Hyperscaler AI capex to surge 57% in 2027, straining credit markets

The AI buildout is becoming a capital story, with Morgan Stanley projecting hyperscaler capex to jump 57% in 2027 and widen a financing gap that will test credit markets.

Morgan Stanley expects Microsoft, Alphabet, Amazon and Meta to lift combined capital spending 57% in 2027 from 2026, pushing AI financing deeper into credit markets as free cash flow trails the buildout.

"The scale of AI investment has long made clear that traditional financing channels alone will not suffice. What is truly impressive is the speed, breadth and creativity of the market's response," Vishwanath Tirupattur, chief fixed income strategist at Morgan Stanley, said in an Aug. 17 report.

The four hyperscalers, which carry average credit ratings around AA, expect AI investments to generate returns on invested capital above 25%, according to Morgan Stanley internet analyst Brian Nowak. Cash goes out up front while revenue arrives later, pushing 2027 free cash flow forecasts lower and widening a financing gap that will require more credit issuance. Spreads on hyperscaler debt widened roughly 35 basis points for high-grade issuers and about 50 basis points for lower-rated names during the summer before rallying in the final two weeks.

The next phase of AI funding is shifting from data center shells toward compute equipment and energy assets, with Nvidia launching a compute financing platform and a $35 billion Broadcom-backed chip deal emerging. The question for investors is whether returns stay above Morgan Stanley's 25% threshold as depreciation, power costs and financing needs climb.

Financing gap widens as cash flow trails capex

The timing mismatch is the core problem. AI infrastructure demands cash up front, while revenue and free cash flow arrive later. Morgan Stanley's 2027 free cash flow forecasts for the four hyperscalers have continued to move lower, creating what analysts describe as a widening financing gap that will require AI-related credit issuance to increase substantially before operating cash flows catch up with capital deployment.

Credit markets are already responding. The widening was most pronounced in unsecured bonds, where new issuance volumes climbed significantly. Data center asset-backed securities and commercial mortgage-backed securities experienced more modest spread movement, supported by operating assets with established contractual cash flows.

The major hyperscalers, along with semiconductor suppliers such as Nvidia and Broadcom, are relatively insensitive to modest changes in borrowing costs given their elevated return expectations. Lower-rated issuers face a different reality. Oracle, rated mid to low BBB across rating agencies, along with independent data center developers, REITs and former bitcoin miners, have less balance-sheet flexibility and a lower tolerance for rising funding costs.

Chip financing moves to the front line

The next phase of AI financing is expected to extend well beyond buildings. Morgan Stanley anticipates that private capital and asset-backed structures will increasingly fund servers, chips and energy infrastructure. Nvidia has announced a compute infrastructure financing platform, and a $35 billion chip financing transaction backed by Broadcom has emerged.

Nvidia has signed memorandums of understanding with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to raise more than $500 billion in third-party capital for AI infrastructure, according to Reuters. Chief Executive Jensen Huang said the company has the option to backstop up to $125 billion, or 25% of potential deals, likening Nvidia's chips to "revenue-generating assets." Apollo and Blackstone are financing a $35 billion expansion of AI computing capacity for Anthropic using Broadcom's custom chips.

Bank of America analyst Tom Curcuruto estimates Broadcom's chip-financing vehicle could grow to $370 billion of senior debt by mid-2029 to fund 20 gigawatts of compute, implying around $150 billion of net new supply in 2027 alone. Meta struck a $27 billion financing deal with Blue Owl Capital to fund its biggest data-center project.

Yet the financing wave is drawing scrutiny. CreditSights analysts wrote that Nvidia is effectively "writing a put," a structure that is "pro-cyclical and exacerbates boom-bust potential." Moody's Ratings warned that a substantial increase in Broadcom's contingent obligations could limit its financial flexibility, while S&P Global Ratings said it will treat residual value support as a "contingent debt-like obligation" added to adjusted debt.

A particular source of unease is circular financing, where AI suppliers help fund the same customers that purchase their hardware. Nvidia's plan to invest up to $100 billion in OpenAI would flow back into data centers filled with Nvidia chips. Similar arrangements involving Oracle and OpenAI have added to fears that some AI demand may be artificially inflated rather than driven by genuine customer adoption.

For investors, the critical metric is the gap between AI capital expenditure and free cash flow. Microsoft, Amazon, Alphabet and Meta will need to demonstrate that AI revenue and utilization are rising quickly enough to offset accelerating depreciation, power costs and financing needs. Persistent credit spread widening, particularly for Oracle and leveraged infrastructure operators, could slow deployment or raise required returns.

The next hyperscaler earnings cycle will be judged on more than AI revenue growth. Investors should watch 2027 capex guidance, free cash flow revisions, financing activity and management commentary on AI returns. According to TipRanks data, analysts see the most upside in Oracle at 74%, followed by Nvidia at 36.6% and Meta at 32.4%, with Amazon offering 27.8%.

This article is for informational purposes only and does not constitute investment advice.