DeepSeek's low-cost AI model triggers tech stock selloff, sparks 'Sputnik moment' debate

On Monday morning, technology stocks across the globe entered a sharp decline that would mark the sharpest single-day drop in years for major indices. The Nasdaq 100 futures fell nearly 4 percent, suggesting the index was headed for its worst day since September 2022. The S&P 500 futures dropped 2 percent. Nvidia, the chipmaker that has become synonymous with the artificial intelligence boom, fell 10 percent in pre-market trading. Oracle lost 8 percent. Palantir, which specializes in AI data analytics, shed 7 percent. From Tokyo to Amsterdam, investors were selling.

The catalyst was a Chinese startup most Western investors had barely heard of. DeepSeek, a small company based in Hangzhou, had released an AI assistant the previous week that claimed to achieve what the industry had begun to assume was impossible: building a competitive large language model at a fraction of the cost everyone else was spending. The company said it had trained its DeepSeek-V3 model using less than $6 million, a figure that stood in stark contrast to the $100 million or more that Western companies like OpenAI and others had been pouring into similar projects. By Monday, DeepSeek had climbed past ChatGPT in Apple's app store download rankings.

The numbers behind DeepSeek's claim were striking enough to unsettle the market's confidence in the entire AI investment thesis. The startup's researchers had published a paper detailing how they achieved this efficiency using Nvidia's H800 chips—not the company's most advanced processors, but rather a reduced-capability model originally designed to circumvent U.S. export restrictions to China before being banned outright by American sanctions. The implication was clear: if a small team in China could build a competitive AI model so cheaply using older hardware, what did that say about the hundreds of billions of dollars Western technology companies had committed to AI infrastructure?

Marc Andreessen, the venture capitalist and influential voice in Silicon Valley, called DeepSeek's R1 model an "AI Sputnik moment," invoking the Soviet satellite launch that had triggered the space race decades earlier. He described it as one of the most impressive breakthroughs he had witnessed and called the open-source release a gift to the world. The comparison carried weight: it suggested that the technological dominance the West had assumed in artificial intelligence might be shifting, just as American confidence in space exploration had been shaken in 1957.

The selloff extended far beyond chipmakers. ASML, the Dutch company that manufactures the equipment used by Taiwan's TSMC, Intel, and Samsung to produce advanced semiconductors, fell nearly 11 percent. SoftBank Group in Japan dropped more than 8 percent, just days after announcing a $19 billion commitment to fund Stargate, a data-center joint venture with OpenAI. The timing underscored how quickly investor sentiment could reverse when assumptions about the future of technology spending came into question.

What made the moment particularly unsettling was the scale of the valuations that had been built on the premise of Western AI dominance. Nvidia's stock had risen more than 200 percent over the previous 18 months, climbing to a valuation of 56 times its earnings—a multiple far exceeding the broader Nasdaq, which traded at 16 times earnings. The entire rally in technology stocks had been powered by the belief that artificial intelligence would generate enormous returns, justifying massive capital expenditures by companies like Microsoft and Apple. Now, with a cheaper alternative emerging from an unexpected quarter, investors were forced to reconsider whether those valuations could hold.

Jon Withaar, a senior portfolio manager at Pictet Asset Management, acknowledged the uncertainty. He noted that the details of DeepSeek's claims had not been independently verified and that nothing had been confirmed with absolute certainty. But he also recognized what was at stake: if the cost to train advanced AI models had genuinely dropped from nine figures to six figures, it would be transformative for the industry and its users, even if it meant lower returns for the companies that had bet so heavily on expensive infrastructure. Masahiro Ichikawa, chief market strategist at Sumitomo Mitsui DS Asset Management, captured the broader anxiety: the assumption that American technology—Nvidia chips, ChatGPT—represented the world's most advanced capabilities was no longer guaranteed. That perspective, he suggested, might be beginning to shift.