A trillion dollars went up in smoke last Monday. The “culprit” was DeepSeek, the Chinese open-source chatbot alternative to ChatGPT. All of a sudden, the immense computing power seemingly needed for state-of-the-art AI proved to be a nice-to-have and nothing more. Here are the biggest tech stock dives we saw on Monday.
The biggest name to take a beating yesterday was Nvidia. Shares of the leading AI chipmaker slumped by a substantial 16.86 percent. A more significant feat is that it saw the company evaporate a record number of dollars from its total market value: 589 billion to be exact. It has doubled the previous record, which was already held by Nvidia. On Sept. 3, it lost $279 billion, which it regained in the weeks after.
It is not, however, the biggest percentage-point loss of the day among tech companies. Broadcom‘s shares fell by 17.4 percent. As a result, its stock market value dropped back below $1 trillion. As a critical network equipment supplier, the company plays a major role in building out AI infrastructure. Although Broadcom took the hardest stock market hit among IT-focused companies, there are other organizations that had an even worse day. We’ll discuss those in a bit.
A similar yet smaller hit was seen at memory maker Micron Technologies with a loss of 11.67 percent on Monday. HBM modules are the basis for the huge memory capacity on AI chips from Nvidia and others, which Micron builds along with its rivals SK Hynix and Samsung. On the software front, Oracle suffered a heavier reduction by 13.79 percent. The question, however, is whether this is entirely because of the DeepSeek revelation or also because of the rumour that Oracle is considering buying TikTok.
Smaller losses were seen at Nvidia’s direct competitors AMD (-6.37 percent) and Intel (-2.59 percent). Chip machine builder ASML slumped 5.75 percent, far less impressive than a loss due to leaked quarterly results in 2024. Taiwanese chip maker TSMC, which is responsible for manufacturing virtually all AI accelerators, even rose slightly on the Taipei Stock Exchange. We therefore cannot really call yesterday a ‘Black Monday’ of sorts by any means.
AI doubts affect more than just IT
DeepSeek’s revelation that advanced AI can also be built and offered cheaply has major implications. Many of those effects will take a while to sink in: after all, what are other AI players going to do with DeepSeek’s innovations in open-source models? These doubts may have tempered the panic that could have set in had DeepSeek-R1 been closed-off and more impressive. Meta‘s shares actually rose 1.91 percent in value last Monday. Amazon also had nothing to fear with a small 0.24 percent growth. Competitors Microsoft and Google parent company Alphabet lost 2.14 percent and 4.03 percent, respectively; normally not too significant a figure.
The real pain, however, was seen at some U.S. energy companies. The ambitious $500 billion Stargate project recently announced by President Trump leans firmly on expanding the U.S. energy infrastructure. But now that DeepSeek has cast doubt on the need for possibly even nuclear power plants next to data centers, players in this sector are taking a major hit on Wall Street. For example, Constellation Energy Corporation (CEG) lost a hefty 20.85 percent in stock market value, while rival Vistra Corp was Monday’s biggest loser thanks to a big 28.27 percent loss.
Some signs of recovery already
There are far too many fluctuations, tiny losses and even minor gains to really speak of stock market panic. AI doubts are certainly there and they stung at Wall Street’s valuations, mostly targeting Broadcom and Nvidia. However, the pre-market data is a lot rosier for both: Broadcom is heading for a slight recovery (3.9 percent) at the time of writing, as is Nvidia (4.33 percent). Even the energy sector’s players are showing signs of positive corrections before market open later today.
The fact that an AI company like Meta is not affected by DeepSeek shows enough market confidence in a reaction to the new Chinese challenger. Llama 4 may well incorporate all the lessons DeepSeek is teaching. On top of that, the immense computing power helps U.S. AI players as well, as they can train their LLMs faster, serve their models with less latency ánd build capacity. The familiar AI scaling laws of still apply: bigger models are better, more data is better, faster chips are better. Since China won’t easily get its hands on the latest Nvidia chips, Western players still have a big advantage. As long as DeepSeek won’t outperform its rivals with a closed-source model whose inner workings are black magic, the AI hype will not be floored by this Chinese player just yet.
Also read or listen: DeepSeek, hot on OpenAI’s heels, hit by cyberattack