Nvidia

On Monday, Chinese company DeepSeek claimed that after investment of a mere $5.6m it was able to replicate the functionality of, what were deemed, the leading AI models. DeepSeek R1, its new “reasoning model”, was released on 20th January 2025 and has already become the top free app on the Apple Store (source: Apple Store), startling its rivals and stockmarkets.

Nvidia had become the world’s most valuable company in 2024, as its revenues from AI computer chips soared. This spending surge on Nvidia’s computer chips was predicated on the hope that the AI models developed using these chips could deliver value to Nvidia’s clients and their end customers. As of today, there is scant evidence that this massive investment has delivered any value. It is also noteworthy that over 40% of Nvidia’s sales came from fellow Magnificent 7 members (source: yahoo! finance).

[The Magnificent 7 are a group of 7 US companies that have been responsible for a large portion of all US investment returns for the past 2 years]

Concerns have been rising among investors that this huge investment (c$200bn estimated for 2025, 

following an estimated $165bn in 2024, source: Bloomberg Intelligence) may be malinvestment; companies overinvesting because of the hype and fear of missing out on what evangelists claim will be a transformative technology. Since the midpoint of 2024, Nvidia’s share price has stalled as the hype met reality.

The addition of the news on Monday of this apparent tech breakthrough allied to those lingering concerns wiped $590bn off Nvidia market capitalisation of $3,500bn (source: Bloomberg). 

Some context:

• This is the largest one day loss of value in global stock market history. 
• Annual US military spending is not much more than this. 
• Nvidia was valued at “only” $590bn in March 2023. 

This was front page news across the globe. The lines of discussion focused on this new entrant, the merits of the tech, the ramifications to the competitive landscape, the proportional investor reaction to the news etc. 

Our first reaction to the news and events was not in line with this received market commentary. Our first reaction was that this was probably market abuse. 

[Market abuse relates to behaviours that unfairly disadvantage investors. An example would be a company releasing a claim of an astonishing breakthrough, with the intention to damage the share prices of its competitors, in order to profit from the subsequent share price movements. Market abuse is often focused on those areas of the market that are the most volatile.]

DeepSeek was founded in 2023. It is owned and funded by a hedge fund called High-Flyer. 

Alarm bells should be ringing. 

What are the chances that 1) with minimal investment 2) a hedge fund 3) has trounced the US AI behemoths 4) in under two years?  

Incentives drive behaviours. Show us the incentive and we’ll show you the outcome. Hedge funds are not known for their ability to develop consumer technology. Their raison d’etre is to make money for their founders and investors and they do this by profiting from market swings. There was a substantial market swing on Monday and it seems inconceivable that this hedge fund did not profit from that. Consider the counter-argument: what would be the reaction of the chief investment officer of a hedge fund, that has just developed and released a competition-crushing piece of tech, if his investment team had not shorted those competitors!?

[Shorting a stock is investing in a way that profits when a share price goes lower. It is a strategy often employed by hedge funds in particular.] 

Short sellers have reportedly made $6.75bn in profits from the fall in Nvidia’s share price on Monday  (source: FT, S3 Partners). The scale of the financial opportunity in the movement of Nvidia’s share price is obvious. 

AI associated stock market companies are ripe for market abuse for several reasons. 

• The technology is itself opaque. There are a few major competitors investing tens of $bns in an arms race, and these competitors are not seeking revenue or profitability at this early stage. They have a much greater incentive for privacy than disclosure over the effectiveness of this investment.
• There is a gold-rush dynamic with spending on AI chips. The Magnificent 7 are estimated to have purchased 1.27m chipsets from Nvidia in 2024 (source: FT). The all-in cost of a Hopper 100 chipset to purchase and fit in a data centre is estimated at around $50,000 (source: MacroStrategy), which gives an estimated total spend of around $63.7bn.
• There is little evidence to support the notion that the huge investment from tech firms in Nvidia’s chips has created any value for the Magnificent 7 buyers so far. Their R&D budgets have swelled but their profitability from AI is not disclosed (presumably because the disclosure would reveal limited revenues and significant losses). 
• Extending this, there is no evidence of material value being created for the customers of AI products created by the likes of Microsoft or Alphabet. 
• There is ample evidence that firms associated with the success of AI have seen their market capitalisations soar over the last three years.

One of our primary concerns over the last two years has been that the AI boom in the US stock market had clear indications of hype being a component. It also had some deep underpinnings of an exciting new technology in its early stages. 

Another was that the concentration of the growth and value of the S&P 500 on the Magnificent 7 has created growing risk management questions (from investors’ perspectives). On Friday last week the proportion of the S&P 500 represented by the Magnificent 7 was 33% (source: Bloomberg). 

Our investment position has the following components:

• The Magnificent 7 are outstanding companies.
• They have the profitability to reflect their monopolistic positions.
• In the most part they are not obviously overvalued.
• The heightened risk of investing one third of our clients’ US equity exposure in those companies presents a significant barrier. 
• We want to maintain an exposure to these companies, but not to the extent that is represented by the S&P 500.
• Our investment weighting of these stocks is currently around half that of the S&P 500 index.

The news on Monday and the subsequent share price reaction revealed that there is considerable speculative content (both pushing prices higher and lower) in AI stocks. It also reveals that investor sentiment is retreating from levels we would characterise as “messianic zeal” to at least a vestige of healthy scepticism. 

Overall we doubt that DeepSeek represents an enduring threat to the leading AI systems. But we must expect that as the AI gold-fields continue to be competed over that several things will happen; there will be more claims and counter claims, more media coverage and speculation, a deeper appreciation of the threat to incumbents, and ultimately, an increase in share price volatility.

It is pleasing when adherence to one of the key tenets of investing – effective risk management – delivers for our clients. The fall in US equity markets was largely confined to Nvidia and the other AI tech stocks – the equal weighted version of the S&P 500 was flat on the day (source: Bloomberg). Limiting our client’s exposure to the Magnificent 7 better protected client portfolios from the market volatility. For now, we will make no adjustment to our US positioning, but the position remains dynamic.