Jensen Huang does not do quiet product launches. At Computex 2026 in Taipei, the Nvidia CEO unveiled RTX Spark — a superchip that combines a Blackwell GPU, a 20-core Arm CPU, and up to 128 GB of unified memory into a single package destined for Windows laptops and compact desktops. The announcement wiped billions off the market caps of Intel, AMD, and Qualcomm in a single trading session.
"For forty years, you launched apps," Huang told the packed audience. "With RTX Spark and Microsoft Windows, you ask — and the PC does the work." That line encapsulates exactly what Nvidia is selling: not just faster hardware, but a fundamentally different relationship between user and machine, mediated by on-device AI agents running without any cloud dependency.
What the RTX Spark Superchip Actually Contains
Strip away the marketing and the silicon is genuinely impressive by any measure:
- A Blackwell RTX GPU core carrying 6,144 CUDA cores alongside fifth-generation Tensor Cores capable of FP4 precision inference — the same architecture found in Nvidia's data center products, shrunk for mobile power envelopes
- A 20-core Grace CPU based on Arm architecture, co-designed with MediaTek — Nvidia's first serious foray into PC processor silicon
- Up to 128 GB of unified memory shared between CPU and GPU workloads with no transfer overhead, connected via Nvidia's NVLink-C2C chip-to-chip interconnect
- Peak AI compute of 1 petaflop — enough to run frontier-class language models entirely on-device
- Full support for DLSS 4.5, hardware ray tracing, Nvidia Reflex, and the complete GeForce gaming stack
The Software Ecosystem Already Waiting
Nvidia has lined up over 100 software partners ahead of launch. Adobe is rebuilding Photoshop from the ground up as a fully GPU-accelerated application for RTX Spark, unlocking generative AI editing workflows and HDR color grading at speeds not previously possible on laptop hardware. Riot Games and Xbox have both committed platform support, and over 1,000 existing RTX-enhanced titles carry over immediately.
The AI agent layer runs through Nvidia's OpenClaw and Hermes Agent frameworks — local, private, no subscription required. Microsoft is deeply embedded in the platform, with Satya Nadella expected to join Huang to demonstrate the "new era of Windows" built around Spark's capabilities.
When You Can Buy One
Systems from ASUS, Dell, HP, Lenovo, Microsoft Surface, and MSI are scheduled for fall 2026. Acer and Gigabyte will follow. Nvidia has already named the next two generations: Vera Rubin with LPDDR6 memory, and Rosa Feynman beyond that — signaling a long-term platform commitment rather than a one-off product line.
Pricing has not been announced, but given the silicon density involved, expect these machines to occupy the premium tier of the laptop market when they arrive.