- Nvidia is the first company to hit a $5 trillion market cap, after hitting $4 trillion in July 2025, $3 trillion in June 2024 and $1 trillion in June 20231
- Excessive spending by major cloud computing companies is outpacing revenue whilst share prices are outpacing profits, fuelling fears of a market bubble
- The productivity gains of AI, promising labour enhancement, is actually causing labour displacement with increasing amounts of layoffs at big tech companies like Amazon2
It seems we can’t go a day without seeing another article about the AI bubble, whether it is in one or it isn’t, and so we thought we’d add another one to the pile and give readers the Bowmore opinion.
What is a bubble?
A market bubble, can be characterised by three key stages.
1. Speculative Frenzy – A new idea or technology captures attention; investors start piling in because they believe it’s the next big thing or simply are afraid of missing out. Market bubbles are not a concept new to this century, in fact, our office is split on whether the first market bubble was Dutch Tulips in the 1630s or the South Sea bubble of 1720 (technically the first stock market bubble). More recently though, you had the Dot-Com bubble which saw internet stocks skyrocketing in the late 90s on hype and speculation. The internet was undeniably a world-changing technology, but irrational exuberance drove stock prices higher than companies profits justified.
We saw it again in 2007/08 in the US housing market where house prices looked like they could only go up, before they didn’t and mortgage defaults started rising which led to banks feeling the strain. The graph below shows a history of asset bubbles since 1977 including the current run of the Disruptors – includes the likes of Microsoft, Alphabet, Nvidia etc.
2. Divergence of fundamentals and valuations – The speculative frenzy takes us to stage two, where the buying of ‘promised potential’ causes prices to go up quicker than profits. This is where you see concentration build up and valuations become stretched. The top 10 stocks in the S&P 500 make up more than 40% of the index now, for reference, in the dotcom bubble the concentration of the top 10 peaked around 27%. The below chart from Montanaro shows the valuations of the US Growth index through a metric called Price to Book ratio (P/B), an indicator of what investors are paying per dollar of a company’s net assets.
3. Bubble pop – All trends come to an end. There’s always a different catalyst that pops the bubble – whether it’s companies running out of capital (Britain’s Canal building bubble of 1800) or failing to turn ideas into cash flow (Dot-Com bubble), or even macroeconomic factors (Gold bubble of 1980 deflated by a sharp drop in inflation).
So is AI a bubble?
Well, it definitely ticks a couple of the boxes. A new technology, AI, has caused a concentrated rally within the sector. Nvidia, for example, has just hit $5 trillion dollars in size, having been one tenth of that just three years prior. We have seen a divergence between fundamentals and valuations as evidenced by the above Price to Book chart. Our note back in September, titled, ‘The Price is Wrong’ focused in on this divergence – a stock market at all time highs, with a labour market cracking and inflation going in the wrong direction. Where AI was supposed to enhance productivity in the workforce, we are instead seeing signs of labour displacement – people losing their jobs to AI (Amazon has just cut 14,000 jobs)2 and companies with lower hiring intentions. And what goes hand in hand with Unemployment? Recession.
Lastly, there is the much-overlooked depreciation story. AI companies are spending an astronomical amount on datacentres and chips (Nvidia set to make $500 billion of GPU sales in 2026).1 Meta sold off 11% yesterday as investors found out about further spending commitments and soaring AI costs from the company. The problem with this spending is that it is not one-off, these chips depreciate as they break and become worn out. It is estimated that the latest range of Blackwell chips last 3 – 6 years depending on usage, and this recurring expense is not being factored in.
A rational bubble
Mohamed El-Erian, Chief Economic Advisor at Allianz, described the AI bubble as a ‘rational bubble’. This is hard to disagree with as, for all intents and purposes, it meets the definition of a bubble, but this is not like 2000 or 2007. What the dark blue line in the below chart from Goldman Sachs shows is that the spending from these AI companies is funded largely by their cash flow, instead of by debt as was more common in the late 90s. They are spending what they can afford to and their balance sheets aren’t stretched. The danger though is the light blue line which shows how the spending is not translating into revenue (sales) yet.
There’s definitely a risk in the monetisation of generative AI, but this spending is actually just part of a race to a far greater prize: AGI, Artificial General Intelligence, where AI can think and behave like humans, not just perform tasks we give it. Whilst this is a decade off, it has the power to reshape industries, massively enhance productivity and boost economic growth.
Bowmore portfolios
We think there is no denying the capabilities of AI, it is a new technology that is here to stay and will only get better. However, as stewards of capital, we cannot ignore the risks posed by elevated valuations and uncontrollable spending. We don’t think this is a bubble that will burst with companies suddenly going bust like in 2000, but rather one that will slowly deflate as the market gradually grows impatient with expenditure not translating into profits. It will come as no surprise then that we are materially underweight US Equities relative to global trackers that hold 65%+ in the US. We prefer regions with both higher growth prospects and discounted valuations like Japan and India. We have trimmed our exposure to US tech and, this week, to US Corporate Bonds, although we recognise that the innovation in the region is hard to argue against and maintain some exposure through dedicated low-cost trackers. We are also actively tapping into more niche, tech-adjacent themes like Cyber-security or Biotech that are making the most of AI.
Source: Alpha Terminal
The value of your investments can go down as well as up, so you could get back less than you invested. Past performance is not a guide to future performance.
Sources:
1Fortune: Nvidia is officially the world’s first $5 trillion company
2CNN, October 2025