From a Technical Narrative to a Capital Narrative: The Next Main Battleground in the AI Construction Cycle

On August 17, Morgan Stanley's Chief Fixed Income Strategist Vishwanath Tirupattur pointed out in his latest report that the AI investment cycle is shifting from technology-driven to capital-driven.

Hyperscale cloud vendors have significantly increased their capital expenditure plans, with public and private credit markets taking on a larger share of AI infrastructure financing. Investors are starting to make finer distinctions between different borrowers and business models, while financing structures are rapidly iterating, extending deeper into the value chain and increasingly focusing on individual components such as chips.

Tirupattur wrote:

The scale of AI investment has long shown that traditional financing channels alone will struggle to keep up. What's truly impressive is the speed, breadth, and creativity of the market's response.

AI is no longer just a technology story—it's evolving into a capital markets story, and understanding capital flows is becoming as important as understanding technological innovation itself.

Financing Gap Continues to Widen, Credit Issuance Pressure Remains

Computing power supply continues to lag behind demand, prompting hyperscale cloud vendors to keep increasing capital expenditures to secure future capacity.

Morgan Stanley's equity research team recently estimated that the combined capital expenditures of four major hyperscale cloud vendors—Microsoft, Alphabet, Amazon, and Meta—will increase by 57% in 2027 compared to 2026.

Behind these spending plans is strong confidence in investment returns. In the report "Path to 25%-50% Returns on Generative AI Investment," Morgan Stanley Internet Analyst Brian Nowak noted that related investments are expected to achieve ROIC (Return on Invested Capital) of over 25%.

The issue, however, is that there is a significant time lag between the deployment of capital expenditures and monetization, resulting in near-term pressure on free cash flow. Morgan Stanley analysts have continued to revise down forecasts for the four largest hyperscale cloud vendors’ free cash flow for 2027. With capital expenditures rising and cash flow declining, the financing gap will further widen in 2027. Analysts believe that AI-related credit bond issuance will remain at high levels and may even need to increase further.

Credit Spread Divergence: Who Is Under Pressure and Who Can Withstand It

This summer's performance in the credit market revealed structural differences within the AI financing ecosystem.

AI-related credit spreads widened significantly: spreads for high-rated issuers expanded by about 35 basis points from the beginning of the year, while lower-rated names expanded by about 50 basis points, followed by a notable retracement in the most recent two weeks.

Even more noteworthy is the divergence between different financing channels.

High-rated unsecured bonds were hit hardest. The reason is the drastic increase in issuance volumes, and investors face broader risk exposure, which is directly tied to the overall uncertainty of the AI investment cycle.

Data Center ABS and CMBS are more resilient. These structured products are backed by operating assets that have been built, powered on, and leased out, with contract cash flow largely predetermined. In addition, a restrained issuance pace effectively isolates them from volatility in the unsecured market.

Tirupattur further distinguished behavioral differences between two types of issuers:

• High-rated end: Hyperscale cloud vendors such as Microsoft, Alphabet, Amazon, and Meta have an average rating around AA, with significant financing needs but ample rating flexibility. Leading semiconductor companies like Nvidia and Broadcom are also included in this category. Given their ROIC expectations, these issuers are not sensitive to small changes in financing costs—spread widening will not materially slow their pace of borrowing.

• Low-rated end: Such as Oracle (which receives mid-to-low BBB ratings from various rating agencies), as well as former bitcoin miners, REITs, and other data center developers. Their balance sheet flexibility is limited, making them more sensitive to financing costs. For such borrowers, spread widening is a material constraint—financing costs themselves become a natural stabilizer of supply.

Advancing the Frontier of Financing: From Data Centers to Chips

Analyst Tirupattur pointed out that the next phase of AI financing will see a clear structural shift.

The focus of incremental capital expenditures is shifting from data center shells to computing equipment (servers and chips) and energy assets. These assets naturally fit asset-level financing arrangements, creating more room for private capital participation.

Clear signals have recently emerged in the market:

• Nvidia announced the launch of a computing infrastructure financing platform;

• A Broadcom-backed $35 billion chip financing deal was implemented, setting a new record in size;

• The Beignet deal set for the end of 2025 as well as the recent Sopaipilla deal both represent innovative new directions in financing structures.

Tirupattur believes that the rise of large-scale component financing will largely depend on high-rated issuers leveraging their rating and balance sheet advantages—by providing backstop arrangements, credit enhancements, and residual value guarantees, they can help private capital underwrite increasingly large pools of AI infrastructure assets.

Capital Flows Will Be As Important As Technological Innovation

As AI evolves from a technology cycle to a capital cycle, understanding the details of financing is becoming more important than ever.

The AI financing ecosystem is expanding rapidly, but not uniformly. Different credit channels absorb different levels of risk, and issuer behavior will become increasingly differentiated, accelerated by differences in capital needs, rating flexibility, and sensitivity to financing costs. The ultimate result will be a more complex and more layered market, where financing outcomes will play a growing role in determining competitive outcomes.

As Tirupattur stated: "In the next stage of the AI buildout cycle, understanding the flow of capital is almost as important as understanding the flow of innovation. AI is no longer just a technology story—it is increasingly becoming a capital markets story."