Full Breakdown
Nvidia Unveils RTX Spark AI Chip and AI-Focused Windows PCs at Computex 2026
6/7/2026, 10:53:40 AM
Nvidia’s RTX Spark SoC and AI-Centric Windows Machines
At Computex 2026 in Taipei, Nvidia announced its RTX Spark system-on-chip, an Arm-based design that pairs a 20-core Grace CPU (co-developed with MediaTek) with a Blackwell-generation RTX GPU. The chip delivers up to 1 petaflop of AI compute, 6,144 CUDA cores, and up to 128 GB of unified memory. Nvidia also introduced the DGX Station for Windows, a desk-side AI supercomputer that can handle up to one trillion parameters. Partner OEMs—including HP, Dell, ASUS, Lenovo, Microsoft Surface, and MSI—will ship laptops and compact desktops using RTX Spark in the fall of 2026.
Background: The Push for an “AI PC”
Nvidia’s prior AI hardware, such as the DGX Spark workstation, positioned the company in data-center markets. The “AI Together” theme at Computex signaled a shift toward bringing comparable compute to the desktop, echoing Microsoft’s earlier Copilot+ effort and the broader industry desire to reduce cloud latency for AI agents.
Key Players and Their Roles
- Nvidia – CEO Jensen Huang (vision of an “agentic computer”), VP of Enterprise Platforms Chris Marriott (enterprise AI infrastructure).
- HP Inc. – SVP and Division President Jim Nottingham (enterprise Windows integration).
- OEM Partners – ASUS, Dell, Lenovo, MSI, Microsoft Surface, Acer, GIGABYTE.
- Software Ecosystem – Adobe (native support for Photoshop and Premiere Pro on RTX Spark).
Technical Specifications and Pricing
| Device | CPU | GPU | Unified Memory | AI Compute | Approx. Price |
|---|---|---|---|---|---|
| RTX Spark laptops (N1) | 10 × Cortex-X925 + 10 × Cortex-A275 | 6,144 RTX Blackwell cores (RTX 5070 class) | up to 128 GB | ~1 PFLOP | $1,799 (analyst estimate) |
| RTX Spark laptops (N1X) | same | same | up to 128 GB | same | $2,899 (analyst estimate) |
| HP ZGX Fury GB300 workstation | Nvidia GB300 Grace-Blackwell Superchip | Top-end Nvidia GPU | up to 784 GB | up to 20 PFLOPS FP4 | $94 k–$200 k (reseller reports) |
| Nvidia DGX Station (Windows) | GB300 Superchip | Nvidia GPU array | large (unspecified) | up to 1 trillion-parameter inference | $94 k+ (mid-range) |
Why It Matters
If RTX Spark delivers the promised performance, Windows PCs could achieve an “Apple Silicon moment,” offering high-bandwidth unified memory and GPU-centric AI acceleration previously limited to Macs. Enterprise customers could run trillion-parameter models locally, reducing reliance on cloud services and per-token fees. The launch also intensifies competition with AMD’s Ryzen AI Ultra line and Qualcomm’s Snapdragon X Elite Copilot+ devices.
Official Statements & Responses
- Nvidia described the DGX Station as a “desk-side AI supercomputer” designed for enterprises scaling AI agents.
- HP confirmed that the ZGX Fury GB300 will ship alongside Nvidia’s DGX Station in Q4 2026, targeting Windows-based enterprise environments.
- Microsoft is collaborating with Nvidia to improve Windows 11 support for the RTX Spark Arm platform, including Prism emulation for legacy x86 applications.
Criticism & Opposition
Industry analysts caution that first-generation Arm-plus-GPU designs face uncertain battery life, Windows-on-Arm compatibility issues, and premium pricing that may limit adoption to deep-pocketed developers and content creators. PC-gaming journalists highlighted public skepticism toward “agentic” PCs, noting concerns about loss of direct user control and the reliability of AI-generated outputs.
Conflicting Reports & Gaps
- Pricing: Analyst notes suggest $1,799–$2,899 for laptops, while Engadget projects a $4,000+ starting price.
- Release timing: HP cites “later 2026,” Nvidia mentions “fall 2026,” and HP’s Q4 2026 window overlaps but is not precise.
- Memory: Laptops are limited to 128 GB, whereas the workstation advertises 784 GB, creating a gap in performance expectations.
- CPU performance: Nvidia’s 20-core Grace CPU uses Cortex X925 and A275 cores, described as slower than newer Arm cores, yet no benchmark data are provided.
Verbatim Quotes
- “As enterprises scale AI agents across their organizations, they need AI infrastructure that can connect directly to the applications and workflows that power their business,” — Chris Marriott, VP of Enterprise Platforms, Nvidia
- “Over 70% of enterprise PCs run Windows, and our customers have asked for AI supercomputing power that can seamlessly integrate into their existing environments,” — Jim Nottingham, SVP and Division President, Advanced Compute and Solutions, HP Inc.
- “Your personal computer is really the world's largest edge device, and it's 40 years old,” — Jensen Huang, CEO, Nvidia
- “We're now reinventing the computer. It's an agentic computer. It's an agent now, it's an assistant, not a tool. And that's the big idea,” — Jensen Huang, CEO, Nvidia
- “And while it continually improves, the level of public trust around it seems shaky at best.” — PC Gamer commentary
What’s Next
Nvidia, HP, and partner OEMs plan to begin shipments of RTX Spark-based laptops and the ZGX Fury GB300 workstation in the fourth quarter of 2026. Rumored stripped-down variants for 2027 could broaden access beyond enterprise and AI-enthusiast segments, while pricing, battery life, and Windows-on-Arm software support remain key uncertainties.
