Big tech’s fat profits conceal unsettling cashflows

Arguments dismissive of the scale of big tech’s transformation have collapsed under the weight of the growing bill. Comparisons to the dotcom bubble are wrongheaded because the big spenders today generate ample cashflows, went one argument. Not any more. Their cashflow pressures cannot be that great because firms are still buying back bucket-loads of their own stock, many said. During the most recent quarter, buy-backs collapsed. A third is that big tech trades “only” at 23 times the firms’ forecast earnings. Yes, but when the denominator of this equation captures almost nothing of their spending, is it at all useful?

Nowadays investors judge the success of these firms on the basis of concentrated revenue contracts stretching far into the future, rather than dispersed sales received today. Mostly these contracts involve selling computing capacity to model-makers like OpenAI and Anthropic, which are themselves incinerating vast piles of cash. Total future revenue agreements have risen to $2trn, from $730bn last year, at Amazon, Google, Microsoft and Oracle (Meta is a buyer, rather than a seller, of computing capacity).

Simple balance-sheets with intangible assets and generous cash buffers have been replaced by ones which are complicated, asset-heavy and indebted. Since the start of last year the big five have raised $260bn from bond markets, a quarter of all such borrowing by listed American non-financial firms. What started as a local affair has become a global bacchanal. Nearly a third of the haul from selling bonds this year is in currencies other than the dollar. Alphabet, Google’s parent, will soon issue its first bonds denominated in yen.

Much larger obligations lurk off-balance-sheet. The biggest are $820bn of future payments to lease data centres yet to be built, up from $270bn a year ago. Commitments to spend money on other things, like packing their data centres with chips, have risen as fast. Amazon, Google, Meta and Oracle now disclose $680bn of such obligations. Other bills are tied to special-purpose vehicles: separate entities with their own balance-sheets. Last year one assembled to build Meta’s new data-centre in Louisiana issued the biggest single corporate bond in history. Oracle’s finance chief recently talked about “uncoupling” the firm’s cashflows from its capex, presumably with similarly advanced financial engineering.

This vast nexus of AI contracts combines an absolute faith in technologists with a naive trust in lawyers. Occasionally the market is discerning about what these contracts really mean; Oracle’s shares have been hammered since investors realised how dependent its future revenue is on OpenAI. More often the market is not. Bankers increasingly whisper about decaying documentation in AI financing agreements. “When we ask our lawyers to find ways that a hyperscaler might wriggle away from or re-negotiate a lease contract, often they come back with a very long list,” says the boss of one big lender that has steered clear of some more esoteric financing structures in the AI boom.

So far Silicon Valley’s capital spending has been a great act of charity to America’s tech industry. The firms have assumed the role of central planners, attempting to make the complex chain of returns on investment work across the AI economy: data-centres are useless if businesses don’t find models useful enough to pay for, and model-makers cannot raise enough capital to make them.

In the process, the hyperscalers have sacrificed their own returns. Only shares in Alphabet have beaten the NASDAQ index during the past year. Big tech has also liberally lent its creditworthiness across capital markets. Many firms that contract with the giants can take those contracts to the bank (literally) and raise more debt. Moreover, the hyperscalers’ capex has become someone else’s free cashflow. Broadcom, Micron, Nvidia and Sandisk, four chip companies, are all minting real fortunes outfitting big tech’s data centres. Together they account for a quarter of the expected profit growth in the S&P 500 index this year.

Clearly this is unsustainable without enterprises becoming much more willing to pay for AI. But for now there are no brakes on the train. The hyperscalers’ capex bills this year will be twice as great as analysts predicted they would be a year ago. If AI models keep getting hungrier for computing power and the cost of equipment keeps rising, this forecast will soon fade into the distance, as those that came before it did. After two years of consistent shock-and-awe, nothing would be less shocking.