Over the past two years, as advances in artificial intelligence turned into a massive tailwind for the US stock market – carried aloft by a relatively small number of companies with a credible claim to be at the center of the AI story – skeptical observers compared the phenomenon to the Internet boom of the late 1990s. That boom, as anyone around at the time will no doubt recall, ended with a spectacular crash at the turn of the millennium. So long to Pets.com and its ilk, popped into weightless effervescence as the dot-com bubble burst. So too, the AI doubters have insisted, will be the fate of those enterprises today with their promises of world-changing feats of economic productivity brought about by their high-octane large language modules (LLMs).
The vibe of Schadenfreude was no doubt making its rounds among the critics of AI share prices this week, watching as the likes of Nvidia and Microsoft tumbled on Monday when a barely-known Chinese operation called DeepSeek burst onto the scene and, well, deep-sixed just about everything that had hitherto been conventional wisdom about generative AI – especially the billions and billions of expenditure dollars thought to be needed to make bigger and better LLMs, and the many megawatts of energy needed to power the data centers that bring them to life. DeepSeek appears to have capabilities more or less on par with those of ChatGPT, the iconic platform introduced by OpenAI in late 2022. But – to explain why all those AI-related stocks sank on Monday – DeepSeek can do all that at a fraction of the cost. Nor does this appear to be some kind of elaborate Chinese ruse. The company behind DeepSeek, a hedge fund called High-Flyer that specializes in AI applications for trading in financial markets, has made its training results public and its systems available on an open-source license. Anyone who wants to build their own programming on top of these systems can do so without permission. Scant wonder, then, that DeepSeek shot to the top of app store downloads in the US this week.
Getting back to those share prices for Mag Seven names (the Wall Street shorthand for those hitherto high-flying AI stocks). Is DeepSeek the catalyst for the bursting of the bubble – if indeed that is the right term to use? We imagine there will be plenty of short-term volatility in these shares as bargain-hunters do their thing, skeptics do theirs, and asset allocators interpreting all of this as a signal for repositioning between value and growth stocks do theirs. For our part, we think this is a new chapter in what we expect to be a very long story, with an ending that is nowhere near in sight. The capabilities of AI LLMs has grown significantly since the release of that first ChatGPT app. Now we have evidence that further growth is achievable with potentially greater cost and energy efficiencies – which is likely to spark more interest and produce new entrants whose names we don’t yet know. In turn, the overall potential market size is likely to grow. The established players will have a role to play here. At a recent tech conference, Nvidia CEO Jensen Huang pointed to robotics as the next major use case, with AI-empowered robots taking over a large swath of agent functions throughout the massive (and often unproductive) services economy.
“Agentics”—the techy buzzword for this – could be a word you hear circulating around a great many cocktail parties this year.
There will be winners and losers, in other words, as there always are with radical breakthroughs. Think back to Internet 1.0 in the late 1990s. Lots of those names went out of business, while others (such as Amazon) went on to be the behemoths of successive chapters in that story. There will be risks and there will be opportunities. But for better or for worse, artificial intelligence is very likely going to be an increasingly influential component of the global economy. The newest chapter of the story just got under way.