Big Tech spending and the AI bubble

Why does Big Tech capex decide the pace of the AI bubble?

Capital spending, or capex, is money invested in data centres, servers, networks and land. The largest cloud and platform companies are the biggest buyers of AI hardware, so each budget decision flows to chip makers, builders, power suppliers and software vendors. A change of a few percent in their plans moves the whole chain.

That makes capex the most useful single number for anyone who tracks the AI bubble. It is public, it is reported every quarter, and it comes with company guidance for the year ahead. Companies state their plans in results and calls, so you can compare words with later numbers. A plan that is raised quarter after quarter shows confidence, and a plan that is held or trimmed shows caution.

How do you read Big Tech capex without getting lost in the figures?

Use 3 comparisons. First, capex against operating cash flow: a ratio near the total means little room for error. Second, capex against its own history: a steep rise needs a matching rise in revenue. Third, capex against depreciation, since new equipment loses value over a few years and its cost enters profit gradually.

Then ask what the equipment is for. Capacity built for the company’s own services behaves differently from capacity rented to outside customers. The first group of customers is inside the company, and the second group signs contracts. Both can be sound, and each has a different failure mode if demand falls. Note the split in the company’s own disclosures, because it tells you who carries the risk of unused capacity.

Which returns would justify Big Tech spending on AI?

Spending needs revenue from AI products, from cloud contracts, from better advertising results, or from lower costs. The EQOS Asia AI study records 95% of AI projects failing in production (MIT 2025) and about $200B a year invested (April 2026). Those figures describe the market: they explain why investors ask when returns arrive.

Look for evidence in 3 places: AI revenue named in results, cloud backlog from signed contracts, and customer usage that grows each quarter. A company that reports these items gives you something to measure. A company that reports only intentions leaves the question open, and the open question weighs on the share price. Write down the three items for each company at the same date, so the comparison stays fair.

How do Microsoft and Alphabet differ on AI spending and risk?

Microsoft sells cloud services and business software, so AI spending can return through subscriptions and cloud contracts. Our page on Microsoft and the AI bubble looks at spending, returns and the ties to model builders. Its exposure rests on how quickly business customers pay for the AI add-ons. Watch cloud growth and the share of customers who adopt the paid tools.

Alphabet runs search, video and a large cloud business, and it designs its own chips. Read Alphabet and the AI bubble for the way AI affects search income and cloud growth together. Its question is whether AI strengthens or reduces the profit of its core product. Watch search revenue, cloud growth and the cost per query over several quarters.

What separates Amazon and Meta from the other AI spenders?

Amazon earns AI revenue mainly by renting computing capacity through its cloud arm, and it also uses AI inside shopping and logistics. The guide on Amazon and the AI bubble explains why its returns depend on customer contracts for capacity. Contracts give visibility, and they also carry the duty to deliver that capacity. Watch backlog growth and how quickly new capacity turns into billed revenue.

Meta spends heavily and sells no cloud service, so its returns must come from better advertising, engagement and products. The page on Meta and the AI bubble sets out how that single path concentrates the question. Advertising income funds the spending today, which supports the company through a slow period. Watch advertising growth against the rise in spending, since the gap between the two is the exposure.

Why is Oracle’s AI spending a different kind of bubble risk?

Oracle supplies databases and cloud infrastructure, and the lens for its page is how much of new data centre building borrowing funds. Debt changes the risk profile, because interest payments continue whatever customers do. Our review of Oracle and the AI bubble explains how to test the funding side first.

The test is simple. Compare the cash that signed customer contracts will bring with the interest and repayments due. A company that matches them has a buffer, and one that lags depends on lenders staying calm. Check the maturity dates of the debt too, because a long schedule gives time for revenue to arrive. This is the clearest way to separate funded growth from growth that needs ever more outside money.

What happens to suppliers if Big Tech trims AI spending?

Suppliers feel a cut before the giants do. Chip designers, memory makers, builders and power firms book orders ahead, so a smaller budget reduces their future sales within a few quarters. Companies with only a few buyers feel it most, which is why the supplier chain appears on the map of exposed companies.

The EQOS Asia method page explains how a score frames such a shock, and the Λ simulator lets you test your own supplier or customer case. This page is general information, not investment advice. A budget cut is one scenario among several, so track the plans each quarter and update your view with the new numbers.

Frequently asked questions

About the author

Stéphane Jambu

President, EQOS Forecast Asia

Stéphane Jambu leads EQOS Forecast Asia and works as a consultant in e-commerce, search visibility, and AI. LinkedIn profile