AI Companies Are Becoming Major Borrowers. What Does That Mean for Bond Investors?
The world’s largest technology companies are financing an unprecedented expansion of data centres and computing capacity. Their growing use of corporate debt is changing the risks investors take when buying investment-grade bonds.
Artificial intelligence is turning some of the world’s largest technology companies into major borrowers. Alphabet expects capital expenditure of US$180–190 billion in 2026, roughly six times its 2022 spending, while Meta has forecast US$130–145 billion. Much of the money will finance data centres, computing equipment and the infrastructure needed to develop and operate increasingly powerful AI systems.
The expansion has begun to reshape corporate bond markets. According to S&P Global Market Intelligence, technology companies accounted for 27% of US investment-grade corporate bond issuance in 2026, compared with their historically smaller presence in the market. Their growing borrowing requirements are giving fixed-income investors considerably more exposure to the commercial prospects of AI.
The financial position of the largest technology groups remains strong in many respects. Several generate substantial operating cash flow from advertising, cloud computing and established software businesses. Yet their investment programmes are absorbing increasing amounts of that cash, while new bond issuance and financing arrangements provide additional capital.
For Swiss investors holding global corporate bonds, the question extends beyond the creditworthiness of individual technology companies. AI-related borrowing is becoming more prevalent across bond funds, infrastructure investments and private credit, exposing portfolios to businesses whose financial results increasingly depend on the same investment cycle.
The Scale of AI Investment Is Changing Corporate Borrowing
Technology companies traditionally financed much of their expansion from operating cash flow. Their profitable businesses, substantial cash reserves and relatively modest borrowing helped establish some of the largest groups as strong investment-grade issuers.
Data-centre expansion requires capital on a different scale. Operators must finance land, construction, electrical connections, cooling equipment, servers and specialist processors, often years before they recover the expenditure through customer revenues.
Alphabet told investors in June that its planned 2026 capital expenditure would reach US$180–190 billion, compared with approximately US$31 billion in 2022. The company expects spending to increase substantially again in 2027 and has identified operating cash flow, debt issuance and equity financing as sources of funding.
Meta has also accelerated investment. Its second-quarter results showed capital expenditure, including principal payments on finance leases, of US$31.08 billion over three months. The company expects annual spending of US$130–145 billion in 2026.
The increase is taking place across the industry. S&P Global estimated in September that AI-related capital expenditure by major hyperscalers had risen from approximately US$80 billion in 2019 to more than US$700 billion in 2026.
Much of this investment supports cloud computing as well as AI. The companies are buying capacity to serve existing customers, develop new products and meet expectations of continued growth in demand for computing power.
Financing such expansion from cash flow alone would limit the funds available for acquisitions, dividends, share repurchases and other business activities. Debt allows companies to spread some of the financial burden over a longer period, although the obligation to pay interest and repay principal remains regardless of whether the new infrastructure generates the expected revenue.
Strong Credit Ratings Do Not Eliminate the Risks
The largest technology issuers enter this investment cycle with advantages that smaller infrastructure businesses lack. Established revenues, access to international capital markets and diversified operations give them considerable flexibility when financing new projects.
Meta’s second-quarter results illustrate both the financial strength and the demands of its investment programme.
The company reported quarterly revenue of US$60.80 billion and operating cash flow of US$31.86 billion. Capital expenditure reached US$31.08 billion, while reported free cash flow fell to US$784 million.
Meta also held US$90.26 billion in cash, cash equivalents and marketable securities at the end of June, against US$83.66 billion in long-term debt.
Its ability to generate cash remains substantial, but the difference between operating cash flow and spending on infrastructure has narrowed sharply. During the first half of 2026, Meta raised approximately US$24.91 billion in net proceeds from a bond issuance completed in May.
Bond investors must assess how these figures may develop as expenditure continues. A profitable company can maintain a strong credit rating while generating less free cash flow, but repeated borrowing to finance investment may eventually change its debt profile.
S&P Global Ratings warned in August that financing AI infrastructure increasingly involves debt, leases, guarantees and other arrangements beyond conventional corporate bond issuance. Its assessment identified pressure on free operating cash flow across the largest hyperscalers during 2026 and 2027.
The analysis raises questions about the amount of investment financed directly by the companies and the obligations they may assume through partnerships or project structures.
The Returns on AI Investment Will Take Time to Establish
Technology companies expect their new infrastructure to support revenue growth from cloud services, AI applications, advertising and enterprise software. The timing of those returns remains difficult to estimate.
Data centres must be built, connected to electricity networks and equipped before they can generate income. Once operating, they require continuing expenditure on power, maintenance and replacement equipment.
Advanced computing hardware also has a relatively short commercial life compared with buildings and other long-lived infrastructure. As processors become more powerful and efficient, older equipment may lose its competitive value before the associated financing has matured.
These characteristics make the relationship between capital expenditure and future cash flow particularly important.
An equity investor may accept several years of weaker profits if the company ultimately establishes a dominant position in a rapidly growing market. A bondholder receives contractual interest payments and repayment of principal, with limited participation in any increase in the company’s value.
A successful AI investment programme can strengthen a borrower’s finances, but creditors do not receive the same potential gains as shareholders. If anticipated revenues fail to materialise, they remain exposed to deteriorating cash flow and potentially higher credit risk.
The terms of individual bonds determine how much additional compensation investors receive for accepting that uncertainty.
AI Exposure Is Spreading Beyond Technology Bonds
Data-centre developers, electricity suppliers, utilities, equipment manufacturers and specialist cloud providers all require capital to support the expansion. Some finance investments through ordinary corporate bonds, while others use private credit, asset-backed securities, project finance or property-related structures.
S&P Global Market Intelligence reported in August that AI infrastructure financing was becoming more widely distributed across these markets, including commercial mortgage-backed securities and other asset-backed instruments.
A data-centre development may therefore involve several separate borrowers. The technology company purchasing computing capacity could enter a long-term lease or service agreement, while a developer borrows against the property and another institution finances the equipment.
The financial exposure depends on the contracts. Some arrangements provide predictable payments over several years, while others depend on utilisation, future customer demand or the ability to refinance assets.
Investors need to establish who must make the payments, what collateral supports the debt and how the project would perform if its principal customer reduced spending.
S&P Global Ratings warned in October that some AI-related obligations may be difficult to identify through conventional corporate debt figures alone. Guarantees, partnerships and other financing arrangements can leave investors with an incomplete picture of the commitments associated with large infrastructure programmes.
For holders of corporate bonds, the implications extend to the wider credit market. An investment-grade portfolio may own securities issued by a hyperscaler, a utility expanding electricity production and a data-centre landlord whose rental income depends on the same customer.
Corporate Bond Funds May Be More Concentrated Than They Appear
Traditional credit portfolios diversify risk by holding bonds issued by numerous companies across different industries.
S&P Global Market Intelligence reported in October that correlations among hyperscalers, semiconductor companies and other AI-related issuers had increased during 2026. Credit spreads on some AI-exposed technology bonds had also widened relative to the broader US investment-grade market.
A bond fund might hold securities issued by an established technology group alongside debt from chip manufacturers, electricity utilities and specialist infrastructure businesses. The companies operate in different sectors, but their financial performance may be affected by the same slowdown in AI investment.
A spread represents the additional yield investors demand above a comparable government bond or another reference rate. When it widens, the market value of an existing corporate bond generally falls, assuming other factors remain unchanged.
Changes in credit spreads can occur well before a company experiences difficulty making its payments. Investors may demand greater compensation because expenditure is rising, free cash flow is weakening or future refinancing looks less certain.
In October, S&P Global Market Intelligence identified widening spreads among AI-exposed issuers even as the wider investment-grade market remained relatively resilient.
Portfolio managers consequently need to examine concentration across industries rather than relying exclusively on the sector classifications of bond issuers.
Debt Maturities and Refinancing Costs Deserve Closer Attention
Major technology companies can issue bonds with maturities extending well beyond the period in which their current infrastructure investments are expected to generate returns.
Longer maturities allow borrowers to secure funding for many years and reduce the frequency with which they must return to capital markets. They can also increase the interest expenses a company commits to paying over time.
The maturity of the bond determines part of its sensitivity to interest rates. Longer-dated fixed-rate securities generally experience larger price changes when market yields move.
Credit risk adds another consideration. A company’s balance sheet may change substantially between the issuance of a bond and the date on which it must be repaid.
Large capital-expenditure commitments, acquisitions, leases and additional borrowing can affect the financial position of an issuer over that period.
The refinancing schedule matters particularly for companies expecting to repay existing debt through new borrowing. Higher market yields or weaker credit ratings may make that process more expensive.
A company with substantial cash reserves and recurring operating income generally has greater flexibility than an infrastructure developer whose revenues depend on a limited number of projects or customers.
Credit analysis should therefore distinguish between established technology groups and companies whose financing depends heavily on the success of the AI build-out.
Could Technology Borrowing Raise Costs for Other Companies?
The expansion of AI infrastructure is increasing demand for capital at a time when governments and businesses are already issuing substantial amounts of debt.
S&P Global Market Intelligence estimated in September that technology companies accounted for 27% of US investment-grade corporate bond issuance and 20% of high-yield issuance during 2026.
The figures demonstrate the industry’s growing presence in credit markets, although they do not establish that technology borrowing has directly raised financing costs for other sectors.
Corporate borrowers compete for investors’ capital, and the volume of bonds coming to market can influence pricing. The consequences depend on overall demand for fixed-income securities, monetary conditions and the amount of risk investors are willing to accept.
Andrew Balls, PIMCO’s chief investment officer for global fixed income, has raised the possibility that hyperscalers’ substantial demand for financing could make it more difficult for other businesses to attract capital at affordable rates.
Large technology companies may be less sensitive to modest changes in borrowing costs because of their cash reserves and expected returns on major investment programmes. Smaller or more heavily indebted businesses could face greater pressure.
The effect would not necessarily be uniform across credit markets. Strong demand for corporate bonds can accommodate substantial new issuance, while periods of market volatility may prompt investors to demand higher premiums from borrowers they regard as less financially secure.
For bond investors, the changing supply of debt warrants attention alongside the fundamentals of the companies issuing it.
What Swiss Investors Should Examine
Swiss investors holding global investment-grade bond funds may already have exposure to the technology companies financing AI infrastructure.
The extent of that exposure depends on a fund’s benchmark, issuer selection, maturity profile and credit-quality requirements. Managers following broad corporate bond indices can acquire larger positions in sectors that account for a growing share of outstanding debt.
Investors should examine the holdings rather than relying on a fund’s broad sector allocation.
The credit quality of individual borrowers matters, but so does their exposure to common business risks. A diversified portfolio can become more sensitive to AI investment spending as technology companies, infrastructure operators and related businesses account for a larger proportion of its assets.
Currency exposure also affects returns.
Most of the largest technology issuers borrow substantially in US dollars. Swiss investors holding their bonds without currency hedging remain exposed to movements in USD/CHF.
A stronger franc can reduce the return from dollar-denominated securities even when the issuer makes every scheduled payment. Hedging may reduce that exposure, but the costs depend on the interest-rate difference between the currencies, market conditions and transaction expenses.
The decision to own longer-dated corporate bonds should also reflect the investor’s need for liquidity. A bond may offer an attractive yield when purchased but lose market value if interest rates or credit spreads increase before it matures.
Families holding private credit or infrastructure investments should review those exposures alongside publicly traded bonds. AI-related financing may appear in several parts of the family balance sheet, including funds whose underlying assets are less frequently valued.
Assessing AI Debt Within a Multi-Asset Portfolio
The growth of AI borrowing is changing the composition of corporate credit markets, but it has not made every technology bond equally risky.
Established companies with diversified earnings, substantial cash reserves and investment-grade ratings differ considerably from highly leveraged infrastructure developers or specialist cloud businesses dependent on a small number of customers.
The comparison also depends on the price of the bond.
Higher yields and wider credit spreads can compensate investors for additional risk, although they provide little protection if the borrower experiences a substantial deterioration in its financial position.
For family offices, the review should examine exposure to individual issuers, the maturity of their obligations and the extent to which future cash flows depend on continued AI investment.
Debt issued by the largest technology companies may also overlap with equity holdings in the same businesses, creating exposures on both sides of a diversified portfolio.
The size of those positions, their sensitivity to changing market conditions and the family’s future liquidity requirements determine how much risk the overall allocation can accommodate.
How to Assess AI Exposure in a Corporate Bond Portfolio
1. How much AI-related credit exposure do we hold?
Request a list of corporate bonds, private credit positions and infrastructure investments connected to AI spending. Include hyperscalers, semiconductor companies, data-centre operators, utilities and specialist cloud providers.
2. Are borrowers generating enough cash to finance their investments?
Compare operating cash flow with capital expenditure, lease payments, interest expenses and other commitments. Review whether free cash flow is sufficient to support existing debt without repeated new borrowing.
3. What debt must be repaid or refinanced?
Examine outstanding debt, scheduled maturities and the interest rates on existing borrowing. Ask how the issuer’s finances would change if refinancing costs increased or AI revenues developed more slowly than expected.
4. Are we relying too heavily on the same investment cycle?
Identify companies whose earnings or debt repayments depend on spending by a small number of technology groups. Look for common exposures across bonds, equities, private credit and infrastructure funds.
5. What return are we receiving for the risks?
Compare credit spreads, ratings and maturities with similar bonds outside the AI sector. For Swiss investors, assess yields after currency hedging and consider whether the investment’s price volatility is compatible with planned cash requirements.
