Jensen Huang just plugged in the future of AI, and it’s running at full bandwidth. On the opening day of Computex 2025, Nvidia introduced NVLink Fusion, a next-gen chip interconnect platform designed to supercharge AI by letting custom silicon from companies like Marvell and MediaTek talk faster, think deeper, and scale wider.
The new NVLink Fusion isn’t just a speed upgrade. Nvidia declares that the AI race isn’t just about having the fastest GPU—it’s about orchestrating multiple processors to behave like one giant, distributed brain. It’s tech infrastructure as strategy, and Nvidia wants to be the socket in everyone’s server.
What It Means:
Instead of limiting its chip-to-chip magic to its own Grace and Blackwell processors, Nvidia is licensing NVLink Fusion to outsiders. If you’re a chip designer with big AI dreams, you now have a shortcut to Nvidia-grade interconnects.
Why It Matters:
From ChatGPT-scale LLMs to massive AI training clusters, performance hinges on bandwidth between accelerators. Nvidia knows this, and it’s putting its hooks into the future of AI infrastructure before anyone else can.
Oh, and there’s more:
- Desktop AI rigs for researchers? In production now. Called DGX Spark, these are plug-and-play AI powerhouses made for labs, startups, and probably billion-dollar unicorns still in stealth. Huang says they’ll ship in “weeks.”
- New chips, endless roadmap: The Blackwell Ultra will drop later this year. Then comes Rubin, followed by Feynman in 2028. Nvidia’s not just iterating—it’s running a marathon in light-speed shoes.
- Taiwan’s roots run deep: Huang also revealed plans to build a new Nvidia headquarters in Taipei, signaling how central Taiwan is to Nvidia’s supply chain and strategy.
Nvidia’s NVLink Fusion is a power move, not just a product. It turns Nvidia into the underlying fabric of AI compute, even when others build the silicon. First, they sold the GPUs. Now they’re selling the glue. And if you’re a competitor, you’re going to need a lot more than duct tape.
AI for the City of Tomorrow — Linker Vision and Nvidia Deploy Smart Urban Brains
If you think smart cities are just buzzwords and traffic sensors, Linker Vision is here, with AI agents, digital twins, and synthetic data pipelines that are already changing how cities think, watch, and respond.
Unveiled at Computex 2025, Linker Vision’s AI platform leverages the full might of Nvidia’s stack, combining Omniverse, Cosmos world models, and Metropolis to deploy AI that does everything from summarizing security camera footage to orchestrating cross-agency responses in real time.
What’s Under the Hood:
- Digital twins created in Omniverse simulate city events before they happen—from car crashes to typhoon evacuations.
- Synthetic data is generated via Cosmos WFMs to train AI agents without needing real-world disasters.
- Vision Language Models (VLMs) interpret camera footage like detectives, spotting anomalies, flagging threats, and prioritizing response—all without human intervention.
Real Impact, Real Deployment:
Partnering with Kaohsiung City and Chunghwa Telecom, Linker has deployed this tech to power AI-driven emergency response, public safety, and municipal coordination. The system breaks data silos across departments—enabling one integrated, intelligent control room for the entire city.
CEO Paul Shieh put it clearly: “We’re not just watching—we’re understanding. This is AI for digital governance, not just digital gimmicks.”
Why It Matters:
By building an end-to-end loop—simulate, train, deploy—Linker Vision isn’t just installing smart cameras. It’s giving cities an AI nervous system. With Nvidia as the underlying AI compute layer, this is real-time analytics with global scalability.
The Future is Agentic:
These aren’t just scripts. They’re AI agents capable of adjusting responses based on context, like spotting a traffic jam and rerouting responders without waiting for a 911 call. As the world urbanizes, this level of autonomy will define public safety, logistics, and governance.
Wrap-up:
Between Nvidia’s NVLink Fusion and Linker Vision’s AI cities, Computex 2025 isn’t just a product showcase—it’s a preview of AI at planetary scale. From rack servers to street corners, the GPU empire is spreading fast. And if you’re not riding the wave, you might end up stuck in traffic, with an AI camera watching.
