Nvidia CEO Jensen Huang speaks at a press event Tuesday in Taipei. CNA photo June 4, 2024
Taipei, June 4 (CNA) Nvidia CEO Jensen Huang said that his company is making investments in Taiwan because Taiwan Semiconductor Manufacturing Co. (TSMC) is beyond “normal good,” at a press event Tuesday in response to a question about geopolitical risks.
The tech giant held a global press conference in Taipei on the first day of Computex, one of the world’s largest computer and technology trade fairs.
Asked whether he has geopolitical concerns when making a series of investment plans in Taiwan such as setting up Nvidia’s first research and development center in Asia and a Taiwan headquarters, Huang responded by saying, “we build in Taiwan because TSMC is so good; it is not normal good.”
TSMC has advanced technology, an incredible work ethic and super flexibility, he said, adding that the two companies have worked together for over a quarter of a century and “know each other’s rhythm.”
“We can build very complicated things at very high volume at very high speed with TSMC,” Huang said. “This is not normal, and you can’t just randomly ask somebody else to do it.”
The CEO also called Taiwan’s industry ecosystem “incredible.”
From upstream of TSMC to downstream of TSMC, from chip making, packaging, testing to assembly and logistics, the ecosystem is rich with “amazing companies doing amazing things,” he said.
“If we can build it somewhere else, we’d think about it, but quite frankly, [it’s] very difficult,” Huang said. “So yes, there are many considerations in life, but the most important consideration is possibility.”
‘Buy more, save more’
At the press conference, Huang, who has been dubbed “the godfather of AI,” also shared his idea about sustainability concerns in the coming artificial intelligence (AI) era.
Accelerated computing should be how computing is done because it actually saves power, he argued, repeating the logic of “buy more [GPU], save more” that he championed in his Sunday keynote speech in Taipei.
The graphics processing unit (GPU), initially designed to accelerate computer graphics and image processing, is what Nvidia specializes in. It has become one of the most important types of computing technology in the AI era because of its parallel processing power.
The Nvidia CEO believes more-GPU-supported accelerated computing could make computing more efficient and thus energy-saving.
He emphasized that since generative AI is not about training, but inference, “when you inference, the amount of energy used versus alternative way of doing computing is actually much smaller.”
He cited as an example the climate simulation technology that Taiwan’s Central Weather Administration has adopted and was mentioned during his Sunday speech.
The new generative AI model for weather forecasting is said to be able to have 2-kilometer resolution with 1,000 times the speed and 3,000 times the energy efficiency of conventional weather models.
“3,000 times less power, not 30 percent,” Huang stressed, pointing out that the potential for energy savings is enormous if AI is used more for inference.
The Nvidia founder added that “AI doesn’t care where it goes to school.”
He was referring to the fact that AI consumes the most power when it is being trained, but that training can be moved to “where people don’t want to live” so it does not compete with the public for power consumption.
He argued that the world has a lot of excess energy from the sun and the power plants, so data centers for AI training can be set up “where there isn’t a population.”
