The math doesn’t add up.
More than $5 trillion will be invested in data centers and artificial intelligence (AI) infrastructure worldwide over the next five years but achieving a 10% return on the spending would demand approximately $650 billion in annual revenue, according to JPMorgan Chase.
The investment bank’s report reveals that achieving such a return – which represents roughly 0.58% of global GDP – is the equivalent to every iPhone user paying an additional $34.72 monthly fee or every Netflix Inc. subscriber contributing $180 per month.
Annual data center funding needs are expected to surge from approximately $700 billion in 2026 to more than $1.4 trillion by 2030. While hyperscalers currently generate more than $700 billion in operating cash flow annually and reinvest around $500 billion into capital expenditures, this won’t be sufficient.
The analysis projects leveraged finance will contribute roughly $150 billion over five years, with data center securitizations adding up to $200 billion. High-grade bond markets could deliver $300 billion in AI and data center funding in the coming year alone, totaling $1.5 trillion over the five-year period.
Even with these varied sources, JPMorgan identifies a $1.4 trillion funding gap that will likely require private credit and potentially increased government support, particularly if national defense concerns surrounding AI intensify.
However, the bank emphasizes that consumers won’t bear the cost directly.
Corporations and institutions, instead, are expected to shoulder the bulk of the cost through productivity gains from AI-assisted automation, data analysis, and enhanced decision-making across healthcare, finance, and manufacturing, JPMorgan concluded.
JPMorgan’s base case projects 122 gigawatts of data center capacity to be constructed between 2026 and 2030, driven by what the report describes as “astronomical” demand for computing power. Yet power constraints present a significant bottleneck. Current lead times for natural gas turbines have stretched to three or four years, while nuclear plants historically require over a decade to build.
“Adding 150GW of power in a timely manner is a remarkable challenge, particularly in light of grid upgrade requirements,” the report said.
JPMorgan’s report asserts the AI market will continue to enjoy sustained expansion and is not about to burst in a hype-driven bubble as predicted by bear market observers like Michael Burry, the American hedge fund manager who recently de-registered his fund after predicting an AI meltdown. (The report did not name Burry directly.)
Despite its optimistic projections, JPMorgan does not assume a straight-upward trajectory and explicitly raises concerns about repeating the telecom and fiber build-out experience, where revenue failed to justify continued investment.
“The path from here to there will not just be ‘up and to the right,'” the report warned, noting that unexpected breakthroughs could trigger overcapacity in which billion-dollar data centers sit idle because of insufficient demand.
The report also acknowledged the high-stakes nature of the AI ecosystem: Even if the build-out succeeds overall, it cautioned, the “winner-takes-all nature of portions of the AI ecosystem” will likely produce both spectacular winners and equally spectacular losers based on the massive capital involved.
While companies like OpenAI have reportedly reached $20 billion in annualized revenue and Anthropic targets $26 billion by 2026, these figures remain unconfirmed and don’t yet translate to net profits. OpenAI is part of the $500 billion Stargate Project with partners Oracle Corp. and SoftBank Group; Anthropic recently vowed to splurge $50 billion on an AI infrastructure plan for data centers in Texas and New York.
