AI: a credit investor’s perspective on the data centre build-out

Technology companies are investing heavily in data centres, computer chips and the infrastructure needed to support AI development. Investors are eager to provide funding, allowing companies to raise large amounts of capital to support expansion.

Against this backdrop, the enormous build-out of data centres is in the spotlight. Capital expenditure from the top AI-linked tech firms, or hyperscalers, is forecast to increase from an estimated USD 833 billion this year to USD 1.3 trillion in 2027¹, creating substantial financing needs across public and private credit markets.

How should credit investors approach this booming sector?

For investors, the key lesson is to separate a great technology from a great investment. Some of the biggest opportunities can emerge from transformational innovations, but the winners are not always the companies attracting the most attention during the initial excitement. For example, the latest phase of the AI expansion has seen a broadening of the investment opportunity as businesses involved in sectors such as electrical equipment and connectivity enjoy surging demand. It is not only semiconductor companies powering the AI boom.

Yet with increased choice comes greater pressure to identify the right businesses. That is why it is important to focus on businesses with strong balance sheets, sustainable cash flows and resilient business models. While AI may reshape the economy over the coming decades, successful investing still requires discipline, careful analysis and a healthy respect for the lessons of history.

While we remain constructive on near-term demand for AI-related infrastructure, there is less line of sight into the long-term monetisation of AI and the sustainability of current investment levels. As a result, our approach emphasises assets with strong contractual protections and visible cash flows rather than taking direct exposure to long-dated AI growth assumptions.

Our preferred exposure is through:

• Large, diversified investment-grade data centre operators with strategic assets in major metropolitan markets.
• Select high yield data centre projects that are fully leased to leading hyperscalers under long-term contracts.

These investments provide exposure to strong counterparties while benefiting from contractual cash flow protection. In the high yield segment, we focus on projects that have secured long-term lease commitments from hyperscalers before completion, include protections that effectively prevent tenant termination without full repayment of debt obligations, and have clear visibility to completion and energisation. This allows for cash generation to be directed toward deleveraging (see Exhibit 1).

Our conviction is strongest over the next two years. Demand for high-quality data centre capacity during 2026-2027 is expected to remain exceptionally strong, supported by hyperscaler AI investment plans.

There are risks, of course. These including the possibility that construction delays postpone project completion, a potential slowdown in hyperscaler demand growth – particularly if AI adoption or monetisation falls short of expectations – and oversupply of data centre capacity over the longer term. At present, we view these risks as manageable given the scarcity of available capacity and the strategic importance of these assets to hyperscalers, but we continue to monitor the situation.

It is hard to ignore market volatility relating to AI. Recent weakness in both hyperscaler debt and equity markets has been driven by investor concerns over the magnitude of AI-related capital expenditure and its eventual return on investment.

However, we do not believe there has been a significant deterioration in underlying credit quality. We continue to view the largest hyperscalers as fundamentally strong credits with robust balance sheets. But from a relative value perspective, select contracted data centre projects may offer a more attractive risk/reward profile because they provide exposure to near-term infrastructure demand without relying on long-term assumptions about AI profitability.

We carefully watch for potential negative signals, which would include hyperscalers lowering long-term growth expectations, reduced commitment to future AI infrastructure spending, or increasing financing activity for speculative or uncontracted data centre projects. While we continue to see earnings growth and evidence of improving returns on AI investments, we believe our investment thesis is intact.

We remain positive on data centre credit exposure tied to contracted hyperscaler demand through 2026-2027. As in previous tech booms, timing is key, which is why we advocate an active approach balancing both risk and potential reward.

¹ Source: Bank of America, August 2026.