Some of the world’s most highly valued AI companies remain lossmaking, consume substantial amounts of capital and have relatively short operating histories. They may nevertheless be seeking the same investment-grade rating held by profitable, established companies such as Oracle.
That does not necessarily mean such a rating would be inappropriate. Rapid revenue growth, access to equity capital and the strategic importance of leading AI models could all support a strong credit assessment. However, the comparison raises a more important question for institutional investors: can the same rating adequately describe creditworthiness when the underlying sources of financial strength are so different?
What does an investment-grade rating change?
A credit rating is an opinion about an issuer’s ability to meet its financial obligations. In practice, it can also determine which investors can own the debt, whether a bond enters an investment-grade index and how much capital a regulated institution must hold against it.
This gives the lowest investment-grade rating particular importance. Crossing the boundary from speculative grade to BBB− can attract a broader pool of banks, insurers and asset managers, including portfolios whose mandates restrict them from holding lower-rated securities. It may also reduce borrowing costs at a time when AI companies require substantial amounts of external capital.
Failure to secure investment-grade status would not prevent an AI company from borrowing, but it could materially change the economics of its expansion. A smaller pool of eligible investors could demand higher yields, shorter maturities, tighter covenants or additional guarantees. When proposed infrastructure requires tens of billions of dollars in upfront capital, even a modest increase in funding costs could reduce expected returns sufficiently to delay, scale back or prevent some projects from proceeding.
The rating therefore does more than describe risk. It affects which institutions can buy the bonds, the terms on which capital is available and whether some projects remain financially viable.
Does a high equity valuation protect creditors?
A large equity valuation can provide a meaningful cushion beneath a company’s debt. It suggests that shareholders see considerable value in the business and may be willing to provide further capital before creditors suffer losses.
Yet that protection is less certain than it first appears. Equity valuations reflect expectations about future growth, margins and market leadership, while debt must be serviced with cash. If those expectations weaken, the valuation supporting the apparent cushion could fall at the same time that access to new capital becomes more difficult.
This distinction matters for frontier AI laboratories. Their revenues may be growing rapidly, but compute, research and employee costs remain exceptionally high. Their credit strength could therefore depend partly on investors continuing to fund losses while the companies work towards sustainable cash generation.
Credit analysis must consider not only the amount of equity beneath the debt, but how dependable that equity would remain under less favourable conditions.
Can very different companies carry the same rating?
Oracle provides a useful comparison. It has an established software business, substantial operating profits and decades of corporate history. Its credit risk has increased because of the debt and long-term lease commitments associated with its expansion into AI infrastructure.
Frontier AI laboratories present almost the reverse profile. They may begin with smaller established debt burdens and much higher expected growth, but they also have shorter operating records, negative earnings and significant future funding requirements.
Both types of company might conceivably warrant the same rating, but not for the same reasons. Oracle’s creditors can assess existing cash flows against its obligations. An AI laboratory’s rating may place greater weight on expected growth, market position, access to equity funding and the willingness of strategic partners to continue providing support.
For investors, the shared rating should not obscure the different ways in which each credit could deteriorate.
How far does the exposure extend?
The risks are also connected. AI laboratories commit to purchasing computing capacity, infrastructure companies build data centres against that demand, and lenders finance the resulting construction and equipment.
According to the Financial Times, S&P estimates that OpenAI accounts for approximately half of Oracle’s $638 billion of remaining performance obligations. Oracle has meanwhile accumulated substantial debt and off-balance-sheet lease commitments as it invests to meet anticipated demand.
This means securities issued by apparently separate companies may depend on the same underlying assumptions. Continued demand for AI services must enable laboratories to meet their infrastructure commitments, allowing providers to service the debt raised to build that capacity.
If demand fell short of expectations, an AI laboratory might seek to delay or renegotiate future infrastructure commitments just as its suppliers were bringing heavily financed capacity online. Even where existing contracts remained enforceable, doubts about the customer’s ability to meet them could affect the provider’s revenues, refinancing prospects and the value of the debt used to fund construction.
A portfolio containing bonds from AI developers, data-centre operators and technology suppliers may therefore be less diversified than the number of issuers suggests.
What happens after a downgrade?
If an AI company received an investment-grade rating that later proved too generous, the first consequence would probably be a sharp increase in its borrowing costs. The wider effects could be more significant.
A downgrade to speculative grade may remove bonds from investment-grade indices and oblige mandate-constrained investors to sell. Funds able to retain the debt might still reduce their exposure before other holders are forced into the market. Refinancing could then become more expensive precisely when the company needed further capital.
Investors might also reassess related issuers. Data-centre developers, cloud providers and technology companies with concentrated AI exposure could face wider credit spreads even if their own ratings remained unchanged.
The issue is not simply whether one company defaults. It is whether a reassessment of one borrower changes the market’s view of the contracts, counterparties and financing structures supporting the wider ecosystem.
Would stronger companies be affected?
Highly rated companies would not become less creditworthy because an AI borrower had been assessed incorrectly. Indeed, they could initially benefit if investors moved towards issuers with established cash flows, lower leverage and more diversified customers.
The longer-term effect could be less straightforward. Investors might demand additional yield from BBB borrowers, particularly where ratings depend heavily on future growth or continued capital-market access. They may also place less reliance on the rating alone and conduct more detailed analysis of contractual obligations, customer concentration and refinancing risk.
Stronger companies would not directly absorb those losses, but they could face greater scrutiny and higher borrowing costs as investors reassessed similar risks across the credit market.
A test of consistency
Rating agencies do not need to decide whether AI will transform the global economy. They must judge whether individual companies can meet their obligations if growth is slower, capital becomes more expensive or important commercial relationships weaken.
For institutional investors, the same discipline applies. The relevant questions include when cash generation is expected to begin, how much further funding will be required and whether different holdings ultimately rely on the same customers and growth assumptions.
An investment-grade rating can introduce a borrower to vast pools of institutional capital. If that assessment proves sound, it can support the financing of a major technological expansion. If it proves wrong, the consequences may travel well beyond the company whose name appears on the bond.