The companies building the AI future are about to spend money faster than they make it. A lot faster.
Morgan Stanley predicts that hyperscalers and neoclouds will post their most negative free cash flow in 2027, with cash burn approaching $600 billion.
The spending math behind the forecast
The projection comes from an October 15, 2026 Morgan Stanley report on AI-related capital expenditures. It covers the hyperscalers, meaning Microsoft, Amazon, Alphabet/Google and Meta, along with neocloud providers.
The bank forecasts total AI compute investment of approximately $1.58 trillion in 2027. That figure sits about 14% above a baseline cash capex estimate of around $1.388 trillion.
The difference works out to $195 billion. According to the research findings, that increase reflects lease financing obligations layered on top of direct cash spending.
The trajectory is steep on both sides of 2027. The bank pegs AI investment at about ~$980 billion in 2026 and about ~$1.81 trillion in 2028.
So 2027 is not the year spending tops out. It is the year the gap between spending and incoming cash is expected to be widest.
Why 2027 is the pain point
Morgan Stanley’s view is that infrastructure buildout will push capex up faster than operating cash flow can keep pace in 2027.
The report also points to a cumulative shortfall of roughly ~$600 billion for the top four hyperscalers through 2028.
They already are. Hyperscalers had issued approximately $255 billion in unsecured bonds by mid-October 2026, according to the research.
That equals about 30% of projected full-year capex.
Who can afford the bill, and who might struggle
High-rated hyperscalers are expected to keep strong borrowing capacity even as yields rise. The report cites Amazon and Google as examples of firms with an interest coverage ratio of around 30.7x.
Notably, the bank says the cushion holds even if these firms take on a potential $400 billion in new debt in 2027.
Lower-rated entities face a tougher road. Morgan Stanley expects they could run into stricter constraints when financing their operations.
Nvidia’s slice of the pie
Morgan Stanley expects Nvidia to collect about $700 billion from the overall wave of AI investment.
What this means
Bond issuance becomes a key signal to watch. The $255 billion already raised this year suggests the debt market has become a core funding channel for AI infrastructure, not an occasional supplement.
The research also flags the possibility of heightened volatility across AI-related stocks.
Demand for semiconductors and server hardware stands to benefit if the spending materializes as projected.
Morgan Stanley’s numbers show spending still climbing into 2028, at about ~$1.81 trillion.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.