NVIDIA CEO Jensen Huang unveiled the company's first PC super chip and a new CPU designed for AI agents during a keynote at COMPUTEX 2026 on June 1. The RTX Spark chip, developed with MediaTek using 3nm process technology, integrates a Blackwell RTX GPU with a 20-core Grace CPU to deliver 1 Petaflop of AI performance for Windows laptops and small workstations. Huang also announced the Vera CPU, claiming it runs 1.8 times faster than x86 processors and supports diverse AI workloads across industries. The product launches coincide with NVIDIA's thesis that compute power directly translates to revenue and profit as AI factories scale investments from $20-30 billion to $80-100 billion per gigawatt. This marks NVIDIA's formal entry into the PC processor market long dominated by Intel and AMD, with mainstream manufacturers including Dell, Lenovo, ASUS, and HP set to ship RTX Spark-powered devices this fall.
NVIDIA Launches Vera CPU with 1.8x x86 Performance Advantage
Huang stated that traditional CPUs were designed for humans operating in "second-scale worlds," while AI agents function in "nanosecond-scale worlds" with extreme latency sensitivity. NVIDIA built the Vera CPU from the ground up to serve AI agents, achieving the world's highest instructions per clock cycle (IPC) and performance 1.8 times faster than x86 processors. The chip provides up to 1.2 TB/s bandwidth and is the first processor to support PCIe Gen 6 and LPDDR5X memory.
Planned adopters of the Vera CPU include the New York Stock Exchange, Anthropic, OpenAI, and SpaceX, as well as cloud service providers ByteDance, CoreWeave, and Oracle Cloud Infrastructure. System manufacturers Dell Technologies, HPE, Lenovo, and Supermicro are integrating Vera into their AI infrastructure products. NVIDIA disclosed that Grace CPU shipments have approached 2.5 million units to date.
RTX Spark Chip Enters PC Market with 1 Petaflop AI Performance
The RTX Spark super chip, co-developed with MediaTek on TSMC's 3nm process, combines a Blackwell RTX GPU with a 20-core Grace CPU and runs Microsoft's Arm-based Windows operating system. The platform delivers 1 Petaflop of top-tier AI compute power and features 128GB of unified memory, enabling local operation of advanced AI agents, rendering of massive 3D scenes, and editing of 12K ultra-high-definition video.
Huang stated that NVIDIA and Microsoft have "completely re-architected the PC," positioning the technology as a reinvention of personal computing that began with Windows 95 forty years ago. NVIDIA Product Management Senior Director Mark Aevermann said the RTX Spark series offers graphics processing roughly equivalent to the RTX 5070 mobile GPU, with CPU performance "expected to match any other product on the Windows platform."
Starting this fall, Dell, Lenovo, ASUS, and HP will launch laptops and desktops powered by the RTX Spark super chip. Adobe is restructuring Photoshop and Premiere Pro to fully utilize the RTX Spark architecture. Pricing for the products has not been announced.
AI Factory Investment Scales to $100 Billion Per Gigawatt
Huang emphasized that NVIDIA's role has evolved beyond manufacturing GPUs and systems to helping customers build complex "AI factories." Single-gigawatt AI factory investments are rising from $20-30 billion to $80-100 billion, requiring systems to operate flawlessly from first deployment. He stated that in AI factories, electrical power (gigawatt capacity) is a fixed ceiling, making "the number of tokens generated per watt" a direct determinant of company revenue.
NVIDIA introduced the DSX platform to provide infrastructure builders with a complete blueprint for creating AI factories. Huang stated: "We are not just shipping chips; we are providing every infrastructure builder with a complete guide to building AI factories. With the DSX platform, you can simulate the entire factory without spending a cent, verify performance before installing a single rack, and operate with the reliability required for production-grade AI."
The company announced that its new-generation GPU computing platform Vera Rubin is "the most ambitious project in company history" and has entered full production. The supply chain scale for Vera Rubin is twice that of the previous Grace Blackwell generation.
NVIDIA Releases Cosmos 3 Model and Alpha Mayo 2 Autonomous Driving System
In the physical AI domain, NVIDIA released the Cosmos 3 open frontier model. The model uses a Transformer hybrid architecture to understand and simulate the physical world, addressing the scarcity of first-person perspective data in robotics research and development. Based on this technology, NVIDIA also released Alpha Mayo 2, described as the world's first autonomous driving model capable of thinking and reasoning, along with NVIDIA Isaac GR00T, a complete open-source reference platform for humanoid robot development.
Huang spent significant time asserting that "useful AI has arrived" and that agentic AI is creating more jobs rather than reducing them. He cited GitHub code commits nearly tripling from 500 million in 2025 to 1.4 billion in the first months of 2026, stating: "The number of software engineers is increasing. People talk about AI reducing jobs—that's complete nonsense. In reality, more software engineers are being hired."
FAQ
What did NVIDIA announce at COMPUTEX 2026 on June 1?
NVIDIA CEO Jensen Huang announced the RTX Spark super chip for Windows PCs and the Vera CPU designed for AI agents. The RTX Spark chip, developed with MediaTek using 3nm process technology, integrates a Blackwell RTX GPU with a 20-core Grace CPU and delivers 1 Petaflop of AI performance with 128GB unified memory. The Vera CPU runs 1.8 times faster than x86 processors and provides up to 1.2 TB/s bandwidth, supporting PCIe Gen 6 and LPDDR5X memory.
When will RTX Spark-powered PCs become available?
Starting this fall, mainstream PC manufacturers including Dell, Lenovo, ASUS, and HP will launch laptops and desktops powered by the RTX Spark super chip. Adobe is restructuring Photoshop and Premiere Pro to fully utilize the RTX Spark architecture, though product pricing has not been announced.
Which companies plan to adopt the NVIDIA Vera CPU?
Planned adopters of the Vera CPU include the New York Stock Exchange, Anthropic, OpenAI, and SpaceX, as well as cloud service providers ByteDance, CoreWeave, and Oracle Cloud Infrastructure. System manufacturers Dell Technologies, HPE, Lenovo, and Supermicro are integrating Vera into their AI infrastructure products.
