Nvidia has positioned itself as an AI infrastructure company, with CEO Jensen Huang declaring that data centers are now “AI factories” producing valuable tokens.
NVLink Fusion delivers up to 1.8 TB/s per GPU—14x faster than PCIe Gen5—enabling integration of non-Nvidia chips with Nvidia’s GPUs.
Jensen Huang believes in the future, the jobs gap will be filled by AI agents and humanoid robots instead of human beings.
At Computex 2025 in Taipei this week, Nvidia CEO Jensen Huang presented a more aggressive vision of his company’s future, emphasizing artificial intelligence’s need for dedicated infrastructure—and Nvidia’s commitment to give the world what it needs.
“AI is now infrastructure, and this infrastructure—just like the internet, just like electricity—needs factories,” Huang told the audience. “You apply energy in, and it produces something incredibly valuable: tokens.”
The company introduced the idea of these data centers as new “AI factories,” reimagining their purpose as facilities specifically optimized for artificial intelligence workloads.
As a first step toward significant expansion, Nvidia announced NVLink Fusion, opening its “interconnect” technology to other manufacturers. This system enables customers to integrate processors from other companies alongside Nvidia’s GPUs.
“A tectonic shift is underway: for the first time in decades, data centers must be fundamentally rearchitected—AI is being fused into every computing platform,” Huang said in an official press release. “NVLink Fusion opens Nvidia’s AI platform and rich ecosystem for partners to build specialized AI infrastructures.”
This development is particularly significant as it provides developers with more hardware choices, even from third-party providers, for training their models. Companies can now build semi-custom AI systems tailored to specific workloads, optimizing for cost, performance, or specific architectural advantages.
Partners including MediaTek, Marvell, Alchip Technologies, Astera Labs, Synopsys, and Cadence are among the first to adopt this technology.
Another important development, announced separately from Computex, is that Nvidia expanded its partnership with Neocloud providers. Such providers focus almost exclusively on renting high-end GPUs, such as NVIDIA’s H100, H200, A100, and upcoming Blackwell chips, for AI-specific workloads like training large language models (LLMs), computer vision, and inference. Some of these new “cloud partners” include CoreWeave, Nebius, Crusoe, and Lambda.
As part of its vision of establishing “sovereign AI” development, the company announced partnerships with Saudi Arabia and the United Arab Emirates to develop massive infrastructure for AI training and deployment, mainly driven as part of Trump’s chip accords.
The market has responded positively to these announcements, with Nvidia’s stock growing almost 60% since its dip on April 7. On Monday, though, the stock was relatively stable—as it has been for the last 4 days—going from $132.39 to $135.35. NVDA finished Tuesday’s trading day just below the $135 mark.
Image: Tradingview
Nvidia also emphasized that NVLink enables compatibility between Qualcomm processors and Nvidia GPUs. This could lead to AI computers featuring energy-efficient CPUs paired with Nvidia’s high-performance GPUs, or even mobile hardware with Nvidia GPUs for faster inference and enhanced gaming experiences.
The company also revealed plans for a new local headquarters called Nvidia Constellation in Taipei’s Beitou Shilin Science Park. Nvidia is also expanding to Taiwan, where the company plans to expand its quantum operations.
For individual developers, Nvidia introduced DGX Spark, a compact AI workstation built by partners including ASUS, Dell, and MSI. Powered by the GB10 Grace Blackwell Superchip, it offers 1,000 AI TOPS with 128GB memory, supporting models up to 200 billion parameters.
Just for context, Nvidia’s most powerful consumer GPU, the RTX 5090 packs up to 32GB of memory, and starts at $2,499.
“Direct descendants of the DGX-1 system that ignited the AI revolution, DGX Spark and DGX Station are created from the ground up to power the next generation of AI research and development,” Huang noted.
Reservations for the DGX Spark start at $3,999 per unit.
Image: Nvidia
Beauty in the dark side of AI?
Addressing concerns about AI’s impact on employment, Huang presented a different perspective. Rather than causing a job crisis, he sees AI as the solution to an impending workforce shortage.
“We’ll have a shortage of workers by 2030—about 30 to 50 million people short. It’s actually limiting the world’s ability to grow,” Huang said. “Now we have these digital agents that can work with us. One-hundred percent of Nvidia software engineers now have digital agents working with them to help develop better code more productively.”
Looking further into the future: “The reason why humanoid robotics are so important is because it is the only form of robot that can be deployed almost anywhere,” he said, adding that “this whole new industry is going to expose us to giant opportunities ahead.”
Generally Intelligent Newsletter
A weekly AI journey narrated by Gen, a generative AI model.
