The State of AI Investment, October 2, 2026: Record Spending, Rising Worries
The money is still pouring in, but central banks and markets are asking harder questions about who pays for it.
The scale of the buildout
The four big hyperscalers (Amazon, Alphabet, Microsoft, and Meta) are guiding to roughly $730 billion in AI-driven capital spending in 2026, up about 78% from around $410 billion in 2025. Looking further out, UBS estimates that Amazon, Alphabet, and Microsoft will spend about as much on capex this year as their entire cloud revenue, and projects roughly $4.1 trillion of hyperscaler capex from 2026 through 2028. Yahoo Finance
The private labs are raising and delaying
The leading AI labs are chasing capital on a similar scale. OpenAI is seeking at least $30 billion at a valuation near $1.4 trillion after postponing its IPO beyond 2026. Anthropic has pushed its own listing from October to November, with investors expecting a valuation around $2 trillion and a raise of up to $100 billion. Reuters has reported that Anthropic’s confidential prospectus shows a $42 billion loss last year. OpenAI seeks $30 billion in funding at whopping $1.4 trillion valuation after delaying IPO +2
The concerns
Debt is doing more of the work. The Bank of England said this week that AI-related debt issuance has surged in 2026 and that high valuations depend heavily on future earnings and productivity gains. Australia’s central bank similarly flagged the AI boom as a key driver of potential financial instability in its October review, warning that highly valued AI firms could face sharp repricing if expectations aren’t met. cyberdaily
Circular financing. In late July, Nvidia fell nearly 5% on reports it was in talks to backstop $250 billion in funding for OpenAI, renewing worries about suppliers financing their own customers. Yahoo Finance
Rates and energy. July saw the Nasdaq fall more than 3%, long-term Treasury yields reach a 19-year high, and oil briefly cross $100. Higher borrowing costs matter more when so much of the buildout is financed with debt. vaquerowealth
Private credit stress. The private credit default rate stood at a record 6.0% for the twelve months ended May 2026, which matters because private lenders have become a major source of data-center funding. vaquerowealth
The counterargument
Not everyone sees a crisis. During the July selloff, hyperscaler credit spreads barely moved, since their balance sheets are strong enough that default risk is considered remote. One Forrester analyst described the chip selloff as a repricing of expectations after an exceptionally strong rally, not a sign of weakening AI demand. CNBC
What to watch
Three signals will matter most: whether AI revenue keeps pace with infrastructure costs, how the Anthropic IPO is received in November, and whether interest rates keep rising. The debate isn’t whether AI is useful. It’s whether prices and borrowing have run ahead of the profits.
