700 billion on capex this year are not going to leave a single layer of Nvidia's stack on the table if the price-performance lines up.
The piece worth watching is software. Windows on Arm has shipped twice before and twice failed because x86 emulation was slow and native Arm app coverage was thin. What is different now is that the apps people actually buy a $2,900 laptop for, Adobe Creative Cloud, the AI coding stacks, local LLM runtimes, are the apps Nvidia controls the optimization story for. If RTX Spark ships in October with native Photoshop, native Premiere, native Ollama, and native CUDA-on-Arm developer tooling, the Windows-on-Arm transition that Qualcomm could not force in three tries will happen in one generation because the GPU is doing the convincing, not the CPU. We argued in March that the Arm Windows transition needed a GPU-first incumbent to ever land, and Nvidia just became that incumbent. Intel has roughly eighteen months to ship a competitive answer before the premium-laptop ASP curve permanently shifts.
CNBC — Nvidia's new PC chips are CEO Huang's bid to win every layer of AI stack → · Tom's Hardware — Nvidia unveils RTX Spark Superchip at Computex 2026 → · NVIDIA Newsroom — NVIDIA and Microsoft Reinvent Windows PCs → · CNBC — Nvidia jumps into PCs with new Arm-based chip →