Bain says AI needs $6 trillion by 2031

On October 2, IT Home relayed a Bloomberg report from September 29. Bain says that, to support the current wave of capital spending on data centers, the global AI industry needs to generate $6 trillion of revenue a year by 2031. The report says existing consumer and enterprise AI services can contribute at most $1.8 trillion, so another $4.2 trillion of new revenue still has to be found.

That new revenue is described as something that might come from markets still getting started, including automated machines, robots, drug development, mental health, and energy production. “Might come from” is not a signed contract.

David Crawford, the report’s lead author and chair of Bain’s global technology, media, and telecommunications practice, said the industry needs a wave of innovation far larger than mobile internet and cloud computing. He said AI infrastructure construction is running far ahead of demand, and that annual global GDP growth would need to be about 1 percent higher for the investment to keep going. That is a judgment in Bain’s report, not a growth target published by a statistics office.

Bain also says the industry still has to clear several obstacles if it is to keep the current pace of AI development.

Microsoft, Google, Amazon, Meta, Oracle, and other companies are putting trillions of dollars into data centers to meet demand for compute. The report says the scale and construction cost of data centers roughly double every 12 to 16 months. Higher chip prices at Nvidia, SK Hynix, and others, plus higher prices for networking gear and other parts, are pushing costs up.

The report also describes a growing argument about when AI service providers will see returns that match the spending. Critics worry that the ties of interest and supply between equipment makers and AI developers are becoming more complicated, lifting market expectations, which then require still more capital. Bain says attention is now mainly on how far AI can raise worker productivity, but that those gains alone cannot carry the infrastructure spending. The industry also has to create trillions of dollars of new revenue.

Bain expects global data-center investment to reach $5 trillion to $6.5 trillion by 2030, with at least 150 gigawatts of new capacity, adding further pressure on national energy supply. By 2031, annual spending on AI infrastructure — data-center construction, added compute, and upgrades to AI accelerators and memory chips — could reach as much as $1.5 trillion.

Developers are also constrained by shortages of transformers, water, and electricity, and by strong local opposition. The report says that in the United States, in the quarter ending in June, data-center projects halted or delayed for these reasons were worth $68 billion. We have not seen Bain’s original report. The figures above are from this retelling, and the forecasts are not booked revenue.