DeepSeek’s $1 trillion stock market crash: Nvidia, TSMC, and Broadcom recover some losses after AI shock

Yesterday, shockwaves rippled across the American tech industry after news spread over the weekend about a powerful new large language model (LLM) from China called DeepSeek.

News of DeepSeek’s capabilities—not to mention the fact that it is open-source and free for anyone to use and modify—sent U.S. markets reeling, including the tech-heavy Nasdaq, which saw $1 trillion evaporate from its market cap as AI-adjacent stocks such as Nvidia and Broadcom were hit hard. U.S.-listed shares of TSMC, which trade on the New York Stock Exchange (NYSE), also took a dive.

But today, some of those stocks are recovering, at least to a degree. Here’s what you need to know about DeepSeek and its latest market impact.

‘DeepSeek R1 is AI’s Sputnik moment.’

Investor and engineer Marc Andreessen posted on X yesterday that ‘Deepseek R1 is AI’s Sputnik moment.’

DeepSeek R1 is the Chinese firm’s latest reasoning mode. “AI’s Sputnik moment” refers to the time when the Soviets leapfrogged the U.S. in the space race with the launch of the world’s first satellite, a milestone that caught America off guard.

Andreessen and many other tech experts seem to believe that DeepSeek is a similar milestone—and for several reasons.

First, not only did DeepSeek’s AI model outperform reigning U.S. champions like OpenAI’s ChatGPT and Meta’s Llama, but it was made at a fraction of the cost that U.S. tech giants spent developing their homegrown LLMs. DeepSeek reportedly cost less than $6 million to train, while U.S. tech giants have spent hundreds of millions or billions to develop theirs. 

Second, DeepSeek was reportedly trained on midrange AI hardware—Nvidia’s H800 chips. It was previously thought that a model with such industry-defining capabilities couldn’t be trained on anything but the latest high-end chipsets.

Third, DeepSeek’s LLM is also more energy efficient, making it more environmentally friendly—not to mention cheaper to run.

These three factors made it appear that America’s tech giants vastly overspent on training their LLMs, which now appear to be inferior to DeepSeek. This also suggests that America’s major tech giants operating in the AI space, including OpenAI, Meta, and Google, aren’t as impenetrable to competition as once thought.

When the financial barrier to entry into creating an LLM that could compete with America’s best models was thought to be relatively high—a company would need hundreds of millions or billions in capital to enter the race—it gave America’s tech giants a competition buffer. Not many other tech companies, and certainly not upstarts, would have the financial resources to compete.

But now, if they can compete for just a few million dollars, America’s AI tech giants might have a lot more competition in the months ahead, threatening their AI dominance.

Why did DeepSeek knock $1 trillion off U.S. markets?

After news of DeepSeek’s achievements spread, U.S. markets sank yesterday, especially the tech-heavy Nasdaq. By the end of the day, the Nasdaq had lost $1 trillion. The majority of that loss came from a sell-off of Nvidia shares.

As noted by CNBC, Nvidia’s stock (Nasdaq: NVDA) plummeted nearly 17% yesterday, which wiped almost $600 billion from its market cap. Other AI-adjacent stocks like chipmaker Broadcom Inc. (Nasdaq: AVGO) fell over 17%, and OpenAI’s largest investor, Microsoft Corporation (Nasdaq: MSFT), fell over 2%. These and falls in other AI-related tech stocks helped account for that $1 trillion loss.

As for why DeepSeek sent shares tumbling, it’s because its existence—including how little it cost to train and the inferior hardware it was trained on—is a threat to the interests of some of the reigning American AI giants.

If advanced AI models can now be trained on lower-spec hardware, why should companies keep shoveling money to Nvidia for their latest, most costly chips? And if any company can create a high-performance LLM for a fraction of the cost that was once thought to be required, America’s AI giants are about to have much more competition than ever imagined.

That kind of news scares investors who have invested heavily in America’s AI tech giants over the past few years.

How are U.S. tech stocks reacting this morning?

The good news for tech-heavy investors is that in premarket trading this morning, many U.S. tech stocks that plummeted yesterday are recovering today, albeit slightly.

As of the time of this writing, Nvidia shares are up about 5% over yesterday’s close. Broadcom shares are up about 3.4%. TSMC shares are up about 3.2%. However, shares in Microsoft and in chip-tooling maker ASML are relatively flat.

This doesn’t necessarily mean DeepSeek’s effect on U.S. stock markets is over. Today’s slight recovery of yesterday’s biggest losers likely suggests that some investors are seemingly catching their collective breaths as they wait to see how America’s AI leaders respond to “AI’s Sputnik moment” as the week continues.

View the full article