NVIDIA CEO Jensen Huang took the COMPUTEX 2026 stage in Taipei this week to launch what the company describes as the next chapter of accelerated computing, with robotics and physical AI promoted from supporting players to the heart of the message. The June 1 keynote — opening NVIDIA's GTC Taipei conference — drew the full Taiwan ICT supply chain to Taipei Music Center.
Vera Rubin Headlines
Huang detailed the production roadmap for the Vera Rubin AI computing platform, paired with the new Vera CPU and Spectrum-X Ethernet networking fabric for AI factories. The combination targets million-GPU training systems while squeezing more useful work out of each watt of power — a pitch aimed squarely at hyperscalers building out gigawatt-scale data centers.
Physical AI Comes Of Age
Robotics took multiple keynote segments. Deepu Talla, NVIDIA's robotics and edge AI lead, walked the audience through the company's Isaac robotics libraries, Cosmos world models, and Jetson Thor on-device compute. Huang emphasized that the trillion-dollar humanoid and autonomous machine market sits on top of the same accelerated computing stack as cloud AI — including the recent Cadence-NVIDIA multiphysics partnership for sim-to-real.
Spectrum-X For AI Factories
Huang highlighted broad adoption of Spectrum-X, which NVIDIA bills as the first Ethernet fabric purpose-built for AI workloads. New customers were announced alongside reference designs that fold Spectrum-X into Vera Rubin pods and on-prem AI factories — relevant for robotics OEMs running large-scale synthetic data and policy training.
Ecosystem Roster
COMPUTEX 2026 saw participation from Intel, Marvell, Qualcomm, Samsung, NXP, Infineon, Synopsys, Cadence, Texas Instruments, Siemens, Supermicro, Vertiv, Schneider Electric, and Google — a who's-who of partners NVIDIA needs to keep on board as it pushes deeper into networking, edge silicon, and physical AI.
What To Watch Next
Production timing on Vera Rubin remains the swing factor for late-2026 capex plans, and TSMC's CoWoS-L packaging supply has been called out as a constraint in our earlier coverage of NVIDIA's Taiwan supply strain. Robotics teams should also track Isaac Lab integration with Genesis AI's GENE-26.5 and the Cosmos world model release cadence.
Related: NVIDIA-IREN 5GW AI factory deal.
Reporting based on coverage from NVIDIA Newsroom, Tom's Hardware, ServeTheHome, and TechRadar.