Nvidia opened Computex 2026 in Taipei with a major strategic announcement: it's going after the $200 billion CPU market, and it's bringing some of the biggest names in PC hardware along for the ride.
The centerpiece is the RTX Spark, a new PC CPU that Nvidia is calling a "superchip." Delivering one petaflop of AI compute, it's specifically engineered to run AI agents locally, securely, and — Nvidia claims — usefully. The first wave of RTX Spark–powered PCs will ship this fall from ASUS, Dell, HP, Lenovo, and Samsung.
What makes this more than another chip launch is the integration story. Microsoft has worked alongside Nvidia to develop secure sandboxes baked into the hardware, allowing AI agents to run locally without exposing sensitive data to the cloud. This effectively makes each AI agent PC a self-contained secure execution environment — a pitch squarely aimed at enterprise buyers who've been nervous about cloud-resident agents touching corporate data.
The system has enough CPU, GPU, and RAM headroom to run large language models directly on the device, with Nvidia's CUDA software stack providing the glue. Beyond agents, RTX technology promises faster rendering, better in-game visuals, and compatibility across more than 1,000 applications and games.
The timing matters. Apple's M-series chips have spent the last several years demonstrating that tightly integrated silicon can dramatically outperform traditional PC architectures on performance-per-watt. Microsoft's Windows on ARM push never fully gained traction, but RTX Spark changes the competitive dynamic by pairing Nvidia's dominant AI ecosystem — CUDA, thousands of optimized applications — with a fresh CPU architecture.
For enterprise IT leaders, the secure-agent sandbox story will likely be the headline, especially as organizations look to deploy internal AI assistants and automation tools without ceding control to hyperscaler infrastructure. For consumers, the promise of running capable AI workloads locally without a cloud subscription could be compelling.
Whether Nvidia can actually make a dent in the CPU market remains to be seen — Intel and AMD have massive distribution moats — but the AI agent angle gives it a differentiated story that raw GHz benchmarks can't easily counter.
Why It Matters
If AI agents become the primary use case that drives PC purchases, then the company controlling the best agent-optimized silicon controls the future of the PC market. Nvidia's move into CPUs isn't just about chips — it's a bid to become the default infrastructure layer for the AI-native computing era.