is ai a bubble?

by Hoshang Daroga CFA, Investment Director, Elston Consulting

Is the AI boom starting to look like a bubble?

Is the AI boom a bubble?

​To determine if the AI boom has entered bubble territory, we must first define what a bubble is: a period where valuations are stretched and the prices people pay for stocks become disconnected from fundamentals, reality, or underlying earnings. Currently, the tech sector is trading at approximately 35 to 40 times earnings. While these valuations are certainly stretched, they are not yet "absolutely exorbitant" when compared to historical precedents like the tech bubble of the 2000s, where price-to-earnings ratios reached 80 times.

Another factor to consider is market participation; currently, the breadth of the market is narrow, meaning investors are acting somewhat rationally by only rewarding companies that are actively delivering AI solutions or are directly involved in the sector. If the market shifted to a state where every stock rallied simply by mentioning "AI," that would indicate a late-stage bubble. Furthermore, the current growth is largely financed by operating cash flows from major hyperscalers rather than the excessive over-leveraging seen in past market crashes. Consequently, while the market might be in the early stages of a bubble, it does not yet feel massive or disconnected from reality.

What are the warning signs?

There are several specific metrics and behaviours that would signal a dangerous shift in the market. First, if valuations were to climb further and reach that 80 times price-to-earnings threshold, it would be a major cause for concern. Second, a shift toward universal optimism, where all stocks rise regardless of their individual merits, would indicate a lack of investor discipline. Third, investors should watch for a deterioration in financial conditions, specifically if companies begin issuing significantly more debt to build data centres while the Federal Reserve starts raising interest rates.

Finally, "price action" serves as a critical warning; in historical bubbles, stock indices often see 50% year-over-year growth that is entirely disconnected from fundamentals. Currently, US equity earnings growth is more modest at about 10%, with AI-specific companies growing at 20% to 30%. We see the silver market as a cautionary example of a true bubble, noting that silver prices recently exploded from $30 to as high as $115 - a 380% increase in a year which is a very short period of time, which represents a significant disconnect from physical supply-and-demand fundamentals.

How does this boom compare to others?

The most relevant comparison to the current AI surge is the 2000s dot-com bubble, but there are fundamental differences in how this wave is being funded. During the dot-com era, only 7% of companies were profitable and generating positive free cash flow. In contrast, the current AI boom is driven by the massive operating free cash flow of "hyperscalers" like Amazon, Microsoft, and Google. These companies are seeing actual revenue growth of 20% to 40% year-over-year. To illustrate the scale, quarterly cloud revenues have reached $33 billion for Amazon, $31 billion for Microsoft, and $15 billion for Google, totalling over $200 to $300 billion annually.
​
These companies are choosing to deploy this capital into data centres rather than returning it all to shareholders through buybacks or dividends because they believe the risk of not building fast enough far outweighs the risk of overbuilding. Currently, capacity utilization for data centres is near 80%, and vacancy rates are extremely low. For these giants, failing to build out infrastructure means being left behind in a period of rapid AI adoption.

If AI is a bubble, do you think it will pop in 2026?

​Predicting the future is difficult, but based on current fundamentals, a massive "pop"-defined as a 30% to 80% drawdown similar to the dot-com crash - is considered unlikely in 2026. The Federal Reserve is expected to cut interest rates, which supports easier financial conditions and a cyclical economic upswing.
Instead of a total collapse, the most likely outcome for 2026 is that the bubble will slightly deflate, which is described as the "best-case scenario" for investors. This would likely involve a rotation where investors move money out of expensive large-cap stocks and Nvidia and into cheaper areas of the market that are just beginning to reap the benefits of using AI within their own business models. This transition would allow the broader market to gain without experiencing the sharp losses associated with a sell-off.

Will there be a market correction?

While a massive crash is deemed the least likely scenario because real earnings are powering the global market, a correction is still possible in 2026 or 2027. This could be triggered by a "CapEx air pocket," a point where the rapid growth of capital expenditure from hyperscalers begins to flatline. If investors realise that spending will not increase indefinitely at the current rate, it could give the market a "jolt" leading to a 10% to 20% drawdown. However, such a correction would actually be a positive sign of prudent capital deployment and a healthy stabilisation of the market rather than a catastrophic burst.