COMPUTEX 2026 is shaping up to be the biggest edition yet, and NVIDIA is parking its entire GTC roadshow in Taipei to make sure no one misses it. Running June 1–4 at the Taipei International Convention Center, NVIDIA GTC Taipei will bring together developers, researchers, and industry leaders for what’s effectively the AI event of the year in Asia.

What does this have to do with smart glasses? More than you might think. Behind every AI-powered wearable making headlines this year — from Google’s Android XR glasses to the wave of new AR hardware — is the infrastructure, edge silicon, and agentic AI technology that NVIDIA is building. GTC Taipei is where the backend of the smart glasses revolution gets real.

Jensen’s Keynote: The Big Stage

NVIDIA CEO Jensen Huang takes the stage at Taipei Music Center on June 1 at 11 AM local time. If the GTC 2026 main event in San Jose was any indication, expect major hardware announcements and a relentless focus on AI infrastructure scaling. The keynote will be livestreamed, and GTC Live pregame coverage kicks off beforehand with conversations on AI and accelerated computing’s latest breakthroughs.

For the smart eyewear crowd, the announcements to watch are anything touching edge inference, compact AI models, and real-time vision processing — the core ingredients that determine how capable (and how power-efficient) next-gen glasses will be.

Agentic AI Comes to Wearables

A defining theme of GTC Taipei is agentic AI — autonomous AI agents that can run tasks across local devices, cloud VMs, and edge hardware. NVIDIA’s “Build-a-Claw” program lets attendees learn how to build safer AI agents using OpenClaw and NVIDIA NemoClaw, deployable anywhere from a DGX Spark to a local laptop.

This matters for smart glasses because the entire category depends on agentic AI to function. When you ask Gemini on audio glasses to prepare a DoorDash order or summarize missed messages, you’re asking an AI agent to execute multi-step tasks in the background. The agent frameworks being showcased at GTC Taipei are the same ones that will power these experiences — handling task planning, tool use, and safe execution with or without a cloud connection.

Jensen himself demonstrated this at a Meet-a-Claw pre-event, showing how OpenClaw agents can handle software development, marketing, and content creation tasks. The message: practical AI is here, and it’s open source.

Jetson Thor and the Edge AI Opportunity

One of NVIDIA’s biggest wins at this year’s COMPUTEX Best Choice Awards is Jetson Thor — the company’s most powerful edge AI platform, designed for physical AI and autonomous machines. Powered by the Blackwell GPU architecture, Jetson Thor delivers up to 2,070 FP4 TFLOPS in a compact module running between 40 and 130 watts. That’s 7.5x the compute and 3.5x the energy efficiency of the previous generation.

Why should smart glasses watchers care? Because the same miniaturization and efficiency curve that NVIDIA is pushing with Jetson will eventually trickle down to wearable form factors. Real-time AI inference on a power budget — translating a conversation, identifying objects in your field of view, running generative edits on photos — requires the kind of silicon that Jetson Thor represents. The more efficient edge AI becomes, the thinner and lighter glasses can get without sacrificing capability.

Smart glasses today rely heavily on cloud offload for heavy AI lifting. The path to truly capable standalone eyewear runs through chips like these.

Vera Rubin: The Infrastructure Behind AI Glasses

NVIDIA’s Vera Rubin NVL72, which won the COMPUTEX Best Choice Gold Award and a Sustainability Special Award, is on the other end of the spectrum — a rack-scale AI supercomputer connecting 36 Vera CPUs with 72 Rubin GPUs. It delivers up to 10x the inference performance per watt and cuts per-token cost by 10x compared to previous generations.

This matters because every voice query, every photo edit, every “Hey Google” request from a smart glasses user ultimately hits cloud AI infrastructure. The cheaper and more efficient inference becomes at the data center level, the more capable (and affordable) the cloud services backing smart glasses become. Vera Rubin is designed specifically for agentic AI, reasoning, and long-context workloads — exactly the kinds of models that smart glasses will lean on.

Coupled with NVIDIA’s Spectrum-X silicon photonics and co-packaged optics switches, the networking infrastructure keeps latency low as these AI factories scale. Fast inference = responsive glasses.

Alpamayo: Perception AI That Glasses Need

NVIDIA Alpamayo — which won the Smart Vehicle and Smart Cabin category at COMPUTEX — is an open platform for autonomous vehicle development built around 10-billion-parameter chain-of-thought reasoning vision-language-action models. It’s aimed at autonomous driving, but the computer vision and real-time scene understanding technology is directly transferable to smart glasses.

The same challenges that Alpamayo tackles — interpreting ambiguous visual cues, understanding spatial context, recognizing objects and people in real time — are the fundamental problems every smart glasses platform needs to solve. Google’s Android XR glasses use Visual Positioning System for spatial awareness. The underlying AI models that make this possible share DNA with what NVIDIA is building for autonomous machines.

What to Watch at COMPUTEX 2026

GTC Taipei isn’t just about data center hardware. The conference covers:

• Physical AI and Robotics — developers integrating NVIDIA’s simulation frameworks and open models into next-gen factories, robots, and autonomous vehicles, with clear parallels to spatial computing and wearables.
• AI Factories and Infrastructure — how the full NVIDIA stack scales from single clusters to global AI factories, directly impacting the cost and latency of cloud-dependent devices like smart glasses.
• AI for Science — accelerated computing enabling breakthroughs in climate, materials, and biology, with implications for the sensor and materials science behind lighter AR optics.
• Agentic AI and Reasoning AI — building long-running autonomous agents using open frameworks, which is essentially the software architecture behind modern AI wearables.

NVIDIA also revealed plans for its Taiwan headquarters during the pre-show events — a signal of deepening commitment to the Taipei ecosystem that manufactures a significant portion of the world’s smart glasses components.

The Bottom Line for Smart Eyewear

While Google I/O 2026 stole the headlines with consumer smart glasses announcements, GTC Taipei is where the enabling technology lives. The edge AI processors, agent frameworks, simulation platforms, and AI infrastructure being showcased at COMPUTEX will determine how capable, how responsive, and how power-efficient the next generation of smart eyewear can be.

NVIDIA isn’t building glasses. But it’s building everything that makes glasses smart.

COMPUTEX 2026 runs June 1–4 in Taipei. Jensen Huang’s keynote is June 1 at 11 AM. Conference passes are sold out, but the keynote livestream is free.