(Bloomberg) – NVIDIA CORP., whose products have supported the flood of artificial intelligence expenditure, said that new types of AI models that create more complex answers will only increase the need for computational infrastructure.
Most reading from Bloomberg
Jensen Huang, CEO, said that fewer chips and powerful servers would be needed for such software in the future – was needed for this software.
“Understanding R1 was completely bad,” Huang said at a meeting with analysts and investors at the GTC conference on Wednesday in San Jose, California. “The demand for calculation is much higher.”
NVIDIA shares rose by up to 2.4%in New York. This year they headed for a session.
The company seeks to persuade a wider group of industries to invest in AI facilities, which promises to be economical benefits of technology at hand. NVIDIA, which has become the most valuable chip in the world after dizzying growth in the last two years, has been facing more investors in 2025. The key problem is whether customers will maintain their expenses for AI infrastructure.
Deepseek raised these concerns at the beginning of this year when he released his model AI and said he had created powerful technology for less money. However, NVIDIA's largest customers have since reaffirmed their expenditure plans. And this week, Bloomberg Intelligence has found that the expenditure of the largest data centers are actually rising faster than expected.
At the analyst meeting, Huang NVIDIA was asked for customers' efforts to develop their own components – chips that could replace its AI accelerators in data centers. Company as ALPHABET INC. Google cooperate with Broadcom Inc. on the development of your own integrated circuits or asics specific to application. Huang, however, faced that many ASICs were designed, but not always put in data centers.
These large customers need better chips to generate more income from their infrastructure, not cheaper to save costs, claimed.
“All these companies are managed by great CEOs who are really good in mathematics,” he said. “The effects are not just costs.
Competitors chips cannot match the design of NVIDIA HOPPER, its previous generation, he said. And the current Blackwell platform is 40 times more powerful.
