Key Takeaways
Nvidia, Qualcomm, Intel and Arm declared the AI Agent era has arrived, signaling a new demand cycle for chips across the semiconductor industry.
Nvidia, Qualcomm, Intel and Arm used Computex 2026 in Taipei to declare the AI Agent era has arrived, with executives from all four chip giants forecasting a surge in processor demand as artificial intelligence shifts from answering questions to completing tasks autonomously.
"This is the new computing pattern — tokens are now profitable units of revenue," Nvidia Chief Executive Officer Jensen Huang said during his two-hour keynote. "The more you buy, the more you make."
Qualcomm CEO Cristiano Amon declared "2026 is the year of AI Agent" in his own keynote, while Intel said multiple chief executives are calling to demand CPU supply for agentic workloads. Arm added that CPU demand is rising faster than originally projected as AI agents require more orchestration and low-latency processing than traditional inference tasks.
The consensus across the industry's largest chipmakers points to a capital expenditure cycle that could reshape semiconductor demand for years. Nvidia alone said it expects to spend $150 billion annually in Taiwan, where it is building its Constellation campus, while its newly announced Vera Rubin platform — now in full production — is designed specifically for agentic AI workloads.
Vera Rubin and the CPU Renaissance
Nvidia's Computex keynote centered on Vera Rubin, its next-generation data center platform now entering volume production. The system pairs the Vera CPU — which Huang claimed delivers the "highest instructions per clock in the world" at 10 per cycle — with Nvidia's latest GPU architecture in an NVL72 configuration that eliminates cables between compute trays.
The Vera CPU is built for what Nvidia calls "impatient agents," offering 1.8 times the agentic performance of x86 processors, according to the company. Huang said Vera will be "the most optimized agentic CPU in the world" and predicted the CPU market for agentic workloads "will surely be larger than the last because there will be more agents than people."
Assembly times for a single Grace Blackwell rack have fallen to five minutes, Huang said, as Nvidia works to accelerate deployment of its AI infrastructure. The company also unveiled a roadmap extending the RTX Spark consumer chip line through Vera Rubin and Rosa Feynman generations in 2028 and 2030, respectively.
Qualcomm, Intel and Arm Stake Their Claims
Qualcomm used Computex to unveil the Snapdragon C platform, targeting $300 entry-level Windows laptops even as the industry navigates a memory pricing crisis. Amon positioned the chip as a way to keep budget devices affordable while bringing AI capabilities to lower price points.
Intel announced its Arc G3 and G3 Extreme processors, custom chips built specifically for handheld gaming devices. The Panther Lake-based chips feature two fewer CPU cores than Intel's standard laptop processors but include a full complement of Xe3 graphics cores — up to 12 — to run games. The chips will power the Acer Predator Atlas 8 and MSI Claw 8 EX AI Plus, among other devices.
Arm, whose architecture underpins both Nvidia's RTX Spark and Qualcomm's Snapdragon C, said the AI Agent transition is accelerating CPU demand beyond its earlier forecasts. The company posted the same "new era of PC" teaser as Nvidia and Microsoft ahead of Computex, signaling alignment across the Windows-on-Arm ecosystem.
Supply Chain Implications
The four largest Taiwan-based electronics manufacturers — Hon Hai Precision Industry, Quanta Computer, Wistron and Pegatron — appeared on stage together to discuss AI Agent infrastructure buildouts, a rare joint appearance underscoring the scale of expected demand.
Nvidia's RTX Spark family, built on TSMC's 3-nanometer process with 70 billion transistors, 20 CPU cores and 6,144 CUDA cores, represents the company's first consumer PC system-on-chip. More than 30 laptops and 10 desktops are in development with partners including Microsoft, Asus, Dell, HP, Lenovo and MSI, with availability slated for the fall.
For investors, the question is whether the AI Agent thesis can sustain the valuation multiples the semiconductor sector currently commands. Nvidia shares trade at elevated multiples of forward earnings as the market prices in years of data center growth. If the agentic AI transition materializes as quickly as the four chip giants suggest, the demand for CPUs — long considered a mature, low-growth market — could see a structural re-rating. If adoption disappoints, the inventory build across the supply chain could mirror the corrections seen after previous technology hype cycles.
This article is for informational purposes only and does not constitute investment advice.
