DeepSeek’s Game-Changing Implications for the AI Play
The DeepSeek surprise that rattled markets on Monday, January 27, may actually be good news for AI in the long run. What’s more, it demonstrated the market’s breadth beyond the tech sector. Here’s why we think the DeepSeek story holds some interesting possibilities.
AI giant Nvdia’s stock price plummeted by 17% on Monday on news that a Chinese artificial intelligence company called DeepSeek has created a rival to OpenAI's ChatGPT that requires remarkably lower development costs than that of US tech giants. This was a truly rare shock, given that the company had been around for years but no one had any idea it had developed this capability until recently. The initial estimates are breathtaking -- DeepSeek’s model cost a purported $6 million to train, compared to the current industry benchmark of around $100 million, and a fraction of the $1 trillion OpenAI, Google, and other major U.S. companies have stated they will invest in AI in coming years. It’s no surprise that this news struck at the heart of AI-related company valuations.
We have long been saying that the AI momentum trade had some real vulnerabilities. Interestingly, we had been expecting something different in the way of a shock -- antitrust issues, tax issues, and regulation have been more likely candidates. But given the valuation levels for these companies, it was always the case that any shock, no matter what the source, would be impactful.
In the long run, according to our tech analysts, DeepSeek’s innovation will be great for the whole industry. It means lower costs for software companies (think Microsoft), lower costs for Cloud computing companies (think Amazon’s Web Services), and lower costs for any tech business where the cost of inference is significant (think Apple and Meta). It will be interesting to hear from the big tech companies that are reporting earnings about this development; for instance, Meta and Microsoft are reporting this week. As of Tuesday morning, January 28, AI generally had only fallen back to levels from earlier this month. The Goldman Sachs AI Basket, for instance, which holds everything “AI related” with about 108 companies from chips to energy to data center companies, is still only back to early January levels. This suggests that in the short term there could be more rotation/broadening to come.
But perhaps the single best data point from Monday’s session is related to increasing market breadth. In fact, despite the market declining 1.5%, more than 300 of the 500 stocks in the S&P 500 Index rose that day, a feat which has never happened in the nearly 70-year history of the index.1 This supports the view we’ve had since early 2024 that we expect the market to broaden out beyond tech at some point. Our healthy outlook for US equities this year has more to do with the non-tech names - and we don’t need outsized tech returns to achieve positive US equity returns this year.
On AI specifically, we eventually see the benefits of AI spreading to smaller names in the US, countries outside the US, and sectors outside tech, though we’ve been reticent to predict how fast that happens. DeepSeek has changed the name of the game though. Take for example biotech companies using Large Language Models (LLMs) for drug discovery. They could see far lower costs and could dramatically speed up advances in the space.
In the long run, this may indeed turn out to be a pivotal moment in the evolution of the tech industry: countries may no longer need to have large incumbent tech firms to afford the capital expenditure required to propel higher economic growth. From a macroeconomic perspective, this is potentially a huge advancement in democratizing AI technology.