The Hypothetical
- sebastienpautrot
- Sep 11
- 2 min read

12 September 2025
The current speculative episode in Artificial Intelligence (AI) has many parallels with the first dot.com bubble in 2000. Indeed, on some measures the divergence between price and fundamentals is even greater. As Soros said on bubbles, “financial markets, far from accurately reflecting all the available knowledge, always provide a distorted view of reality. The degree of distortion may vary. Sometimes it is insignificant, at other times it is quite pronounced. When there is a significant divergence between market prices and fundamentals, there is a lack of equilibrium conditions. Every bubble has two components: and underling trend that prevails and a misconception relating to that trend. When a positive feedback loop develops between the trend and the misconception, it sets a boom-bust process in motion.”
There is often legitimate fundamental support for a speculative mania. That reinforces the trend and the misconception of reality. It also encourages speculators to buy an asset on the hope that it will be worth more in the future. After the US close on Tuesday, Oracle reported disappointing earnings. However, sales guidance was extremely heroic over the next four years. The stock was trading +25% after hours. As a good friend noted, tech companies often trade on capex spend and pipeline. These get extrapolated to the point that isn’t sustainable then burst. Oracle was classic earnings miss versus pipeline beat. Oracle’s outlook appears even more heroic in the context of their earnings performance since the end of 2021. The stock price has gained almost 200% while earnings are down over the same period (chart 1).
Chart 1

Source: Bloomberg
As we have noted recently, there is a divergence between price/beliefs and fundamentals for the US information technology sector more broadly. The sector also now trades on 10 times sales. That is beyond the peak in 2000 at 8.5 times sales (chart 2). As former Sun Microsystems CEO Scott McNealy famously said in 2002, “at 10 times revenues, to give you a 10 year payback, I have to pay you 100% of revenues for 10 straight years in dividends… that assumes that I have zero cost of goods sold, which is very hard for a computer company…that assumes that I have no expenses… and pay no taxes.” Stated differently, 10 times revenue is probably heroic and not sustainable. Of course, price can decouple from fundamentals for an extended period in a speculative mania.
Chart 2

Source: Bloomberg
We are sympathetic to the positive feedback loop extending into next year if capital spending from the hyper-scalers continues. However, as we noted recently, AI spending has created a spiraling cost model. From a big picture perspective, if AI becomes cheaper, requires less power than previously thought, proliferates faster, or does not generate a sufficient return on capital to meet investor expectations existing capital spending outlays could be very wrong. Euphoric non-linear price action combined with heroic valuation is a warning signal that the trend is mature. From a macro perspective the deterioration in the labour market is also comparable to pre-2001 and 2008 and a red flag for US domestic final demand. We are not convinced that aggressive Fed rate cuts will be enough to stabilise growth and corporate profits over the coming months.
Comments