Bain Says AI Infrastructure May Require $6 Trillion in Annual Revenue by 2031

Bain’s estimate covers yearly spending on facilities, processors, memory and networking. It assumes infrastructure would consume about 25% of sales each year. The consultancy calls that 25% assumption bold but fair because it matches what cloud computing providers spent on infrastructure in earlier years. Even so, the math leaves AI businesses needing far more commercial activity than their current products bring in.

The largest share of that revenue, about $4.2 trillion, would have to come from new products in search, advertising, autonomous systems and physical AI, many of which barely exist today. Enterprise productivity would add $1 trillion to $1.4 trillion as companies use AI for software development, sales, marketing, customer support and IT operations. Consumer subscriptions and advertising would contribute only $200 billion to $400 billion, even as providers push AI products to billions of users.

Adoption speed matters. Leading AI labs are spending more than $9.75 billion on engineering that helps companies put AI to work faster. “The economics of AI infrastructure demand trillions in new revenue beyond productivity gains,” said David Crawford, chairman of Bain’s global technology practice. Crawford adds that the industry needs a surge of ideas big enough to dwarf what mobile technology and cloud computing unlocked in earlier years. Bain says future products could reach into drug discovery, mental health and energy generation, areas where AI has limited presence today.

Data center expansion compounds the financial pressure. Bain says data center size and cost have been increasing at roughly double rates over periods lasting approximately 12 to 16 months globally. According to Epoch AI, Meta’s Prometheus facility in Ohio had 600MW of capacity and an estimated $24 billion cost in 2025. It could reach 2GW and $80 billion by 2027, before increasing to 5GW and $175 billion in 2029 under current projections. By 2030, Epoch AI estimates the project could reach 9GW while requiring as much as $200 billion in total spending.

Those facilities require additional electricity generation, grid connections, advanced semiconductors, skilled workers and equipment capable of supporting sustained operations at scale. The pace at which money is pouring into data centers has made securing enough computing capacity the industry’s most pressing concern, the report said. “But the more important question may be whether enough economic value can be created to justify it,” the report said.

Editor's Summary. Bain & Company estimates AI infrastructure spending may reach $1.5 trillion a year by 2031 and require about $6 trillion in annual revenue. It says most of that revenue would have to come from products that barely exist today, while data center costs and capacity are doubling roughly every 12 to 16 months. The report raises the question of whether enough economic value can be created to justify the investment.