KEY POINTS
- Nvidia unveiled the RTX Spark system-on-chip at Computex 2026, pairing a Blackwell RTX GPU with a 20-core Grace CPU via NVLink interconnect in a laptop form factor.
- CEO Jensen Huang called the PC reinvention "as big as the smartphone" and announced partnerships with Microsoft, Dell, HP, ASUS, Lenovo, and MSI for devices shipping later this year.
- Qualcomm, AMD, and Intel shares sold off on the news as the market priced in Nvidia's entry into the $400 billion PC semiconductor market.
Nvidia declared war on the PC chip market Monday at Computex in Taipei, unveiling a system-on-chip called RTX Spark that pairs the company's Blackwell GPU architecture with a custom ARM-based CPU — and sent shares of Qualcomm, AMD, and Intel lower in the process.
CEO Jensen Huang wasted no time framing the stakes. "This reinvention of the computer is as big a deal as the reinvention of the phone into what we now know as the smartphone," Huang told the Computex audience, arguing that agentic AI will transform every personal computer into an autonomous computing platform. The chip, also referred to internally as N1X, debuts later this year in Windows laptops from Microsoft, Dell, HP, ASUS, Lenovo, and MSI.
Inside the Silicon
RTX Spark is not a discrete GPU bolted onto someone else's processor. It is a full system-on-chip built jointly with Taiwan's MediaTek, combining an Nvidia Blackwell RTX GPU with 6,144 CUDA cores and fifth-generation Tensor Cores featuring FP4 precision alongside a high-performance 20-core Nvidia Grace CPU. The two silicon blocks communicate over NVLink chip-to-chip interconnect, the same fabric Nvidia uses to link GPUs in data center racks. Bringing NVLink down to a laptop-class power envelope is a genuine engineering achievement and one that neither Qualcomm's Snapdragon X Elite nor Apple's M-series chips can replicate.
The FP4 Tensor Cores are the detail that matters most for AI workloads. FP4 precision halves the memory footprint of inference models compared to FP8, which means a laptop running RTX Spark can execute large language models locally that currently require cloud connectivity. Nvidia is betting that the next wave of PC demand comes not from faster spreadsheets but from on-device AI agents that manage calendars, write code, and browse the web autonomously.
Who Gets Hurt
The immediate casualties were visible in Tuesday's tape. Qualcomm, which has spent two years building out its Snapdragon X series as the premiere ARM-based Windows chip, saw shares slide as the market recalculated its competitive position. AMD and Intel, which dominate the x86 PC market, also traded lower. Nvidia's entry threatens all three incumbents simultaneously: it brings superior GPU performance to a segment where integrated graphics have been the norm, superior AI inference capabilities relative to Qualcomm's NPU, and the NVLink interconnect advantage that no competitor can match in 2026.
But Nvidia also needs Microsoft. The RTX Spark announcement was a joint effort, with Microsoft committing to optimize Windows for Nvidia's chip and integrate its Copilot AI assistant natively. That partnership is critical because Qualcomm's existing Snapdragon X laptops have struggled with app compatibility on Windows ARM, and Nvidia needs to avoid the same trap. The MediaTek partnership helps here — MediaTek brings decades of ARM SoC experience from the smartphone market and handles much of the platform validation work.
The Vera CPU Factor
Beyond the consumer PC announcement, Huang revealed that Nvidia's Vera CPU for data centers is now in full production, with millions of units shipping for what Huang described as "a market that never existed before." Vera, available starting this fall, positions Nvidia to sell complete server nodes — GPU, CPU, networking, and interconnect — without relying on AMD or Intel for any component. For traders, this is the more consequential long-term development. A vertically integrated Nvidia that owns every layer of the data center stack commands higher margins and deeper customer lock-in than a GPU vendor dependent on third-party CPUs.
Jim Cramer called the Computex keynote a "[roadmap of winners in the AI boom](https://www.cnbc.com/2026/06/01/jim-cramer-jensen-huang-computex-keynote-revealed-winners-ai-boom.html)," singling out Marvell Technology, which closed up 33% Tuesday after Huang name-checked its networking silicon. The halo effect from Computex announcements tends to last through the week as analysts publish updated models.
What Traders Should Watch
The PC chip market generates roughly $400 billion in annual revenue, and Nvidia has just announced its intention to compete for a share. The RTX Spark laptops ship in Q4 2026, which means the revenue impact hits Nvidia's fiscal Q4 and Q1 2027 results. Watch Qualcomm's investor day on June 18 for its competitive response. The key level on NVDA is $175; a sustained break above would target $190 and confirm that the market is pricing in the PC TAM expansion on top of existing data center growth.
