In an instant, euphoria over artificial intelligence, AI, turned to panic, and US tech companies lost around about $1trn in value. What was the trigger? What does it tell us about the potential of AI as a technology and as an investment? (Clue – they are not the same thing.) And how did other investments respond i.e. should you be worried?
The trigger for this rout was the launch of an AI product called R1 by Chinese company DeepSeek. It was launched on Inauguration Day in the US, and in the days afterwards the implications of R1 quickly began to build. By last Friday the latest Economist had articles analysing how China was surprising everyone by pulling alongside the US on AI, where they were meant to be Masters Of The Universe.
It wasn’t until last Monday that financial markets figured what was happening. The assumptions behind the AI bull market, which had driven the S&P 500 to new bubble highs, were crushed.
There were three planks behind the AI investment mania:
- There will be massively successful products built on AI technology, the best prior example being the iphone for Apple.
- Only the very largest companies can afford the tens of billions of investment required, which creates a huge barrier to entry, and the potential for monopolistic profits, as Apple and Google and Microsoft already enjoy, with AI promising more of the same.
- Only Nvidia’s most advanced chips would be good enough to train the best AI systems.
Tens of billions have already been spent, but there is no obvious ‘killer app’ or product. With the release of DeepSeek’s R1, the second and third assumptions are also doubtful.
DeepSeek has shown that there always was a cheaper and simpler alternative to the massive, expensive, and monopolistic AI infrastructure created in the US, and quite a few analysts, particularly in the US, are very angry:
“This isn’t about China. It’s about how the American tech industry is incurious, lazy, entitled, directionless, and irresponsible… It’s been a con, a painfully obvious one… beneath the hype was an industry that provided modest at best outcomes rather than any kind of ‘next big thing’ ”.
There is a lot more like this.
Behind all of this emotion remains an extraordinary technology.
At its simplest artificial intelligence enables computers to perform tasks that we can perform, but faster or more consistently, or both. It has the potential to revolutionize industries like healthcare, education, and climate change by automating tasks and improving decision-making, or equip machines with the ability to perform tasks that mimic human intelligence. At the least it should boost productivity across a range of industries in the decades ahead. At best “it will power humanity through a transformation comparable to the Industrial Revolution”.
Before Stephen Hawking died he felt the possibilities for AI were much more extreme. He said it will either be the best thing that’s ever happened to us, or it will be the worst thing, and “if we’re not careful, it very well may be the last thing.”
Whatever the future applications, for better or worse, they are undiminished. In fact R1 illustrates that they could come quicker. Where the US giants wanted to hide what made AI tick, aided and abetted by their government, this merely inspired the likes of DeepSeek to innovate around the barriers.
As I said in the 22nd November note, in the context of the US blocking Chinese access to the most powerful chips, and the risk of Trump escalating this conflict:
“The parallel is Napoleon blockading British trade with Europe from 1806. This spurred Britain into action, the result being the 19th centuries domination by British industry”.
As Charles Gave put it:
“Never force your main competitor to become more competitive because you are not competitive enough”
The US has also lost the moral high ground, if it ever had it. The US products are behind walls – you do not know precisely how they work, and their monopolies are not about to share this knowledge with you. DeepSeek promised to develop AI for the public good, and prevent “monopolisation” by a few companies. If you want to download their product and use it to build your own, go ahead, no permission required. When the company releases new models they also publish papers providing a wealth of detail on what is under the bonnet.
Xi will be very happy with the PR potential of “open China, working for global good” vs “closed US, working for its own profit”.
If the potential benefits from AI are now brighter for the world, the investments opportunities are not. This week was perhaps the end of the beginning for AI investment cycle. It is nothing new, whether you turn to canals and railways in the UK in the 18th and 19th centuries, or tech and telecoms in 1998-2000. Greed trumps thoughtful analysis, and investors pile in. But the improved productivity, profits, and products from new technology, typically take much longer to unfold than implied by the headlong rush of naïve investors into “the next big thing”.
That the AI bubble has been pricked by China should not be a surprise. As we and others have highlighted regularly, China has a massive army of STEM graduates, roughly 9 times as many as the US, and this army of smart young people is not complacent, and very entrepreneurial. A study in 2023 found that China leads in 37 of the 44 advanced technologies which they measured (Source: Australian Strategic Policy Institute). Why such a shock that, as the Economist headline put it last week, “Chinese AI has stunned the world”?
Although vast sums were lost in tech stocks over the last week, on Monday, when the shock hit hardest, more shares in the S&P 500 were up than down. The damage was quite narrow. It could take months for the real damage to emerge in the valuations of the major tech stocks, once they are re-priced on the basis of no new-big-thing on the horizon.
Nonetheless, there were no crashes beyond this narrow, albeit very important, sector. The UK did well over the last week, the FTSE 100 and FTSE 250 up 2%. The tech-centric NASDAQ is barely down 2%, with the S&P down 0.8%. The Nikkei was off by 1.1%, but Japanese smaller companies were up around 2%. This is the performance gaps we expected as the less weak yen does not help the giant exporters, and attention has turned to very cheap smaller companies.
Perhaps that is another lesson from the last week. The cheaper markets are turning up, UK and Japanese smaller companies two obvious examples, and we will be looking to China with interest next week, as their markets re-open post celebrations for The Year Of The Snake.
It’s also time for the monthly What’s Hot, What’s Not feature. Four gold funds feature in the Hot funds, remembering that these invest into gold miners, which have been very cheap relative to the gold price, with Ninety One the star, up 12%. Two European funds also show, which is unusual, up 8% a piece. The not so hot funds are dominated by India and UK smaller companies, with India worst, down 6-7%. This showing by UK small caps might just be a contrarian indicator – very cheap, and notable rate cuts not far ahead.
Amongst Hot sectors, the three European sectors dominate, including European smallers, in contrast to the UK. US small caps also feature, and were largely unaffected by this weeks AI panic. The India sector is the worst, down 6% on average over the month, but the other four sectors are all UK-centred, from smaller companies to gilts and corporate bonds – though not good news, the underperformance by these sectors is not immediately worrying.
There won’t be a Friday Note next week, unless there is a big story.