At Computex 2025, Nvidia unveiled a multi-front technique to broaden its function in international AI infrastructure. Going past the everyday efficiency benchmarks or chip shipments bulletins, Nvidia outlined a broader imaginative and prescient. One that’s centered on enabling customized methods, fostering worldwide alliances, and embedding itself deeper into the structure of contemporary compute.
The chip producer is already a key participant within the {hardware} area, it’s now positioning itself as a worldwide infrastructure enabler, shaping not simply how AI runs, however the place it’s constructed and the way it operates. That shift was underscored by the launch of NVLink Fusion, a brand new interconnect designed to offer cloud suppliers and chipmakers extra flexibility in how they construct AI methods round Nvidia’s platforms.
With NVLink Fusion, Nvidia has opened its closed ecosystem to permit customized CPUs and AI accelerators from different firms to attach on to Nvidia GPUs. MediaTek, Marvell, Alchip Applied sciences, Astera Labs, Synopsys, and Cadence are among the many first to undertake NVLink Fusion.
Shutterstock
Along with {hardware} developments, Nvidia is introducing Mission Management, a software program platform designed to handle AI information heart operations effectively. Along with the high-speed connectivity supplied by NVLink Fusion, these developments are usually not solely about bettering efficiency but additionally about enabling hyperscalers and governments to develop AI factories with minimized reliance on U.S. cloud suppliers.
These developments sign Nvidia’s strategic transition from proprietary full-stack options to a extra open, ecosystem-driven AI infrastructure mannequin.
“A tectonic shift is underway: for the primary time in a long time, information facilities should be essentially rearchitected — AI is being fused into each computing platform,” stated Jensen Huang, founder and CEO of NVIDIA. “NVLink Fusion opens NVIDIA’s AI platform and wealthy ecosystem for companions to construct specialised AI infrastructures. This unbelievable physique of labor now turns into versatile and open for anyone to combine into.”
The brand new interconnect powers Nvidia’s GB200-based rack-scale methods. The corporate claims that this delivers as much as 1.8 TB/s per GPU — 14 occasions sooner than PCIe Gen5, which is the present industry-standard interface. That is the sort of tightly built-in compute infrastructure wanted to coach next-generation AI fashions and energy full-scale AI factories.
Computex
Nvidia just isn’t solely broadening the scope of who can construct with its know-how, however it is usually making some calculated bets on the place its infrastructure ought to take root. The corporate is planning to make vital investments in Taiwan, which is residence to Nvidia’s key manufacturing companion TSMC, and a rising hub of AI and semiconductor innovation.
Huang revealed that NVIDIA is strengthening its partnership with Foxconn Hon Hai Know-how Group and collaborating with the Taiwan authorities to develop a robust AI manufacturing facility supercomputer. This initiative will present cutting-edge NVIDIA Blackwell know-how to researchers, startups, and key industries, together with TSMC. “Having a world-class AI infrastructure right here in Taiwan is absolutely vital,” Huang stated.
Moreover, Nvidia plans to ascertain a brand new native headquarters named "Nvidia Constellation" in Taipei's Beitou Shilin Science Park, additional embedding itself in Taiwan's tech ecosystem. Throughout his keynote handle at Computex, Huang launched a dramatic video exhibiting NVIDIA’s Santa Clara workplace launching into area and touchdown in Taiwan.
The information of bold growth plans in Taiwan comes at a time when Nvidia is below strain from the U.S. authorities to restrict superior chip exports to China. Some {industry} consultants view the Taiwan investments as strategic alignment to mitigate potential commerce restrictions and keep operational resilience.
At Computex, Haung praised the AI developments made by DeepSeek. His remarks function an acknowledgment, extra like validation, that AI innovation is turning into extra decentralized, and that non-Western AI initiatives and initiatives are rising in sophistication. On condition that U.S. lawmakers just lately accused DeepSeek of information harvesting, AI theft, and espionage, Huang’s recognition is notable. It suggests Nvidia is taking a extra pragmatic strategy that prioritizes technological collaboration over geopolitical tensions.
Shutterstock
Whereas Nvidia is forging worldwide collaborations to advance AI infrastructure, it's additionally specializing in empowering particular person builders and researchers. The corporate launched DXG Sparx — a compact AI workstation designed to ship information center-level efficiency on to the desktop.
The DGX Spark is constructed across the GB10 Grace Blackwell Superchip, which Nvidia says delivers as much as 1,000 AI TOPS and comes with 128GB of unified reminiscence. This setup lets builders work with giant fashions immediately on their very own machines as a substitute of relying on cloud-based sources.
With its compact design and energy-efficient efficiency, the DGX Spark is well-suited for researchers and builders working with restricted sources. Nvidia says it will possibly course of fashions with as much as 200 billion parameters, enabling refined AI duties like fine-tuning and inference with out counting on exterior computing energy.
“AI has revolutionized each layer of the computing stack — from silicon to software program,” stated Huang. “Direct descendants of the DGX-1 system that ignited the AI revolution, DGX Spark and DGX Station are created from the bottom as much as energy the following technology of AI analysis and growth.”
The bulletins at Computex 2025 replicate Nvidia's strategic shift from a GPU-centric firm to a worldwide AI infrastructure chief. By embracing open requirements, forging worldwide partnerships, and investing in regional AI ecosystems, Nvidia is solidifying its function on the coronary heart of the world's digital transformation.