Bain Says AI Must Create $4.2 Trillion in New Revenue to Support Data Center Buildout

The global AI industry must generate $6 trillion in annual revenue by 2031 to justify the capital committed to data centers, Bain & Company says. Existing consumer and enterprise AI services could contribute up to $1.8 trillion, leaving $4.2 trillion that must come from new products, markets and applications.

The figure turns the data center boom into a financial test as well as an infrastructure race. Billions are flowing into chips, power and facilities, while AI providers still face questions about how quickly usage can become durable revenue. Bain’s 7th Global Technology Report frames the challenge around economic value created beyond productivity gains from current software.

The $4.2 trillion gap is expected to come from applications that expand AI into new industries. Autonomous machines, robotics, drug discovery and other forms of AI-driven product development are among the areas identified as possible sources of growth.

Consumer subscriptions, advertising and enterprise software remain important parts of the existing market, but Bain says they won't be enough on their own. New revenue must come from products that create fresh demand and expand the economy, rather than simply making current office tasks faster. The National’s coverage similarly describes new product development as the biggest contributor to the additional revenue target.

Bain estimates that data center construction could require $5 trillion to $6.5 trillion in investment and add roughly 150 gigawatts or more of capacity by 2030. That would nearly triple global data center capacity over 5 years, with facilities and project costs doubling on a 12-to-16-month cycle, according to The National’s account of the report.

The buildout is running into several constraints at once: electricity, advanced chips, skilled workers and permits. Capacity decisions made today affect supply years later, so delays in power connections or construction can limit AI deployment even when demand is strong. Bain says the scale requires system-level responses, including new power and computing technologies, public-private coordination and shared resources.

The revenue target matters because data centers carry large fixed costs before AI products generate reliable cash flow. Operators must pay for land, energy, networking, servers and increasingly scarce accelerators, while model providers and customers absorb computing costs through subscriptions, usage fees or software contracts.

Productivity improvements can support demand, but Bain’s analysis places the decisive burden on innovation that creates entirely new value. That shifts attention from model capability alone to business models, deployment economics and whether AI can support high-value activity in physical industries. Robotics and autonomous systems have a larger revenue ceiling than many software features, but they also face manufacturing, safety and adoption hurdles.

AI companies face pressure to turn infrastructure commitments into products with measurable revenue. Providers that depend on a narrow set of enterprise copilots or consumer subscriptions will need broader applications, stronger customer retention and pricing that covers rising compute costs.

The same pressure affects the broader technology supply chain. Chip companies, cloud providers, utilities, data center developers and equipment suppliers are all investing against expected AI demand. If new applications arrive quickly, those investments support expansion. If revenue growth falls short, customers can delay capacity commitments, leaving expensive facilities and hardware competing for fewer workloads. Bain’s report therefore links software innovation directly to infrastructure returns.

The next test is whether AI adoption produces revenue at the scale required by the infrastructure pipeline. Existing services can provide a substantial base, but the industry needs new markets to close the $4.2 trillion gap by 2031.

Power projects, chip supply, permitting and labor will shape how fast capacity comes online. At the same time, companies will have to prove that robotics, autonomous machines, drug discovery and other emerging applications can move from investment themes to large commercial businesses. The outcome will determine whether today’s data center expansion becomes durable economic infrastructure or an overbuilt capacity cycle.