Professional investors are cautiously convinced that the sudden arrival of Chinese AI chatbot DeepSeek R1, which has caused huge market bloodletting over the past week, is a game-changer. Only, they aren’t yet sure which particular game it is changing.
Major technology stocks have seen their values slashed by proportions not seen since the dot-com bubble burst. The leading example is chipmaker Nvidia, which has been thumped almost 15% this week, logging a record-breaking loss of about $500bn in market value. But other big tech stocks weren’t spared, including Broadcom (down 14%) and Microsoft (down 4%), while power companies and even some property companies took a hit.
Yet after the initial heart-attack inducing falls, US markets have stabilised, with shares in some big tech companies like Meta even rising. Still, the question lingers: to what extent has the advent of DeepSeek changed the broader AI investing environment, which has been a money-spinner for investors over the past year.
The reason DeepSeek caused such a rumble is no secret: its model is as effective as other frontier operators like ChatGPT, but it was trained, theoretically, using approximately 2,000 graphics processing units over 55 days at a cost of $5.58m. That is laughably less than the hundreds of millions reportedly spent by competitors.
Tech writer Noah Smith summarises the key early takeaways from DeepSeek’s emergence as this:
- Large language models (LLMs) don’t have very much of a “moat” – a lot of people are going to be able to make very good AI of this type, no matter what anyone does.
- The idea that America can legislate “AI safety” by slowing down progress in the field is now doomed.
- Competing with China by denying it the intangible parts of LLMs – algorithmic secrets and model weights – is simply not going to work.
So how should investors respond? Is this a wake-up moment, or, oddly, a reason to invest even more?
Tech bosses have shrugged off the panic. Meta’s Mark Zuckerberg signalled that he still intends to invest “hundreds of billions of dollars” in AI infrastructure, while Dario Amodei, CEO of Anthropic, has said that DeepSeek’s models are impressive but not unexpected. Amodei reckons DeepSeek’s efficiency gains fit within the established trajectory of AI cost reductions.
Paul Theron, CEO of local fund manager Vestact, has been telling his clients to stick with their investments. “Sit on your hands, don’t overthink this,” he says.
The arrival of DeepSeek is more of an “interesting left-field development” than a “game-changer”, Theron says. There are going to be a huge number of new developments, he reckons, and this is likely to be just one of them.
While DeepSeek has caused investors to confront their own prior assumptions, that too was to be expected.
Where’s the return?
There are, from an investment perspective, two other dynamics that are critical in assessing the fallout this week.
First, markets as a whole have been running very hard, which always increases the risk of volatility. Even after the fallback, the S&P 500, the index of the biggest US companies, is up 24% over the year, and is trading on a p:e of 27, which is relatively high. At those levels, the smallest piece of bad news is going to hit the market hard.
Second, there is concentration risk, which is the risk of overexposure to a particular sector or size of company. Both are relevant here, because the previous conventional wisdom was that, because they were so expensive to create, LLMs could only really be built by very large tech companies.
As it is, the weight of the 10 largest companies in the S&P 500 is massively higher than it has been in half a century, Schroders chief investment officer Johanna Kyrklund pointed out in a note to clients.

“From a portfolio standpoint, having such a high exposure to just a handful of stocks does not feel prudent. Given the concentrated nature of the market, this is not a time for unintentional bets,” she says.
Yet, Theron says the outperformance of these large companies is not accidental: they have enormous customer bases, they have huge “moats” that prevent easy competition, and they have huge margins.
Anchor Capital CEO Peter Armitage says the advent of DeepSeek has meant that “a lot of naivety has been revealed”.
“They opened the curtain and everybody went ‘wow’,” he says. But he points out that this arena is so new that it is likely a new curtain will be opened in this field every few months.
Armitage says that one of the things that is not clear yet is what the actual business models of these companies will be. At this stage, AI is a kind of “have to have”, so big tech companies are including AI in their existing packages, rather than becoming an additional revenue stream, he points out. At the same time capital expenditure on AI in the tech world has shifted from 5% to 60%, while capital expenditure has risen by more than 20%.
“It reminds me a lot of the internet between 1995 and 2001, where everybody got a premium rating and got valued at a hundred times turnover just on the basis of eyeballs,” he says. “The actual business models and who made money out of them only emerged over three or four years.”
Armitage feels this is not a game-changer, but it’s definitely a reveal. “I’m incredibly excited about the way it can change the world and the speed it will do it. I’m not bearish about tech; the only bearish element is actually, as an analyst working out who’s spending money, will the money get a return?”
That is a question DeepSeek, for all its fancy design, has not yet answered.
Sign up to Currency’s weekly newsletters to receive your own bulletin of weekday news and weekend treats. Register here.