In Taiwan, celebrity spotting involves finding Jensen Huang, the CEO and co-founder of NVIDIA, the company at the heart of the generative artificial intelligence revolution.
He arrived in Taipei much before COMPUTEX 2025 (one of the largest computer and technology trade shows in the world) kicked off. In his almost two-hour keynote speech, the man covered a range of topics, including one that everyone is talking about — where is AI headed, and what is the function of NVIDIA.
Sporting his trademark black leather jacket, Huang said NVIDIA is “not a technology company only anymore”. “We are an essential infrastructure company and how can you plan your infrastructure, your land, your power, your electricity, all of the necessary financing around all over the world, how would you possibly do that if you didn’t understand what I was going to make. So we describe our company’s roadmap in fair detail.”
Jensen wants his company's new overseas headquarters to be built in Taipei's Beitou Shilin area The Telegraph
The man said no company talks about five-year roadmaps. NVIDIA does. When he started the company, he was trying to figure out how big the opportunity was in 1993. “I came to the conclusion NVIDIA's business opportunity was enormous…. Three hundred million dollar chip industry to a data centre opportunity that represents about a trillion dollars, to now an AI factory and AI infrastructure industry that will be measured in trillions of dollars, and this is the exciting future that we're undertaking.”
He described his company as more of an “AI infrastructure company that’s essential all around the world” as the new “infrastructure is an infrastructure of intelligence”. “Right now, when we say there’s an intelligence infrastructure, it makes no sense. But I promise you, in 10 years you will look back and you will realise that AI has integrated into everything. AI is now part of infrastructure, and this infrastructure, just like the Internet, just like electricity, needs factories. These factories are essentially what we build today. They are not data centres of the past.”
The “AI factories” he spoke about apply energy to produce something “valuable” called tokens. “Companies are starting to talk about how many tokens they produced last quarter. And how many tokens they produced last month. Very soon, we will be talking about how many tokens we produce every hour, just as every single factory does. The world has fundamentally changed,” he said.
A 30-degree Celsius Taipei didn’t bother the 62-year-old CEO. He wants Taiwan to play a very important role in the AI revolution. The world’s leading supplier of AI chips intends to acquire a larger office in Taiwan and work with technology firms to develop a “supercomputer”.
Huang said the Taipei mayor still wanted to determine whether residents support the new NVIDIA office, which images indicate will be a new building, dubbed ‘NVIDIA Constellation'. It will be in Beitou Shilin. The mayor needs to determine whether residents would support the new office. Silicon Valley to Bangalore, whenever important tech companies arrive, living costs and house rent shoot up. It remains to be seen if that would be the case here.
Opens up its platform
As the rush to build data centres and AI models continues, Huang announced a new NVLink Fusion system that allows the building of more tailored artificial intelligence infrastructure that can combine the company’s high-speed links with semiconductors from other providers.
So far, the company has offered computer systems built with its own components. The new solution will allow data centre customers more flexibility but still retaining NVIDIA technology at the centre.
An office of NVIDIA in Taipei. Mathures Paul
The CEO of the California-based company said: “Maybe what you would like is to use your own CPU. You've been building your own CPU for some time and maybe your CPU has built a very large ecosystem and you would like to integrate NVIDIA into your ecosystem. Now we make it possible for you to do that. You could do that by building your custom CPU. We provide you with our NVLink chip-to-chip interface for your ASIC. We connect it with NVLink chiplets. And now it connects and directly abuts into the Blackwell chips and our next-generation Rubin chips. And again, it fits right into this ecosystem.”
The company said that AI chipmaking partners for NVLink Fusion already include MediaTek, Marvell, Alchip, Astera Labs, Synopsys and Cadence.
The company is in a dominant position in GPUs used for general AI training. But competition is close at hand. NVIDIA’s largest competitors in AI computing include Cloud providers such as Google, Microsoft and Amazon, all of which are building their own custom processors.
Let’s talk about robots
Machines need to understand and interact with the laws of the real world. Huang spoke about Isaac Groot N1, a foundation model for robotics that offers humanoid robots the ability to reason and plan, besides perceiving their surroundings.
“We would like to (also) build physical robots and these physical robots first it start with the ability to learn to be a robot. The ability to learn to be a robot can't be done in the physical world productively. You have to create a virtual world where the robot can learn how to be a good robot. That virtual world has to obey the laws of physics,” said Huang.
The company announced NVIDIA Isaac GR00T N1.5, the first update to its “open, generalised, fully customisable foundation model for humanoid reasoning and skills”; NVIDIA Isaac GR00T-Dreams, a blueprint for generating synthetic motion data; and NVIDIA Blackwell systems to accelerate humanoid robot development.
Humanoid and robotics developers Agility Robotics, Boston Dynamics, Fourier, Foxlink, Galbot, Mentee Robotics, NEURA Robotics, General Robotics, Skild AI and XPENG Robotics are adopting NVIDIA Isaac platform technologies to advance humanoid robot development and deployment.
Developers can first post-train Cosmos Predict world foundation models (WFMs) for their robot. Then, using a single image as the input, GR00T-Dreams generates videos of the robot performing new tasks in new environments. The blueprint then extracts action tokens — compressed, digestible pieces of data — that are used to teach robots how to perform these new tasks.
The man-in-jacket said “the NVIDIA story is the reinvention of the computer industry”. “ We started out as a chip company with a goal of creating a new computing platform. And in 2006, we introduced CUDA (Compute Unified Device Architecture), which has revolutionised how computing is done. In 2016, we realised that a new computing approach had arrived. It required a reinvention of every single layer of the technology stack. The processor was new. The software stack was new. So we invented a new system. That system was called DGX-1. I donated the first one to a non-profit company called OpenAI. And it started the AI revolution.”
Soon, he said, the company realised that this new way of doing software, which is now called artificial intelligence, is unlike traditional ways of running software. “Whereas many applications ran on a few processors in a large data centre — we call it hyper-scaling — this new type of application requires many processors working together, serving queries familiar to people. That data centre will be architected fundamentally differently. We realised there were two types of networks. One, for north-south because you still have to control storage, you still have to have a control point, you still have to connect to the outside. But the most important network was going to be east-west. The computers are talking to each other to try to solve a problem. We recognised the best networking company in east-west traffic for high-performance computing, large-scale distributed processing… a company that was very dear to our company and very close to our heart… a company called Mellanox and we bought them in 2019.”
(The reporter is in Taipei on an invitation from BenQ)