Rocking the AI boat

  • Nvidia stock fell over 15% this week, breaking the record for market capitalisation lost in a day ($589bn)1
  • DeepSeek ‘s AI model reportedly only cost $6m to develop2
  • Large US tech companies represent 39% of the S&P 500 and 41% of total AI research and development spend, but are only 13% of total sales3

Earlier this week, an undesirable Wall Street record was broken in the largest amount of market cap lost in a day. Nvidia fell a little over 15% which saw a whopping $589 billion wiped off its market cap. To put this into perspective, the value it lost in one day is more than 3 times the value of the largest UK company (AstraZeneca). This fall beat out the previous record of $279 billion lost in a day, set also by Nvidia in September last year. This knocked it off the top spot as the most valuable company in the world with Apple and Microsoft overtaking.

It wasn’t just Nvidia though that suffered heavy losses, the whole AI supply chain was severely impacted. This included fellow chip designers like Broadcom, chip manufacturers like ASML, data storers like Oracle and infrastructure specialists like Vertiv; they all saw double digit percentage falls. Such is the weighting of AI in the S&P 500 and Nasdaq 100 that it dragged these US indices down 1.5% and 3.1% respectively.

Why did this happen?

The answer to this question is a Chinese company called DeepSeek. Whilst the magnificent seven in the US have been spending hundreds of billions of dollars between them on AI, DeepSeek claims to have spent a mere $6 million to develop its highly competitive AI model. Spending on AI has been under intense scrutiny for years. Before Nvidia beat the record for the largest one day drop in market cap for its first time in 2024, the previous record was held by Meta in February 2022 which was caused by the company’s excessive spending on the metaverse and AI. The first chart below shows the increase in capital expenditure (capex) from AI’s big spenders over the years.

The second chart shows that whilst the large US tech companies represent 39% of the S&P 500 and 41% of total R&D spend, they are only 13% of total sales. The market took one look at DeepSeek, that has raced to the top spot in the App Store across 51 countries just days after launching, and once again questions the spending of the mega-cap tech companies (and their return on investment). Additionally, and perhaps more importantly, DeepSeek used Huawei’s chips to run inference (generate responses) instead of Nvidia’s. The US banned the export of Nvidia’s most advanced chips to China (they’re also far more expensive than Huawei’s.

Source: FactSet, 29 January 2025 via Citywire

Source: JPM Guide to the Markets – UK, 27 January 2025

There are a few unanswered questions with DeepSeek, however. Firstly, the US is investigating whether DeepSeek used Nvidia’s banned chips after all. It is broadly recognised that chips are smuggled illegally from countries such as Malaysia, Singapore and the UAE into China. This could leave Nvidia’s economic moat and pricing power intact. Secondly, there are serious doubt over the spending figures. Unfortunately, data coming out of China is not always reliable and so the spend could be played down intentionally – it did wipe over a trillion dollars off US tech’s market cap. Some analysts also expect that the amount spent was just on computing power, rather than all the development costs.

Lastly, DeepSeek was built from Meta’s open-source AI, Llama. They couldn’t have achieved what they did without the prior expenditure from Meta and Llama was intentionally accessible to accelerate innovation and AI development. It will in turn help Meta improve their AI models and continue to innovate and push the boundaries. It is a testament to their product if anything. Satya Nadella, Microsoft CEO, made a comment along similar lines: “as AI becomes more efficient and accessible, we will see exponentially more demand”.

Bowmore Portfolios

We see this as a bit of a wake-up call for US tech, who need to start demonstrating to impatient investors that their capital expenditure will translate into consumption and profits. We are beginning to see more use cases of AI and DeepSeek is just another example of how AI developments can be leveraged. Whilst it may not always be beneficial to the big tech in the short term, broader adoption is generally good, and Microsoft and Meta’s early lead in infrastructure could be a real advantage over the next few years.

The key now will be monetization and is what we’re watching. We have a meaningful allocation to US Equities and these developments do not change that. We do have a more balanced approach, tilting slightly away from the concentration at the top end of the market, which has certainly helped this week. It also plays right into our outlook of a broadening of the US tech rally we saw in 2024.

Source: L SEG Datastream, as at 30/01/2024

1  Forbes, Biggest Market Loss In History, January 27 2025 

2  Reuters, Microsoft, Meta back big spending, January 30 2025 

3 JPM Guide to the Markets – UK, 27 January 2025 

10 Jan 2025

Europe: where’s the growth coming from?

06 Dec 2024

The Santa Rally

Get in
touch