Many FTSE investment trusts were trading sharply lower today (27 January). In the blue-chip index, Scottish Mortgage Investment Trust fell 5.5%, while a host of FTSE 250 funds also slumped.
| Fall | |
|---|---|
| Polar Capital Technology Trust | -7.2% |
| Allianz Technology Trust | -7% |
| Baillie Gifford US Growth Trust | -5% |
| Edinburgh Worldwide Investment Trust | -3.9% |
| Monks Investment Trust | -3.8% |
| JPMorgan American Investment Trust | -3.7% |
The common theme here is that these are tech-focused trusts. They have varying degrees of exposure to Nvidia (NASDAQ: NVDA) and related artificial intelligence (AI) stocks, which are all down significantly in the pre-market over in New York.
As I write, the Nvidia share price is due to open 12% lower!
What on earth is going on?
Basically, a previously obscure Chinese start-up called DeepSeek has set the cat among the giant Nasdaq pigeons, tiggering what looks like a sizeable tech sell-off.
DeepSeek is a Chinese large language model (LLM) developer, similar to OpenAI’s ChatGPT. It released its R1 AI model on 20 January. However, it claims to have built this model far more cheaply, by using Nvidia’s less powerful H800 chips, as opposed to the tech giant’s top-shelf H100 version.
There are export restrictions on Nvidia’s H100 chips being sold in China, meaning scarcity might have sparked truly disruptive innovation.
Nvidia’s earnings growth is premised upon massive ongoing capital expenditure on AI infrastructure. Microsoft, for example, is planning to spend a colossal $80bn on AI in 2025. But apparently DeepSeek trained its latest top-performing model in two months for $5.6m — a fraction of the $100m-$1bn range cited by rival Anthropic last year.
The Chinese app is currently at the top of Apple‘s App Store rankings. Venture capitalist veteran Marc Andreessen said that this is potentially “one of the most amazing and impressive breakthroughs I’ve ever seen”.
Doubts
If hundreds of billions of dollars of investment in high-end chips and extensive computing power isn’t necessary, then the potential risk to Nvidia’s growth seems obvious. But my suspicion is that there’s more than meets the eye here. There are already doubts about the true cost of training this LLM.
More to the point, surely executives and AI engineers at some of the best tech companies on earth haven’t been completely blindsided by the exact costs needed to train these models. Have they?
Potential buying opportunities ahead
So, might there be buying opportunities coming up? Potentially yes, as it appears the market is selling first and asking questions later.
When there is widespread selling like this, it can mean high-quality stocks end up on sale. The baby gets thrown out with the bathwater, as it were.
Allianz Technology Trust and Polar Capital Technology Trust might be worth considering if the selling gets worse. Each has diversified exposure to many unstoppable technology trends, including cloud computing, e-commerce, semiconductors, and cybersecurity.
Meanwhile, Scottish Mortgage appears to have been ahead of the curve, having significantly reduced its Nvidia holding in recent months. In November, manager Tom Slater said: “Companies must find ways to offer competitively priced AI systems while managing the skyrocketing costs of training them. This raises concerns about the sustainability of current capital equipment spending, including Nvidia chips.”
Perhaps this disruption is what is needed for AI to go truly mainstream.
We have Microsoft, Meta Platforms, and Tesla reporting earnings this week. It’ll be fascinating to hear what management says on this if asked, particularly shrinking violet Elon Musk.
