Nvidia, long dominant in enterprise and data center computing, is betting that a new agentic AI platform will accelerate adoption of AI PCs.
The chipmaker unveiled RTX Spark, which is aimed at turning personal computers into local AI supercomputers. Nvidia said the platform can handle 120-billion parameter large language models and offers a context window that can fit a prompt of around 750,000 English words, or a million tokens.
The RTX Spark expands Nvidia’s platform from “cloud AI factories to personal AI endpoints,” wrote BofA Global Research analyst Vivek Arya in a June 1 research note.
He noted that RTX Spark combines a Blackwell RTX GPU with 6,144 CUDA (GPU) cores and a 20-core Grace CPU over NVLink-C2C, with MediaTek collaborating on the custom CPU design. Arya said rivals such as Apple Mac Mini has 10 to 12 CPU cores and the MacBook Pro has 15 to 18 CPU cores.
Arya said RTX Spark likely targets the top 10% of the Windows market.
RTX Spark is designed to run AI agents on the device itself rather than in the cloud. Nvidia and Microsoft are working together on new Windows security features and software tools for AI agents.
Dell, HP, Lenovo, ASUS and Microsoft Surface plan to launch RTX Spark systems later this year, Nvidia said. Adobe is also optimizing Photoshop and Premiere for RTX Spark systems, while developers including OpenClaw and Hermes Agent are building AI agent applications for the platform.
“The PC is being reinvented,” said Jensen Huang, founder and CEO of Nvidia, in a statement. “For 40 years, you launched apps. Click. Type. With RTX Spark and Microsoft Windows, you ask — and the PC does the work.”
“RTX Spark brings everything NVIDIA has built — CUDA, RTX, our AI platform — into a single superchip,” Huang continued. “Local agents. Frontier models. Creative workflows. RTX games. All on a laptop. This is the new PC. The personal AI computer.”
New AI CPU for agents
Nvidia also unveiled Vera, its first CPU designed specifically for AI agents in data centers, signaling the company’s ambitions to reshape the broader server CPU market long dominated by Intel and AMD.
Nvidia said Vera is now in full production and delivers 1.8 times faster task completion than traditional x86 CPUs on workloads tied to agentic AI, reinforcement learning and data processing. Companies including Anthropic, OpenAI, Oracle and the New York Stock Exchange are exploring or adopting the processor, Nvidia said.
Arya called Vera the “single greatest new addition since the GPU.” He predicted that Nvidia is poised to become the largest server CPU vendor by next year.
Vera builds on Nvidia’s Grace line of CPUs, which were designed for general AI and high-performance computing workloads rather than specifically for AI agents. Vera redesigns the CPU around AI agents that can run code, use tools and orchestrate multistep tasks.
Unlike traditional x86 server processors, Vera uses Nvidia’s custom Arm-based Olympus cores and is designed to move data more efficiently between CPUs and GPUs. Nvidia said Vera includes 88 custom cores, LPDDR5X memory and tight integration with Nvidia GPUs through its NVLink interconnect.
“AI agents will be the largest users of computing,” Huang said in a separate statement. “Vera is the first CPU designed for that future.”
Be First to Comment