The AI infrastructure boom is expected to attract an estimated $10.3 trillion in investment through 2032, according to Axios’s account of new research. The crucial point is that the buildout is no longer just a technology-sector wager: it is beginning to compete with governments, businesses and households for borrowed money.
Watch Desk analysis
What happened
The research, from Columbia Business School professor Stijn Van Nieuwerburgh and presented at the Brookings Papers on Economic Activity, estimates that AI infrastructure investment could equal 3.6% of GDP each year. Axios reports that Morgan Stanley expects Big Tech to need roughly $2.9 trillion to expand computing capacity through 2028, with more than half coming from outside investors.
That financing is already becoming expensive. Axios says Meta’s $30 billion Hyperion data centre is being financed largely through outside investors, with borrowing costs at least one percentage point above what Meta might have paid on its own debt. The paper estimates that this could add more than $5 billion to the project’s costs over its lifetime.
The numbers
- $10.3tn investment through 2032
The estimated scale of the AI infrastructure buildout. - 3.6% of GDP annually
The share of economic output the investment could represent. - $2.9tn through 2028
Morgan Stanley’s estimate for Big Tech’s computing expansion. - $3.7tn annual revenue required
The infrastructure’s estimated revenue target to generate a 10% return within six years.
Why it matters
The practical consequence is a wider financial risk than the familiar question of whether AI companies can sell enough software. If data centres depend increasingly on bond markets, banks, private credit, pension funds and sovereign wealth funds, disappointing demand could spread losses well beyond the companies building them.
Van Nieuwerburgh told reporters that outside capital may allow the buildout to race ahead of actual demand, creating the familiar possibility of oversupply followed by falling prices. His comparison is not subtle: he said the AI buildout is roughly 50% larger than the second-biggest investment boom, the construction of America’s railways.
Our read
This is the part of the AI story that does not fit neatly into a product launch or a triumphant benchmark. The servers are real, but so is the bill, and the bill is increasingly being shared with investors who may not have much say over the assumptions behind it.
The useful question for readers is not whether AI is “the next railroad”. It is whether the revenue arrives before the debt does. Markets have a long memory for infrastructure that was meant to transform everything and instead transformed the write-down column.
What to watch
- Whether hyperscalers keep borrowing at higher rates despite tighter financial conditions.
- How quickly AI revenue grows relative to new data-centre capacity.
- Which pension funds, private-credit lenders and other investors carry the exposure.
- Whether data-centre oversupply begins to push down prices or delay planned projects. Sources and evidence: Axios, reporting on research by Columbia Business School professor Stijn Van Nieuwerburgh presented at the Brookings Papers on Economic Activity; the supplied report also cites Morgan Stanley estimates and Meta’s Hyperion financing as examples.
Discussion spark: Should investors and governments slow the AI infrastructure race until revenue catches up with the debt, or is large-scale overbuilding the unavoidable price of securing technological leadership?
Sources and evidence
- AI's never-before-seen capital grab (24 September 2026, 17:31 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.