Nvidia CEO Jensen Huang refutes claims of AI chip demand slowdown, asserting that compute demand will grow exponentially.

Global chip leader Nvidia (NVDA.US) is accelerating its expansion with the help of artificial intelligence (AI), and its CEO Jensen Huang has made a strong defense against the market's concerns that AI models will reduce computing power needs at the GTC Global Technology Conference in San Jose, California. The chip industry magnate clearly stated that the demand for computing infrastructure for new AI models not only will not decrease but will grow exponentially.

Huang responded to the controversy over the R1AI model launched by Chinese startup DeepSeek, which some industry insiders feared might reduce reliance on high-end chips and servers. Huang said bluntly: "The market's understanding of R1 is fundamentally wrong. The actual computing power needed is far beyond current expectations." He emphasized that the massive amount of data needed to generate more complex answers by new AI models will inevitably push computing power demand to new heights.

On the capital market, Nvidia's stock price has fallen 14% this year, but it rose 2.4% during the GTC conference, showing investors' confidence in the AI narrative. Huang's key data support his view: the orders for the latest Blackwell architecture chips are far more than those for the previous generation Hopper chips in the same period of their life cycle, and these orders only come from cloud service providers, not including other industry customers' procurement needs.

Moreover, Huang showed his calmness in the face of challenges from tech giants such as Google and Broadcom (AVGO.US) developing dedicated AI chips (ASICs). He pointed out that most ASIC chips eventually fail to be truly deployed because enterprise customers are more concerned about improving business revenue through top-notch computing power than simply cutting costs. "These CEOs are all masters of arithmetic," he said meaningfully, "the revenue growth brought by computing power upgrades is on a completely different scale from the savings in hardware costs."

Looking to the future, Huang showed his optimistic expectations for the AI industry. Even if the US economy is in a recession, he believes enterprises will increase their AI investments and concentrate limited resources on the most promising areas.

In terms of trade policies, Nvidia has already made early preparations to cope with tariff risks by localizing its core chips through TSMC's (TSM.US) Arizona factory, ensuring supply chain security.

Behind this battle for computing power demand is Nvidia's accurate grasp of the speed of AI technology iteration. When the industry is still discussing whether Moore's Law is invalid, Huang's data showing that the performance of the Blackwell architecture is 40 times that of the previous generation announced that the growth rate of computing power demand is redefining the development rules of the semiconductor industry. For investors, the real test may be who can keep up with the computing power arms race led by Nvidia.