MS Summer of AI Financing

M August 16, 2026 06:00 AM GMT

Sunday Start | What's Next in Global Macro Morgan Stanley & Co. LLC Global Idea

Vishwanath Tirupattur Strategist

The summer of 2026 may ultimately be remembered not for a new model release or a breakthrough chip, but for developments in AI financing that highlighted how quickly capital markets are adapting to the demands of the AI buildout. In just a few months, hyperscalers materially increased already ambitious capex plans, public and private credit markets took on a larger role in funding AI infrastructure, investors differentiated more sharply among borrowers and business models across the AI financing ecosystem, and financing structures evolved rapidly, extending deeper into the value chain and increasingly targeting individual components, particularly chips. None of this was wholly unexpected. The scale of AI investment has long suggested that traditional funding channels alone would prove insufficient. What has been remarkable is the speed, breadth, and creativity with which markets have responded. Yet beneath the headlines lie important lessons for investors as the next phase of the AI investment cycle unfolds.

The starting point remains unchanged: demand for compute continues to outstrip supply, resulting in upward revisions in AI infrastructure capex expectations as hyperscalers commit additional capital to secure future capacity. Our equity research colleagues now estimate that total capex for the four largest hyperscalers (Microsoft, Alphabet, Amazon, and Meta) will rise 57% in 2027 versus 2026. These spending plans reflect growing conviction that such investments can generate 25%+ returns on invested capital (see Morgan Stanley Internet analyst Brian Nowak's report, 'The Paths to 25-50% GenAI ROIC'). At the same time, the lag between capex deployment and monetization continues to pressure near-term cash generation, with our analysts' 2027 free cash flow estimates for the four hyperscalers continuing to move lower. The result is a widening financing gap in 2027, reinforcing our view that AI-related credit issuance will remain substantial and may need to increase further before cash flows from these investments begin to catch up.

Developments in credit spreads this summer have been equally telling. Credit spreads for hyperscalers widened meaningfully, at one point reaching roughly 35bp wider on the year for higher-quality issuers and ~50bp wider for lower-rated names, before rallying sharply in the last two weeks. More notable than the absolute move, however, was the divergence across financing channels. Spread widening was most pronounced in higher-quality unsecured bonds, where issuance volumes accelerated sharply and investors remain exposed to a broader range of risks tied to the AI investment cycle. By contrast, spread widening in data center ABS and CMBS was Morgan Stanley does and seeks to do business with more modest. These structures are backed by operating assets that have already companies covered in Morgan Stanley Research. As a result, investors should be aware that the firm may have a conflict of been constructed, powered, and leased, with contractual cash flows largely interest that could affect the objectivity of Morgan Stanley established. Combined with a more measured pace of issuance, these characteristics Research. Investors should consider Morgan Stanley Research as only a single factor in making their investment helped insulate securitized credit from the volatility seen in unsecured markets. decision. For analyst certification and other important disclosures, The divergence across credit markets also reflects differences in issuer incentives refer to the Disclosure Section, located at the end of this and sensitivity to funding costs, which will shape issuance volumes going forward. report.

At the higher end of the quality spectrum, the major hyperscalers, with average ratings of roughly AA, combine substantial financing needs with significant ratings flexibility. We would also include leading semiconductor companies such as Nvidia and Broadcom in this group. Given their ROIC expectations, these issuers are relatively insensitive to modest changes in borrowing…