Technology stocks fell for a second consecutive day on Friday as concern grew over the hundreds of billions of dollars the industry is committing to artificial intelligence. The Nasdaq, the index of the largest US-listed technology firms, ended October with its worst monthly slide since March 2025.
The scale of the commitment is the story. Meta, Alphabet, Oracle, Amazon and Microsoft have between them committed to spending $800bn in 2026. Amazon alone has committed $220bn, going into data centres, network equipment and servers, and Alphabet’s own capital budget for the year runs to a comparable figure.
What investors are actually objecting to
Not the spending itself — markets have rewarded heavy capital investment many times before. The objection is the widening gap between what is going out and what is coming back.
Amazon’s own filing showed cash going out faster than it comes in, to the tune of $42bn — a figure six times larger than three months earlier, which means the gap is not merely wide but widening. Alphabet, meanwhile, grew revenue by a quarter and still disappointed the market, because the advertising line came in light. Chief executive Andy Jassy said Amazon was spending in part because it had the opportunity to. That is a reasonable thing for a chief executive to say and an unsatisfying one for anyone holding the stock.
Why “bubble” is the word being used
A bubble is not simply a high price. It is a situation where the price depends on a future return that nobody can yet evidence, and where the spending is justified by the fear of being left behind rather than by a measured payback.
Both halves of that are now visible in public filings. The companies are explicit that they are spending to avoid being displaced; they are considerably less explicit about which products will earn the money back and when. That does not mean the investment is wrong — the internet build-out of the late 1990s was a bubble that also laid the fibre everything since has run on. It means the timing of the return is unknown, and markets price unknown timing as risk.
What this looks like from here
Three practical consequences for Bangladesh, none of them dramatic.
Cloud prices. The $800bn is being spent on capacity that has to be sold. Compute has been getting cheaper per unit, and an industry that has over-built sells it cheaply — which is good for anyone here renting GPU time rather than buying it.
Outsourcing demand. A spending squeeze at the large platforms historically pushes work towards cheaper suppliers rather than away from them. For the BPO and software-services sector that is not obviously bad news.
Hardware prices. This is the one to watch, and it runs the other way: memory and storage are being absorbed by AI data centres, and Dhaka’s component prices have already been moving on it for a year.
What would end the argument
One number, reported consistently, would settle most of this: revenue directly attributable to AI products, separated from the cloud and advertising lines it currently hides inside. None of the five companies reports it that way.
Until they do, investors are being asked to fund an $800bn programme on the strength of aggregate figures that cannot distinguish a genuine new business from a rebadged old one. That is not a conspiracy; it is a disclosure gap, and it is why a quarter of good overall results can still produce a bad day on the market. The companies that start breaking the number out will be the ones confident it looks good.
