Big Tech Is Spending $725 Billion On AI This Year – And Even Record Profits Aren’t Calming The Markets Down

Meta reported record quarterly profits last week – revenue was up 33% to $56.3 billion, EPS came in at $10.44 and by any conventional measure it was a strong quarter. Its stock dropped over 6% after hours anyway.

The reason was not the profits – it was the spending. Meta raised its full-year capital expenditure guidance to between $125 billion and $145 billion, up from its previous range of $115 to $135 billion. According to Fortune and Tom’s Hardware, when you add up the projections from Meta, Alphabet, Microsoft and Amazon, Big Tech’s combined AI infrastructure spend for 2026 now stands at $725 billion – a 77% increase on the $410 billion spent in 2025.

In Q1 alone, the four companies spent a combined $130 billion on capital expenditure, led by Alphabet at $35.7 billion and Meta at nearly $20 billion.

The Maths Behind The Madness

To put $725 billion in context: that is roughly the GDP of Switzerland – spent in a single year on AI infrastructure.

The size of the commitment makes it one of the largest coordinated capital deployments in the history of technology, and happening mainly on the basis that AI-driven revenue will eventually justify it – though the “eventually” part is doing a lot of work in that sentence.

Alphabet’s results offered a more comfortable narrative, with Google Cloud growing 63% year-on-year to $20 billion in Q1, giving investors a cleaner line between the infrastructure spend and the revenue. Microsoft’s shares held steady for similar reasons. Meta’s problem is that its AI investment case is less clear in the numbers right now. Zuckerberg has been clear that the company is betting on AI across advertising, assistant products and the metaverse long game, but the return on $725 billion in spending across four companies has yet to appear visibly on any income statement.

Why Record Profits Weren’t Enough

The market reaction to Meta’s results tells you something specific.

Analysts have started using the phrase “capital misallocation” in commentary around these numbers, which is the polite way of asking whether anyone has done the maths on whether AI infrastructure on investment of this magnitude will generate returns corresponding with the investment. The straightforward answer, at this stage, is that nobody knows and the timelines involved are long enough with no definitive answer available for several years.

What the markets appear to be pricing in is the risk that the AI infrastructure buildout outpaces near-term monetisation. The positive outlook, which Alphabet’s Cloud numbers support, is that revenue growth catches up with spending and the margins eventually justify the bet. The risk-adjusted scenario is that four companies have collectively committed to spending that requires the AI economy to grow in ways that aren’t yet guaranteed, and that the bill will need to come from somewhere.

So What Does $725 Billion Mean For You?

For anyone building on AI infrastructure from AWS, Google Cloud or Azure, the $725 billion number signals something specific about pricing that founders should understand.
When hyperscalers are spending at this scale and the market is nervous about returns, cloud pricing is one of the levers available to them. Models are currently cheap because competition between providers is fierce and the priority has been adoption over margin. That calculation can change.

The more important point is that the infrastructure being built right now will be the foundation that most AI products run on for the next decade. The companies spending $725 billion are making a very large wager that AI-native products will become the default across enterprise software, consumer applications and hardware. For anyone building in that direction, the infrastructure bet is working in their favour. The question is whether the pricing environment that currently makes building affordable stays that way as the hyperscalers look to recoup their investment.

Record profits did not calm the markets because investors are doing the same maths that every founder building on Big Tech’s cloud infrastructure should be doing right now: $725 billion has to come back from somewhere, and nobody has fully answered where yet.