Brief
- Nvidia has positioned himself as an AI infrastructure company, CEO Jensen Huang declaring that data centers are now “AI factories” producing precious tokens.
- NVLink Fusion offers up to 1.8 TB / S per GPU – 14x faster than PCIe Gen5 – allowing the integration of non -Nvidia chips with the GPUs of Nvidia.
- Jensen Huang believes in future, the gap of jobs will be filled by AI agents and humanoid robots instead of human beings.
In Calpex 2025 in Taipei this week, the CEO of Nvidia, Jensen Huang, presented a more aggressive vision of the future of his business, emphasizing artificial intelligence for the dedicated infrastructure – and Nvidia’s commitment to give the world what he needs.
“AI is now an infrastructure, and this infrastructure – just like the Internet, just like electricity – factories,” Huang told the public. “You apply energy, and it produces something incredibly precious: tokens.”
The company has introduced the idea of these data centers as new “AI factories”, reinventing their objective as a specifically optimized installment for workloads of artificial intelligence.
As a first step towards significant expansion, Nvidia announced Nvlink Fusion, opening its “interconnect” technology to other manufacturers. This system allows customers to integrate processors from other companies alongside the NVIDIA GPUs.
“A tectonic change is underway: for the first time in decades, the data centers must be fundamentally re -wrote – AI is merged in all IT platforms,” said Huang in an official press release. “NVLink Fusion opens the NVIDIA AI platform and the rich ecosystem for partners to build specialized AI infrastructure.”
This development is particularly important because it provides developers with more choice of equipment, even with third -party suppliers, for the formation of their models. Companies can now create semi-personalized AI systems adapted to specific workloads, cost optimization, performance or specific architectural advantages.
Partners such as 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 has expanded its partnership with the suppliers of Neocloud. These suppliers are concentrated almost exclusively on high -end GPU rental, such as the H100, H200, A100 chips from NVIDIA, and the next Blackwell chips, for specific workloads such as the formation of large language models (LLM), computer vision and inference. Some of these new “Cloud partners” include Coreweave, Nebius, Crusoe and Lambda.
As part of its vision of the establishment of the development of “Sovereign IA”, the company announced partnerships with Saudi Arabia and the United Arab Emirates to develop massive infrastructure for the training and deployment of AI, mainly motivated in the context of Trump flea agreements.
The market responded positively to these announcements, Nvidia shares that increased by almost 60% since its decline on April 7. On Monday, however, the action was relatively stable – as it has been the case for 4 days – from $ 132.39 to $ 135.35. NVDA ended the negotiation day on Tuesday below the $ 135 mark.
NVIDIA also stressed that NVLink allows compatibility between Qualcomm processors and NVIDIA GPUs. This could lead to AI computers with energy -efficient CPUs associated with NVIDIA high performance GPUs, or even mobile equipment with NVIDIA GPUs for faster inference and improved game experiences.
The company also revealed plans for a new local headquarters called Nvidia Constellation in the Beitou Shilin scientific park in Taipei. Nvidia is also developing in Taiwan, where the company plans to extend its quantum operations.
For individual developers, NVIDIA introduced DGX Spark, a compact AI work station built by partners such as Asus, Dell and MSI. Powered by the GB10 Grace Blackwell Superchip, it offers 1,000 IA tops with 128 GB of memory, supporting models up to 200 billion parameters.
Just for the context, the most powerful consumption GPU in Nvidia, the RTX 5090 contains up to 32 GB of memory and starts at $ 2,499.
“The direct descendants of the DGX-1 system who sparked the AI revolution, the DGX Spark and DGX station are created from zero to the power of the next generation of research and development of the AI,” noted Huang.
Reservations for the DGX Spark start at $ 3,999 per unit.
Beauty in the dark side of AI?
Responding to the concerns about the impact of the AI on employment, Huang presented a different perspective. Rather than provoking a employment crisis, he considers AI as the solution to a shortage of imminent labor.
“We will have a shortage of workers by 2030 – about 30 to 50 million people short. This is in fact limited the capacity of the world to develop,” said Huang. “Now we have these digital agents who can work with us. A hundred percent of NVIDIA software engineers now have digital agents working with them to help develop better code in a more productive way.”
Looking further in the future: “The reason why humanoid robotics is so important is that it is the only form of robot that can be deployed almost everywhere,” he said, adding that “this brand new industry will expose us to giant opportunities to come.”
Generally intelligent Bulletin
A weekly IA journey told by Gen, an AI generator model.
