AI Spending Boom Shows Cracks as Big Tech Burns Cash

Direct answer: The AI spending boom has hit its first serious wobble. In the past two weeks Alphabet, Amazon, Meta and Microsoft outlined plans to spend more than $740 billion on capital expenditure this year, mostly read by Wall Street as AI infrastructure spending, and investors responded not with applause but with selling. The Nasdaq fell 2.9% on Friday, chip stocks dropped sharply, and Alphabet suffered its worst day on record the day before. The reason is simple arithmetic: the companies funding the AI build-out are now spending more cash than they generate, borrowing heavily to keep going, and still cannot show that the returns will arrive fast enough to justify the bill.

What happened

For most of the AI boom, big spending was treated as a virtue. When a chief executive announced another tens of billions for data centres and chips, the share price usually rose, because the market read it as confidence. That logic held while the spending was comfortably covered by the enormous cash flows of search advertising, e-commerce and cloud computing. This earnings season broke the pattern.

Alphabet provided the clearest example. Google’s parent company reported negative free cash flow of $9.1 billion in the third quarter, its second consecutive quarter in the red, even as quarterly capital expenditure grew to $51.4 billion. Alphabet expects to spend up to $205 billion this year, and expects that spending to keep rising in 2027. At the same time, the business paying for all of it is showing strain. Google’s search advertising business, which makes up roughly half of its revenue, saw growth slow to 12%, its weakest pace in five quarters. The culprit is partly AI itself: the wild success of AI Mode, Google’s AI-powered chatbot feature built directly into Search, is eating into the traditional search business that pays the bills. Google is spending billions on the technology that could undermine the business funding it, and investors noticed.

Amazon is in a similar position. The company reported negative free cash flow of $42.4 billion over the past twelve months, compared with $14.8 billion in positive free cash flow the year before, and says it expects to spend $220 billion on capital expenditure this year. Its cloud business, AWS, grew revenue by 39%, which in any normal quarter would be a cause for celebration. Instead Amazon’s stock tumbled 7% after it reported earnings on Thursday, because the market was focused on the cash burn rather than the growth.

Meta told a similar story. The company reported negative free cash flow of $2.1 billion in the third quarter as capital expenditure outpaced its operating cash, and it is guiding to as much as $145 billion of capital expenditure this year. Meta noted that Muse, its new free personal AI agent, is very compute-intensive to run, in the words of chief executive Mark Zuckerberg. Advertising revenue still grew 26%, yet no amount of good news saved Meta from a 10% share fall last week.

Microsoft was the exception that proved the rule. It recorded $18.6 billion in positive free cash flow in its latest quarter, with capital expenditure at $45.2 billion, less than the $50 billion it had projected. Microsoft said that was not an intentional pullback from AI spending, blaming delays in capacity from third-party providers. The market still punished the companies that rent out AI computing power, whose shares sank on the news, because slower Microsoft spending was read as a sign of cooling demand rather than discipline.

The market reaction was broad and brutal. The Nasdaq fell 2.9% on Friday. Nvidia dropped 4%, Broadcom more than 5% and Qualcomm 11%. Alphabet, which had already plummeted 12% on Thursday, its worst day on record, wiping out $500 billion in market value, fell again. Those are not the moves of a market that has stopped believing in AI. They are the moves of a market that has started asking what the AI build-out actually costs, who pays for it, and when it pays back.

How the AI spending boom actually works

To understand why this matters, it helps to be precise about what “AI spending” means. When Alphabet, Amazon, Meta and Microsoft talk about capital expenditure, they are mostly talking about physical things: data centres, the servers inside them, the specialised chips that train and run AI models, the networking gear that connects them, and the power infrastructure needed to run them. Training a frontier AI model can require tens of thousands of chips running for months. Running that model for millions of users, answering questions all day, requires a second, even larger fleet of machines. None of that is cheap, and none of it is a one-off purchase. Chips are replaced as better ones arrive, data centres need power contracts that run for years, and the whole estate has to be maintained.

In the early years of the boom, this spending was funded from profits. Google search advertising, Amazon’s e-commerce and cloud businesses, Meta’s advertising and Microsoft’s software and cloud all generated far more cash than they needed. That surplus paid for the data centres. What has changed is the scale. Capital expenditure has grown so fast that it now exceeds operating cash flow at several of these companies. The gap is being filled with borrowing. According to S&P Global Market Intelligence, Microsoft, Alphabet, Amazon, Meta and Oracle raised a combined $326 billion in debt in 2026, triple the amount raised last year. Debt is not inherently dangerous for companies of this size, but it changes the question investors ask. A company funding its AI programme from spare cash can afford to be patient. A company funding it from bond markets has to convince lenders, every year, that the returns will come.

There is a second mechanism at work, and it is subtler. AI does not just cost money to build; it changes the economics of the businesses paying for it. Google search is the textbook case. For twenty years, a search query produced a page of links and adverts, and Google was paid when users clicked. An AI answer, by contrast, gives the user the answer directly, which means fewer clicks, fewer adverts seen, and less revenue per query, even as each query costs far more to serve because it runs through an expensive AI model. That is the tightrope Google is walking: it must invest in AI answers to stay relevant, while those same answers erode the advertising model that funds the investment. Amazon faces a version of the same tension in retail, where AI shopping agents could one day bypass the storefront entirely, and Meta faces it in attention, where a free AI agent that answers questions directly competes with the feed that carries its adverts.

None of this means the spending is irrational. Demand for AI computing remains strong. Google, Amazon and Microsoft all saw gains in their cloud businesses this quarter, and companies across the economy are still rushing to adopt AI tools. The question investors are asking is not whether AI is real, but whether the current pace of spending is sustainable, and whether the companies doing the spending will capture enough of the value to justify it. History offers a cautionary parallel. Previous technology booms, from railways to fibre-optic broadband, produced infrastructure that the world still uses, but many of the companies that built it went bust or saw their shareholders wiped out, because they overbuilt and overpaid in the race to be first. The infrastructure survived; the investment returns did not always follow.

The numbers that spooked the market

These figures come from the companies’ own earnings reports as covered by TechCrunch, and they are worth reading slowly, because they explain the market mood better than any single share price. A company can lose money for a quarter while investing for the future, and investors will often tolerate it. What they tolerate less well is a pattern: three of the four largest technology companies in the world all burning cash at the same time, all guiding spending higher, and all relying on debt markets to bridge the gap, while the core businesses that made them rich show signs of strain.

Why Anthropic’s IPO just got harder

The wobble could not have arrived at a worse moment for Anthropic. The AI lab behind Claude filed to go public earlier this week, kicking off speculation about how one of the largest technology listings in years would affect the broader AI industry. Its future growth depends on securing enormous amounts of computing power, which in practice means buying it from, or partnering with, the very hyperscalers whose spending is now under scrutiny.

That context makes a separate report, covered in detail in our companion post on Anthropic’s expected credit rating, especially relevant. The Financial Times reported that Blackstone, the anchor investor in a recent $63 billion deal to finance Anthropic’s chips, told investors it expects the company to receive a high “junk” rating after its IPO, a reversal of its earlier expectation of an investment-grade rating before 2027. A junk rating does not mean a company is failing; it means lenders consider it riskier, and charge more for the privilege of lending to it. For a company whose business model requires vast, ongoing computing purchases, dearer borrowing is a real constraint. It also makes Anthropic more reliant on the hyperscalers and third parties to fund its infrastructure ambitions, exactly at the moment when those partners’ own spending is being questioned.

For smaller companies, the lesson is not about Anthropic’s share price. It is about concentration risk. If your business depends on one AI supplier, that supplier’s financing, pricing and strategy become your problem too. The spending boom made AI cheap and abundant for two years. If the funding environment tightens, the companies that built their products on the assumption that cheap AI would last forever will feel it first.

What it means for UK small businesses

It is tempting to read this as a story about American giants and their share prices, and to conclude that it has nothing to do with a fifty-person firm in Manchester, a trades business in Bristol, or a boutique agency in Edinburgh. That would be a mistake, because the spending boom is the reason AI tools became so cheap, so quickly, for small firms. The giants subsidised the price of AI while they raced for market share. If that subsidy narrows, small businesses will feel it in three practical ways: prices, product choices, and supplier stability.

First, prices. Many of the AI tools small firms use every day, from chatbots to transcription to image generation, are priced below their true computing cost, because the provider is funded by investors or by a parent company willing to absorb losses while it grows. That cannot last forever, and the earnings reaction is an early signal that the patience of capital markets is not infinite. A tool that costs £20 a month today could cost £40 next year, or move features that are currently included into a higher tier. The sensible response is not to stop using AI, but to avoid building a workflow that only makes sense at today’s price. Before adopting a tool, ask what the business case looks like if the price doubles. If the answer is that it collapses, you are not buying a tool; you are renting a subsidy.

Second, product churn. When funding tightens, AI companies consolidate, cancel products, and change direction. We have already seen this pattern in the sector: tools launched with fanfare are quietly retired, free tiers shrink, and start-ups that raised money on a promise are acquired for their teams rather than their products. For a small business, the cost of a tool disappearing is rarely the subscription fee. It is the retraining, the migration, and the week of lost productivity while staff learn a replacement. That is why the boring disciplines matter: keep your own data in formats you control, prefer tools that export cleanly, and avoid letting any single AI product become the only place a critical process lives. If your customer records, your quotes, or your content archive only exist inside one supplier’s system, you have a single point of failure, and the spending wobble is a reminder that even large suppliers change course.

Third, opportunity. A cooling boom is not all bad news for small firms. The first phase of the AI revolution rewarded the companies with the most capital. The next phase may reward the companies with the clearest use cases. When computing was almost free, it was easy to throw AI at every problem and call it strategy. When computing has a real price, the firms that win will be the ones that can point to a specific job, a specific saving, or a specific improvement in customer experience, and show the maths. Small businesses have an advantage here, because they are closer to their customers and can test a tool in a week, with real staff and real customers, rather than debating it in committee for a quarter. The firms that will struggle are the ones that bought AI as a badge, not as a solution to a named problem.

There is also a UK-specific angle worth stating plainly. British small businesses do not buy AI in a vacuum. They buy it under UK data protection law, with customers who care about where their data goes, and with energy and cost pressures that make every subscription a line item to justify. A US spending boom does not change those obligations, but it does change the market small firms buy into. If American providers raise prices or consolidate, UK alternatives and European providers may become more competitive, and the negotiating position of a small buyer improves when they can credibly walk away. The practical move is to keep at least two credible options for any AI capability that matters to your business, and to review your AI subscriptions quarterly, the way you would review any other supplier contract.

A practical checklist for the next quarter

The following checklist is deliberately unglamorous, because the right response to a market wobble is usually process, not prediction. Work through it with whoever owns your technology spending, even if that person is you.

Start by listing every AI tool your business pays for, what it actually does, and what it costs per month. Most small firms are surprised by this list. Subscriptions accumulate, free trials convert silently, and a tool adopted by one team member becomes a business dependency without anyone noticing. Cancel anything nobody can name a job for.

Next, for each remaining tool, write down the job it does in one sentence, and the evidence that it does it well. “Drafts replies faster” is a claim; “cuts first-response time from four hours to forty minutes, measured over the last month” is evidence. Keep the tools with evidence. Challenge the ones without it, and give them a deadline to prove themselves or go.

Then check your exit route for each critical tool. Can you export your data in a usable format? How long would it take to move to an alternative? If the answer is “we don’t know”, find out before you need to know. A thirty-minute export test now is worth a week of panic later.

After that, look at contract length. Monthly rolling terms are worth a small premium over annual deals when the market is volatile, because they preserve your ability to switch. If a supplier offers a large discount for a two-year commitment, ask what happens to pricing, features and support if that supplier is acquired or changes strategy. Get the answer in writing.

Finally, separate your AI experiments from your AI dependencies. Experiments should be cheap, reversible and time-boxed: try a tool for a month, measure it, and decide. Dependencies deserve the opposite treatment: a named owner, a backup option, and a budget that survives a price rise. Most small firms get into trouble by treating a dependency like an experiment, or an experiment like a dependency, and the spending boom encouraged both mistakes by making everything feel free.

Honest limits / what to watch

This story is, at heart, a market reaction to earnings reports, and market reactions are not verdicts. Share prices fall for many reasons, and a bad week does not prove that AI investment is wasted. It is entirely possible that the spending pays off handsomely over five or ten years, that AI revenues grow to dwarf today’s infrastructure bill, and that this quarter is remembered as a buying opportunity rather than a turning point. The honest position is that nobody knows yet, and anyone who claims certainty in either direction is selling something.

There are also limits to the data. Free cash flow is a useful measure, but it is affected by the timing of large payments, tax, and working capital, and a single quarter’s figure can mislead. Capital expenditure plans are intentions, not spending already incurred, and companies revise them constantly; Microsoft’s shortfall against its own projection is a reminder of that. The $326 billion debt figure covers five companies and a calendar year, and it does not by itself prove distress; these are among the most creditworthy borrowers in the world, and lenders are still eager to lend to them. The right reading is narrower: the funding mix is shifting, the market is paying attention, and the era in which AI spending was automatically rewarded appears to be over, at least for now.

What to watch next is straightforward. First, the next round of earnings: do Alphabet, Amazon and Meta return to positive free cash flow, or does the cash burn deepen? Second, capital expenditure guidance for 2027: do the giants hold their spending plans, trim them, or accelerate further? Third, Anthropic’s IPO and its eventual credit rating, which will test whether debt markets are still willing to fund AI growth at scale. Fourth, pricing moves from AI providers aimed at small businesses: price rises, tier changes and product retirements will tell you faster than any share price whether the subsidy era is ending. And fifth, cloud revenue growth at AWS, Azure and Google Cloud: if that growth holds up while spending cools, the boom may simply be maturing. If it fades too, the story becomes much more serious.

For a UK small business, the signal to act on is not the Nasdaq’s daily move. It is the direction of travel in your own supplier contracts. Watch your renewal notices, your per-seat prices, and the features that quietly move between tiers. Those are the numbers that will decide whether your AI bill rises next year, and they are the numbers you can actually control.

FAQs

What is the AI spending boom?

It is the surge in capital spending by Alphabet, Amazon, Meta, Microsoft and Oracle on data centres, chips and power to build and run AI systems, funded from profits and, increasingly, from debt. TechCrunch reports that the four largest of those companies outlined more than $740 billion of capital expenditure this year, which Wall Street reads overwhelmingly as AI infrastructure spending.

Why are investors suddenly nervous about AI spending?

Because the spending has grown so large that several of the world’s most profitable companies are now burning cash, while the returns from AI are still uncertain and the core businesses funding it, such as search advertising, are under pressure. Alphabet reported negative free cash flow of $9.1 billion in the third quarter, Amazon reported negative free cash flow of $42.4 billion over twelve months, and Meta reported negative free cash flow of $2.1 billion in the quarter.

Does the AI spending wobble mean AI is failing?

No. Demand for AI computing is still strong and cloud revenues are growing, with AWS up 39% in the latest reports covered by TechCrunch. The worry is about the cost of meeting that demand and how long it takes to earn a return, not about whether customers want AI. A wobble in funding is not the same as a collapse in demand.

What should a UK small business do about the AI spending boom?

Buy AI on outcomes, not hype. Start with one measurable job, prefer monthly tools you can cancel, avoid long contracts tied to a single supplier, and keep your data portable. Review every AI subscription quarterly, test your export route before you need it, and make sure any tool that matters to your business still makes sense if its price doubles.

Could AI prices rise if Big Tech cuts spending?

It is possible. If investment slows, computing capacity could tighten and providers could raise prices or ration access. That is another reason to avoid building a business that only works if today’s AI prices last forever. Keeping a second credible supplier for any critical AI capability is cheap insurance against a price shock.

Is this the end of the AI boom?

There is no evidence for that conclusion yet. Previous technology booms show that infrastructure can be both overbuilt and ultimately valuable, and that shareholders can suffer even when the technology succeeds. The honest answer is that the boom is entering a more sceptical phase, in which spending must be justified by returns rather than applauded on announcement. That is a healthier test for everyone, including small businesses choosing where to spend their own money.

Sources: TechCrunch, “Cracks are emerging in the AI spending boom” (30 October 2026), https://techcrunch.com/2026/10/30/cracks-are-emerging-in-the-ai-spending-boom/ ; figures for Alphabet, Amazon, Meta and Microsoft and the S&P Global Market Intelligence debt total are as reported there. Internal reading: our earlier coverage of AI data centre funding at https://weblifyai.co.uk/ai-data-centre-funding-33-billion-morgan-stanley/ and Amazon’s cloud scale at https://weblifyai.co.uk/aws-trillion-dollar-business-220bn-ai-spend/.

CTA: Get a Free AI Consultation