Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
(Bloomberg) – Nvidia Corp., whose products have fueled a flood of artificial intelligence spending, said new types of AI models that create more complex answers, only the need for a computer infrastructure.
Most read from Bloomberg
Jensen Huang, Chief Executive Officer, said that concerns caused by the R1 AI model of the Chinese startup from Deekseek – that fewer chips and powerful servers would be needed for such software in the future – are out of place.
“The understanding of R1 was completely wrong,” said Huang at a meeting with analysts and investors at the GTC conference of his company on Wednesday in San Jose, California. “The need for calculation is much higher.”
Nvidia’s shares rose by 2.4%in New York. They had declined by 14% this year.
The company tries to convince a broader group of industries to invest in AI devices and promises that the economic advantages of technology are available. Nvidia, which has become the most valuable chip maker in the world in the past two years after dizzying growth in the past two years, is with more investor test in 2025.
Deepseek stimulated these fears at the beginning of this year when it released its AI model and said it generated the powerful technology for less money. But since then, the largest customers of Nvidia have confirmed their expenditure plans. The analysis of Bloomberg Intelligence this week showed that the expenditure of the largest data center operators is actually increasing faster than expected.
At the analyst meeting, Huang from Nvidia was asked to develop their own components after the efforts of the customers – chips that could replace his AI accelerators in data centers. Companies like Google from Alphabet Inc. have worked with Broadcom Inc. to develop their own application -specific integrated circuits or ASICs for this area. But Huang countered that many ASICs were designed, but are not always used in data centers.
These large customers need better chips to achieve more income from their infrastructure, not cheaper to save costs, he argued.
“All of these companies are operated by great CEOs that are really good in math,” he said. “The effects are not just costs. It is another calculation.”
The competitors’ chips cannot keep up with Nvidia’s Hopper design, his previous generation, he said. And the current Blackwell platform is 40 times more powerful.
