AI Is Facing An $800B Supply-Demand Gap, Bain Study Finds
The world's corporate giants of artificial intelligence are facing a math problem that not even their increasingly advanced models can solve.
AI firms will need to generate $2T in annual revenue to fund the computing power needed to meet demand by 2030, but they will fall roughly $800B short, according to new research from Bain & Co.
Demand for AI compute power — the calculations happening on processors that fuel models like OpenAI’s ChatGPT — is growing twice as fast as chip efficiency, creating the need for more data center development.
Half of global demand for computing power could come from the U.S. by 2030, which Bain projects would require 100 gigawatts of electricity.
Tech companies, with the backing of the White House, are pouring hundreds of billions of dollars into data center development annually in a race to meet the projected need.
Up to 10% of technology spending in the U.S. could be directed towards AI capabilities in the next three to five years, with half of firms’ tech spending being used to run the AI systems, Bain estimates. It’s not nearly enough to meet demand, according to the global consulting firm’s projections.
If the major tech firms shifted their IT budgets to prioritize cloud computing and reinvested all of the projected savings from AI — which is forecast to impact teams in sales, marketing, customer support, and research and development — into supporting the technology, it would still need an additional $800B to support current growth rates, Bain found.
Consumer-facing tech firms continue to build new models and deploy new products, but none have found a way to make their AI tools profitable. OpenAI expected $5B in losses in 2024 on $3.7B in revenue, CNBC reported.
The Bain forecast assumes current levels of productivity growth and innovation in the AI space, but the consulting firm notes that further advances in the tech could shift the cost calculation. Technological leaps in the algorithms that underpin AI systems could improve, reducing the power needed to complete tasks.
When Chinese researchers released DeepSeek, an AI model that is more efficient than but comparable to major U.S. competitors, investors briefly pulled out of AI and chip stocks due to fears that the economic model had been upended. It hadn’t, at least not according to major private equity firms like Blackstone and Apollo Global Management, which are funding much of the AI arms race.
Some technologists expect breakthroughs in quantum computing to cut costs and reduce power needs. Bain suggests the market could grow to $250B, although the consulting firm believes the benefits are years away.
The forecasts for extraordinary growth in power needs come despite analysts and AI thought leaders acknowledging a bubble is forming in the sector.
A 2024 report from Goldman Sachs questioning whether the massive sums of cash that tech firms are pouring into AI would lead to comparable advances was one of the first quiet detractors.
Joe Tsai, the billionaire chairman of Chinese conglomerate Alibaba Group Holding Ltd., said a bubble was forming in the data center development space in March.
Sam Altman, the head of OpenAI and one of the leading public boosters of the tech, conceded last month that there was overinvestment in some segments of the AI ecosystem.
“When bubbles happen, smart people get overexcited about a kernel of truth,” Altman told a small group of reporters, CNBC reported.
