Open Editor's Digest for free
Roula Khalaf, Editor of the FT, selects her favorite stories for this weekly newsletter.
It is difficult to talk about the economic history of the 21st century without discussing the “China shock”. That's the term often used to describe China's entry into global trade, a shift that brought rich countries an abundance of cheap goods, but left businesses and workers behind altogether.
DeepSeek can provide a sequence. A little-known Chinese hedge fund has thrown a grenade into the world of artificial intelligence with a huge modeling language that, in fact, matches the market leader, Sam Altman's OpenAI, for a fraction of the price. And while OpenAI treats the functionality of its models as proprietary, DeepSeek's R1 wears its technical insides on the outside, making it attractive for developers to use and build upon.
Things are moving fast in the AI era; so terrible. Five of the biggest AI-ready technology stocks – chipmaker Nvidia and so-called hyperscalers Alphabet, Amazon, Microsoft and Meta Platforms – collectively spent almost 750bn in market value before US markets open on Monday. It would be even more painful for Nvidia if it proves that DeepSeek is successful without the use of the brightest chips.
Investors in technology companies – including Europeans such as chipmaker ASML, and energy companies investors hoping to get a boost from fuel data centers – are left wondering if their investments will go up in smoke. Hyperscalers were due to plow nearly $300 billion in capital spending this year, according to Visible Alpha estimates. Analysts expect that on Wednesday, when they report earnings, Meta and Microsoft will report investments for 2024 totaling $94bn.
In fact, the game is not over yet. The true power of DeepSeek is still unclear, and it has not yet achieved “general artificial intelligence”, the human-like state pursued by Meta and OpenAI. But the rules may have changed. At the very least, DeepSeek can take on some of the biggest US clients. At worst, it has challenged the core belief that more hardware is the key to better AI. This principle supported the market value of Silicon Valley companies as they invested heavily.
What's wrong with hyperscalers may be a breeze for everyone. For most business users, having the best model is less important than having one that is reliable and good enough. Not every driver wants a Ferrari. Advances in R1-like reasoning can be a big step for “agencies” that deal with customers and perform tasks at work. If those are available at a lower cost, the company's profits should increase.
In this way, this Chinese shock couple may be similar to the first one. It can bring not only destruction but also restoration – even though it is painful for many. Researchers estimate that for every job lost in China's shock, The purchasing power of US households rose above $400,000. The race for AI supremacy is still on; mass production has begun.