According to a report on March 20 from the Financial Associated Press (Editor Zhao Hao), NVIDIA CEO Jensen Huang recently stated that the R1 model released by the Chinese AI company DeepSeek will only increase the demand for computing infrastructure, thus concerns about "chip demand possibly decreasing" are unfounded.

On Wednesday (March 19) local time, Jensen Huang said during a meeting with Analysts and investors at the GTC conference that the previous understanding of "R1 possibly reducing chip demand" is completely incorrect, and that future computing demands may actually be much higher.

In January of this year, the R1 model released by DeepSeek caused a sensation in the market; the model took only two months to develop, cost less than 6 million dollars, and used about 2,000 NVIDIA chips, yet the R1's performance in key areas rivals that of OpenAI's strongest reasoning model, O1.

This led many to question the significance of large American technology companies investing heavily in AI models and datacenters. NVIDIA's stock price plummeted nearly 17% on January 27, and last week its shares continued to fall to the lowest level since September of last year.

However, several major clients of NVIDIA have reiterated their spending plans for AI in recent earnings reports. Media analysis has revealed that the spending of the largest datacenter operators is actually growing faster than expected.

During the analyst meeting, it was mentioned that some of NVIDIA's clients are developing self-researched chips to try to replace NVIDIA's products, such as Google, which is collaborating with Broadcom to develop ASIC chips. In response, Jensen Huang argued that many ASICs cannot be deployed in datacenters.

Jensen Huang believes that these major clients need better chips to derive more revenue from their infrastructure rather than using cheap alternatives to save costs. He added that CEOs are very good at math, so they do not simply focus on costs.

Jensen Huang pointed out that the chip products from competitors cannot compare with NVIDIA's Hopper design, and currently, the performance of Blackwell is even 40 times more powerful than Hopper.

Jensen Huang also presented some numbers indicating that the company has received far more Blackwell Orders during the same period than Hopper. He mentioned that this is just from cloud providers, and spending by Other companies on AI Datacenters will further enhance these figures.

Jensen Huang mentioned that if the USA economy falls into recession, corporate clients may shift more investments towards AI devices—because that is where their Business growth lies.

He also stated that the proposed tariff policy by the Trump administration would have very little short-term impact on NVIDIA, as the company will continue to relocate the manufacturing of the most important parts of its product line to the USA.

Jensen Huang stated that NVIDIA is already utilizing Taiwan Semiconductor's factory capacity in Arizona, and as its suppliers increase capacity, NVIDIA will increase usage.

Editor/Jeffy