By Charudatta Shende, head of client portfolio management fixed income and fixed income strategist at Candriam
Artificial intelligence is promising to reshape how businesses operate and how productivity is generated. Its investment cycle is already supporting activity, well before the full benefits of adoption can be measured.
What is tangible is the commitment of capital to the infrastructure required to make those ambitions possible. The challenge is to finance the future with sufficient protection against the uncertainties of the present.
AI enters credit markets at scale
AI financing has become an increasingly prominent feature of credit markets. The surprise has been the scale and breadth of borrowing by the hyperscalers, the large technology groups behind the expansion of cloud and AI capacity. This marks a change for a sector long associated with the ability to finance growth from its own cash generation.
The scale of infrastructure spending is bringing even the strongest balance sheets back to the bond market. AI is consequently influencing portfolio exposure even where investors have made no explicit thematic allocation.
See also: PortfolioMetrix’s Funk: Bonds are behaving exactly as they should
The financing requirement is also reaching high yield, as specialist developers and operators undertake part of the data-centre construction that hyperscalers choose to procure externally. This type of financing brings construction and project-finance risk into portfolios alongside technology exposure. Strong customers can support these structures, but their credit quality does not automatically transfer to the borrower.
We expect substantial further supply into 2027 as investment commitments are translated into operating capacity. The opportunity set should broaden, while the market’s ability to absorb that supply will increasingly depend on pricing and structure.
A broader opportunity for credit
AI investment is already supporting economic activity, with demand extending into utilities, electrical equipment, engineering and construction. This gives the investment cycle a broader economic footprint and helps explain why its significance extends beyond the fortunes of individual AI providers.
Debt financing can support productive assets whose useful contribution outlasts the initial wave of enthusiasm. The eventual productivity dividend remains uncertain, but the current investment cycle has tangible economic value.. Established franchises, substantial operating cash flows and contracted demand provide a credible foundation for parts of this expansion.
Credit markets also gain greater diversity and additional sources of performance. Investment-grade investors can access more technology debt, while high-yield investors gain a growing infrastructure segment. Completion and the commencement of rental payments can materially reduce risk, although an investment-grade tenant does not guarantee an investment-grade borrower.
There is also the attraction of carry. Where the yield adequately compensates for leverage, execution risk and the bond’s terms, this can be a valuable addition to portfolios. We need cashflows capable of meeting interest payments and repaying principal, with enough flexibility to withstand a less favourable outcome.
The risks begin with physical delivery
Much of the planned expansion in AI depends on data centres that must first be built, connected and commissioned. That pipeline illustrates the distance between announced ambition and revenue-producing assets. Pre-leasing provides comfort on demand, but cannot remove the risk that a facility is delivered late or above budget.
Power is a particularly difficult constraint. A building may be completed before the grid connection, transmission capacity or generation needed to operate it is ready. Delays then become a credit issue: interest accrues while lease commencement and cash generation may be deferred.. As these exposures grow, a common construction bottleneck could affect credit portfolios across rating categories.
Public acceptance creates another hurdle. The concern for creditors is not only whether a project eventually receives approval, but how long that process takes and whether additional conditions change its economics. Commercial demand alone is insufficient to ensure delivery. A project must also secure the infrastructure and local acceptance required to operate.
See also: Saranac’s Urwin: AI has come for the bond market, but not how you think
Financing costs compound these pressures. Elevated sovereign yields increase the underlying cost of borrowing, while heavy supply can require issuers to offer larger new-issue premiums. Even strong hyperscalers face greater investor scrutiny as their funding needs expand. New-issue concessions reward investors for committing fresh capital, but they also demonstrate that the market is becoming more selective.
For borrowers, higher coupons increase the importance of converting investment into cash flow on time. For investors, the additional carry is attractive only if it provides sufficient compensation for the risks assumed.
The financing structure also matters. Leases, joint ventures and project vehicles can distribute obligations across several entities without removing the underlying economic dependence. A long-term commercial relationship is not equivalent to an unconditional guarantee. Strong headline ratings deserve attention, but the terms of an individual security can be just as decisive.
Financial resilience will determine the outcome
AI offers a meaningful long-term opportunity, but the path from investment to dependable cash generation is unlikely to be linear. Fundamental, bottom-up research remains our most useful guide. At issuer level, this means assessing balance-sheet resilience, total financial commitments and the capacity to moderate spending if revenues disappoint. At bond level, covenants, guarantees, seniority and refinancing requirements determine how much protection creditors actually receive.
Issuer and issue selection will be essential to capture the carry on offer while retaining resilience if assumptions prove too optimistic.
For bondholders the more useful discipline is to ask whether the borrower can remain solvent and retain access to liquidity while the commercial model matures.
AI may follow its own course, but creditors have good reason to focus on the capacity to survive the gap between expectations and execution. Amid the debate over whether AI is viable, beneficial or overvalued, financial resilience remains fundamental.
