Does The AI Bubble Start to Lose Air?

TECH & AI

Despite record financial results from US tech platforms, their stock market valuations are taking a hit.

The risks of overvaluation are very real. Is this a wake-up call to deflate the bubble before it bursts?

Microsoft has just announced a voluntary redundancy plan affecting roughly 60,000 employees, about a quarter of its workforce. Google shed 13% in a single day, its worst day on record, wiping around $500 billion off its market value. Amazon fell 5% after its quarterly results. Over October as a whole, Meta lost 15%, while Nvidia and SpaceX each dropped about 10%..

So how are US tech platforms really doing, financially speaking?

On paper, the quarterly numbers are staggering and accelerating:

Amazon brought in over $202 billion in a single quarter, up 12%, with record profits up 61%.
Google posted $127.6 billion in revenue, up 25%. Search alone generated $63.3 billion (up 12%), and Cloud reached $30 billion, roughly double its level a year earlier.
Microsoft reported $45.9 billion in operating profit for the quarter, up 21%, and revenue from Azure, its cloud division, jumped 48%.
Meta’s advertising revenue climbed 26% year on year to $63 billion.

So what explains the share price slide, and are we heading for the bubble to deflate?

I’m not a financial analyst, and I’d much rather see this AI bubble lose air gradually than burst. Still, the risks are real, and they explain why investors and analysts are getting jittery and increasingly wary:

Capital expenditure on AI infrastructure is sky-high and doesn’t really take physical limits into account. Amazon alone is expected to spend around $220 billion in 2026.
Free cash flow is shrinking or turning negative: minus $2 billion for Meta in the quarter, and minus $42 billion for Amazon over the past twelve months.
Debt is piling up, both at established players (Intel, down 30%, being a case in point) and at neoclouds like CoreWeave. Even the digital giants are starting to borrow. As Gilles Babinet reminded us, neocloud debt now stands at $2.1 trillion, whereas at the height of the 2008 financial crisis, outstanding subprime mortgage loans totaled $1.3 trillion.
Doubts over AI’s return on investment are growing. Enterprise adoption has been slower than hoped, and productivity gains are still barely visible in the macroeconomic data.
Circular financing is becoming widespread on an unprecedented scale, with Nvidia effectively acting as the banker of the AI ecosystem.
There is mounting concern about risk pockets, namely unregulated lenders carrying large off-balance-sheet commitments.
The economic and political backdrop is tense, with US midterms on the horizon and Anthropic’s IPO looming.

Anthropic’s IPO: a true outlier?

It would be tempting to blame the current turbulence on Anthropic’s IPO, which could raise as much as $100 billion as early as the week of November 9. Its numbers are more reassuring than expected: revenue in the first half grew from $1.2 billion in 2025 to $16.3 billion, and it relies on a more diversified customer base than it might appear.

But the IPO is a genuine oddball. No company has ever stated so openly the risks it poses to humanity, adopted a governance structure quite like this one (the seven founders hold equal voting rights), or sought a valuation as eye-watering as nearly $2 trillion.

To step back and gain some perspective, I’d recommend the in-depth interview Simon Johnson, the 2024 Nobel laureate in economics, gave to French daily Les Echos.

By Thomas Husson, independent analyst.

I believe in quality over quantity so don’t expect daily or even weekly updates, but rather bi-monthly or monthly news – assuming I’ll only write when I have something meaningful to share.