AI disruption: a credit investor’s perspective

The boom in artificial intelligence (AI) is having a transformative effect on financial markets. From a credit investor’s perspective, we see two major dynamics. First, funding is required to build data centres, power and networks, and substantial amounts will come from investment grade credit. Second, the automatization of complex workflows, and the commoditisation of generic data and content, can reduce the need for intermediaries that previously captured margin through specialised tooling, distribution or information advantages. We see significant disruption caused by AI-facilitated automation in industries such as software – investors should take a thoughtful approach to issuer selection.

Although there are challenges to navigate, we believe these dynamics offer potential advantages for credit investors. AI matters insofar as it alters leverage paths, cash‑flow durability and the stability of industry structures that underpin ratings and recovery assumptions. We believe these disruptions offer opportunities for well-informed investors to benefit in this new era.

Aggregate funding needed for artificial intelligence and data centre projects is projected to reach USD 5.3 trillion over the next five years, with approximately USD 1.5 trillion anticipated to be raised through investment grade bond issuance (Exhibit 1).

Leading "hyperscaler" companies – those managing extensive, globally distributed cloud infrastructures – play a pivotal role in the ongoing AI buildout. According to projections from Goldman Sachs in November 2025, aggregate capital expenditures by the “Big 5” – Amazon, Google, Meta, Microsoft and Oracle – could approach USD 800 billion in 2026.

The funding capacity of the Big 5 remains strong. The leading hyperscalers collectively hold about USD 350 billion in liquid cash and investments and are expected to generate some USD 725 billion in operating cash flow in 2026. That means internal resources are likely to remain the largest funding source; however, external needs will grow. While these firms have increased their debt levels in recent years, net leverage remains low for most, providing significant capacity for additional borrowing (Exhibit 2).

The technical backdrop is favourable for the investment grade corporate bond market to absorb this predicted surge in data centre-related bond issuance without significant supply-driven dislocations. Although gross issuance may rise sharply, we do not expect a significant change in net issuance. This is because record maturities of about USD 1 trillion are coming due in 2026 and record coupon flows of about USD 500 billion are returning to investors. (These large maturities reflect the heavy issuance cycle of 2020–2021, while elevated coupon income reflects refinancing during 2022–2024’s higher-yield environment.) These inflows create substantial reinvestment demand, which should make it easier for the market to take in additional issuance with limited spread widening.

Every new AI capability expands the list of sectors potentially at risk for disruptions. No sector has generated more concern than software, where fears that AI agents will replace core workflows have prompted significant declines in stock prices over recent months – though it is notable that the impact on credit spreads has been more modest.

Software is a relatively small component of the investment grade credit universe, accounting for about 1% of European investment grade bonds and 3% of US investment grade bonds, so the impact on the overall market is likely to be moderate. The market’s focus, however, has been on higher exposures to the software sector in private credit – leveraged loans and business development companies (BDCs)¹. Software accounts for about 22% of the portfolio of BDCs with publicly traded unsecured bonds, for example (Exhibit 3).

Financial strength is a key consideration when we evaluate any company either directly or tangentially at risk to AI disruption. Fortunately, most investment grade software companies have strong balance sheets, including many with cash balances in excess of total debt. So, while we recognise the uncertainty around AI disruption, we also believe the sector remains investable. When we invest in software, we favour companies whose mission-critical functions (for example, human resources or customer relationship management), proprietary data and strong competitive moats (determined by intellectual property, regulation or another barrier) make them more resistant to disruption.

At this stage, we anticipate the most negative AI disruptions to affect issuers in the software-as-a-service (SaaS) sector, to which we currently have no direct exposure in portfolios. Our portfolios’ main exposure to software is indirectly via collateralised loan obligations (CLOs), where allowed by guidelines. Typically, these CLOs have manageable exposure to the software sector, while fund holdings tend to be in the most senior part of the capital structure (AAA rated). These positions can benefit from the embedded protection offered by the senior structure – with credit enhancement well in excess of total software exposure – as well as attractive relative value to corporate bonds.

We have become more constructive on certain technology names given revised funding programmes. We also think data centre buildouts can support the capital goods, banking and consumer non-cyclical sectors, where we see the impact of AI as either minimal or positive, given, for example, regulatory hurdles, cost reductions and/or a focus on tangible products.

We also see opportunities within selected parts of the utilities sector. Issuers in regions with data centres are experiencing structurally higher load growth driven by AI infrastructure. Importantly, much of this incremental demand is underpinned by long-term power purchase agreements with hyperscalers, which improves revenue visibility and reduces demand risk relative to speculative capacity buildouts.

In several cases, technology companies are also providing significant upfront capital contributions for interconnection, substations, generation and dedicated grid upgrades. This structure mitigates stranded asset risk and reduces the cost burden on ratepayers, improving the probability of regulatory approval.

The combination of long-term contracted demand and partial risk-sharing meaningfully improves the sector’s growth profile relative to its historical baseline.

We believe AI will have a tangible impact on many business models, both positively and negatively, which is likely to result in increased dispersion in returns. While SaaS businesses are currently in focus for disruption, we believe there are many sectors where AI will help issuers become more efficient and valuable. We currently have a constructive view on the technology and US utility sectors, for example. Overall, a focus on issuer selection and deep fundamental analysis is warranted to identify issuers and sectors that are well-positioned – and to benefit from the resultant increase in dispersion. 

 

¹ BDCs are tax-advantaged investment vehicles focused on middle-market lending. They are subject to a statutory leverage cap of 2x debt-to-equity (150% asset coverage), must invest at least 70% of assets in qualifying assets, and must distribute at least 90% of net income through cash dividends. Beyond these rules, BDCs are not regulated.