Leading AI equities have declined in response to DeepSeek’s new AI model.
According to Pat Gelsinger, the former CEO of Intel, the markets were mistaken on Monday.
Gelsinger stated that rather than lowering demand for processors, DeepSeek boasts that producing AI at a low cost would increase it.
Former Intel CEO Pat Gelsinger claimed that “the markets are getting it wrong” when investors panicked out of the company’s leading AI equities over worries that DeepSeek would reduce demand for cutting-edge CPUs.
When a new AI model from the Chinese startup DeepSeek claimed to have replicated the success of a leading model released by OpenAI just months ago — and at a fraction of the cost — top AI companies, like Nvidia, saw hundreds of billions of dollars erased from their valuations on Monday.
Compared to the best AI laboratories in America, DeepSeek’s latest model, R1, which was unveiled on President Donald Trump’s Inauguration Day, seemed to have achieved the feat with fewer and less potent processors. The move has sparked worries that demand for chip stocks, such as Nvidia, may decline.
After adding trillions of dollars to its market capitalization since the ChatGPT bubble began, Nvidia saw a $500 billion wipeout on Monday, which led to the largest stock market meltdown in US history.
In a post on X on Monday, Gelsinger implied that the market’s presumptions were incorrect. Instead of decreasing demand, he claimed that making computing “dramatically cheaper” and more user-friendly, as DeepSeek seems to have done, “will expand the market for it.”
The former head of Intel, who retired from the business in December after failing to take advantage of the AI boom, also implied that DeepSeek’s Chinese engineers “had limited resources, and they had to find creative solutions” to get their models to perform better.
According to the AI industry, models are trained and hosted in data centers, where they are powered by increased processing power in the form of chips.
The greatest chips from America have been difficult for Chinese enterprises to obtain due to strict export prohibitions. The amount of processing power that powers DeepSeek’s R1 model, which was disclosed last week, has not been made clear.
