The artificial-intelligence boom is entering a new phase, and the biggest question is no longer simply how much companies can spend on AI infrastructure. The question is how much debt investors are willing to absorb to finance that spending.
That concern is increasingly important as major technology companies and AI infrastructure providers turn to bond markets to fund enormous investments in data centers, computing capacity, power and networking.
A senior JPMorgan bond banker has warned that the growing wave of AI-related borrowing could test investor tolerance, particularly if companies continue issuing debt at a rapid pace.
The shift matters because AI was initially financed largely by the enormous cash flows generated by established technology companies. But as spending requirements have exploded, even highly profitable businesses are increasingly looking toward debt markets.
JPMorgan research indicates that US technology companies have already become major sources of investment-grade bond supply. Net technology issuance reached $131 billion in 2025, more than twice the average annual level of the previous five years. JPMorgan expects technology and AI-related funding to remain a dominant force in 2026.
AI Is Becoming a Debt-Financed Boom
The economics of AI require enormous physical infrastructure.
Training and operating advanced AI models requires data centers containing thousands of specialized processors. Those facilities also need electricity, cooling systems, fiber networks and other supporting infrastructure.
Building all of that costs billions of dollars.
Technology companies therefore face a difficult financial calculation: invest aggressively now to capture future AI demand, or slow spending and risk falling behind competitors.
So far, the industry’s answer has been clear.
Spend.
The problem is that capital expenditure is growing faster than many companies can comfortably finance entirely from operating cash flow.
Goldman Sachs estimates that roughly one-third of hyperscaler capital expenditure could be debt-financed in 2026, with direct bond issuance from hyperscalers potentially reaching around $250 billion.
That represents a major change in the financial structure of the AI boom.
Why Bond Investors Are Important
Equity investors participate directly in the upside if AI companies become more profitable.
Bond investors do not.
A bondholder receives interest and repayment of principal rather than a share of the company’s future growth.
That difference makes credit investors much more focused on leverage, cash flow and the ability to repay debt.
A company can therefore be extremely attractive to equity investors while looking less attractive to bond investors.
This creates an important dividing line in the AI investment story.
The question is not whether AI will transform the economy.
The question for bond investors is whether the returns generated by that transformation will arrive quickly enough to justify today’s borrowing.
The Scale of New Borrowing
The potential scale of AI-related borrowing is enormous.
JPMorgan’s analysis suggests net issuance from the US technology sector could reach about $230 billion in 2026, with AI and technology infrastructure playing a major role. The bank has also estimated that investment-grade debt used to finance AI-related capital expenditure could eventually reach trillions of dollars.
Goldman Sachs estimates that debt financing could account for about 35% of hyperscaler capital expenditure in 2027.
This does not necessarily mean technology companies are approaching a debt crisis.
Many of the largest hyperscalers have enormous revenues, substantial cash balances and strong investment-grade credit ratings.
The concern is instead about the sheer volume of supply.
Investor Appetite Has Limits
Credit investors can absorb substantial amounts of new debt, but demand is not unlimited.
Every new bond issue competes with existing bonds for investor capital.
If supply grows too quickly, issuers may need to offer higher yields to attract buyers.
That can increase borrowing costs.
It can also cause the prices of existing bonds to fall.
This is why the phrase “investor tolerance” matters.
The market may continue buying AI-related bonds, but investors could gradually demand better compensation for taking on additional duration and concentration risk.
Strong Companies Can Still Create Concentration Risk
One of the unusual features of the AI debt boom is that the companies issuing the most debt are not necessarily weak businesses.
Many are among the world’s largest and most profitable corporations.
That creates a different type of risk.
If a handful of technology companies issue enormous amounts of debt, they can become increasingly important components of credit indexes.
JPMorgan Asset Management has warned that bond indexes weighted by outstanding debt can naturally give greater exposure to companies that issue the most bonds.
Passive investors may therefore end up owning more AI-related credit exposure even if they did not explicitly choose to invest in AI.
The Index Problem
Equity indexes are generally weighted by market capitalization.
Bond indexes work differently.
Companies with more debt can have larger representation in certain bond benchmarks.
That means a major borrowing program can increase a company’s importance in credit portfolios.
If several hyperscalers simultaneously increase borrowing, the concentration can become significant.
This creates a strange situation.
The more debt a company issues, the more important it can become to certain passive credit portfolios.
What Happens If AI Returns Disappoint?
The biggest long-term risk is not necessarily default.
It could simply be disappointment.
Companies are spending enormous sums today based on expectations about future AI demand.
If AI adoption grows more slowly than expected, the revenue generated by those investments could arrive later.
But the debt used to finance data centers would still have to be serviced.
Interest payments do not wait for AI adoption.
Neither do bond maturities.
That creates a potential mismatch between the timing of investment and the timing of returns.
The Capital Expenditure Cycle
The AI industry is currently operating under enormous pressure to expand capacity.
Companies fear that insufficient investment could leave them without enough computing power to meet future demand.
That creates a classic first-mover problem.
If one company spends aggressively and another does not, the more aggressive company could gain an advantage.
This encourages everyone to keep spending.
But if every company simultaneously expands capacity, the industry could eventually create excess infrastructure.
That is where debt becomes particularly important.
Debt magnifies both successful investments and unsuccessful ones.
Data Centers Are Expensive Long-Term Assets
AI infrastructure is not like ordinary software.
A software company can sometimes launch a product with relatively modest capital expenditure.
Data centers require billions of dollars in construction and equipment.
They also have long useful lives.
That makes the financing structure important because investors may be asked to provide capital for assets whose economic returns will unfold over many years.
Goldman Sachs has noted that bond investors have shown greater comfort with shorter-duration debt than very long-dated exposure as the AI financing cycle develops.
Longer Maturities Create Another Risk
Long-term AI infrastructure financing can expose investors to interest-rate risk.
If rates rise, the value of existing long-duration bonds can decline.
If inflation remains elevated, investors may demand higher yields before committing to decades of financing.
That makes the cost of AI infrastructure sensitive not only to technology demand but also to the broader bond market.
AI is therefore becoming increasingly connected to monetary policy.
Credit Spreads Are the Warning Signal
One of the most important indicators will be credit spreads.
Credit spreads measure the additional yield investors demand to hold corporate debt rather than safer government bonds.
If investors become increasingly nervous about AI-related borrowing, spreads should widen.
That would mean companies need to pay more to borrow.
JPMorgan has noted that credit-default-swap spreads have widened for some hyperscaler names as markets reassess the implications of rising leverage.
That does not mean investors expect widespread defaults.
It does indicate that credit markets are becoming more attentive to leverage.
Hyperscalers Still Have Strong Balance Sheets
This is where the bearish argument needs to be restrained.
The largest technology companies are not comparable to heavily indebted speculative companies.
Many hyperscalers generate enormous operating cash flows and maintain strong balance sheets.
Their debt-to-EBITDA ratios remain relatively low compared with much of the investment-grade universe.
That gives them substantial room to borrow.
The real concern is therefore not that the biggest technology companies suddenly become insolvent.
It is that the scale of borrowing could eventually become large enough to change how credit markets price technology risk.
AI Infrastructure Companies Face Greater Risk
The situation is different for companies that lack the balance sheets of the hyperscalers.
Data-center developers, specialized infrastructure companies and other AI-adjacent businesses may depend much more heavily on external financing.
These businesses can face higher borrowing costs and more complicated capital structures.
If AI demand slows, they may be more vulnerable than the technology giants that ultimately use their infrastructure.
This creates a potential weak link in the AI financing chain.
Banks Are Also Exposed
Banks play an important role in financing AI infrastructure.
They provide loans, underwriting services, credit facilities and other forms of financing.
If the amount of debt in the sector rises rapidly, banks may eventually want to distribute some of that risk to investors.
That can involve bond markets, private credit funds and other institutional investors.
The result is that AI financing becomes a broader financial-market issue rather than a problem confined to technology companies.
Private Credit Could Become More Important
Private credit is another potential source of financing for AI infrastructure.
Private lenders can structure customized loans for data centers, energy projects and other specialized assets.
That may help companies obtain financing when public bond markets become more selective.
But private credit also introduces transparency concerns because many loans are not priced continuously in public markets.
Investors may therefore have less visibility into the true market value of the underlying debt.
The AI Bubble Debate Is Expanding
For years, concerns about an AI bubble focused mainly on stock valuations.
The debate is now expanding into credit.
That is significant because debt markets can impose discipline on companies in ways equity markets sometimes do not.
Equity investors can tolerate volatility if they believe long-term growth will eventually justify high valuations.
Bond investors have a much narrower margin for error.
They need to know that interest and principal will be paid.
Shareholders Are Also Becoming More Demanding
AI spending has so far received substantial support from shareholders.
Large technology companies have been able to justify enormous capital expenditure because investors believe AI could create new revenue streams and strengthen competitive positions.
But that tolerance may not last indefinitely.
JPMorgan’s own outlook notes that large technology companies are shifting from relatively capital-light businesses toward capital-intensive models, creating greater pressure to demonstrate returns on AI investments.
Eventually, investors may ask whether every additional billion dollars of AI spending is generating sufficient economic value.
The Return-on-Investment Question
This may ultimately become the most important issue.
AI infrastructure spending can be justified if it produces sufficiently large future cash flows.
But the investment must eventually generate returns.
Companies therefore need to demonstrate that AI is doing more than increasing computing capacity.
They need evidence that AI is producing:
- Higher revenues
- Better margins
- New products
- Greater market share
- Lower operating costs
- Stronger customer retention
Without those benefits, debt-financed capital expenditure becomes harder to justify.
What Could Trigger a Market Shift?
Several developments could change investor sentiment quickly.
Slower AI Revenue Growth
If AI-related revenues disappoint, investors may question future infrastructure spending.
Higher Interest Rates
Higher rates would increase financing costs and reduce the attractiveness of long-duration debt.
Excess Data-Center Capacity
Too much infrastructure could reduce returns on investment.
Wider Credit Spreads
Higher spreads would make new borrowing more expensive.
Rating Downgrades
A deterioration in credit quality could force some investors to reduce exposure.
Weaker Economic Growth
A slowdown could pressure corporate cash flows just as debt-service obligations rise.
Why This Is Not Yet a Crisis
Despite the concerns, there is no clear evidence that the AI debt market is currently in systemic distress.
Demand for new investment-grade technology bonds remains strong.
AI-related issuers continue to have substantial access to capital markets.
The biggest companies also have considerable financial resources.
JPMorgan’s research describes strong fundamentals and relatively low downgrade risk among major AI-related issuers.
The issue is therefore one of sustainability rather than immediate collapse.
The Bond Market May Become the Real Discipline
The equity market can remain enthusiastic about AI for a long time.
The bond market may be less forgiving.
If investors decide that technology companies are borrowing too aggressively, they can demand higher yields.
That raises financing costs.
Higher financing costs could eventually force companies to slow capital expenditure.
In that sense, the bond market could become one of the mechanisms that determines how quickly the AI infrastructure boom can continue.
A New Phase of the AI Story
The AI revolution began as a technology story.
It became an equity-market story when investors started rewarding companies expected to benefit from AI.
Now it is becoming a credit-market story.
The next stage could be determined partly by the willingness of bond investors to finance the enormous infrastructure required to keep AI expanding.
That makes the comments from senior bond bankers particularly relevant.
The market is not necessarily saying “no” to AI debt.
It is asking how much is too much.
What Investors Should Watch
Investors should focus on several indicators over the coming years:
- Total AI-related bond issuance
- Technology-sector credit spreads
- Hyperscaler leverage
- Debt-to-EBITDA ratios
- Interest coverage
- Data-center utilization
- AI revenue growth
- Free cash flow
- Capital-expenditure forecasts
- Bond-market demand
- Credit-rating changes
Taken together, these measures can provide a better picture than stock prices alone.
Conclusion
The AI boom is increasingly dependent on debt markets, and that changes the risk profile of the technology sector.
Major companies have strong balance sheets and substantial cash generation, so the current borrowing wave should not automatically be interpreted as a sign of an impending financial crisis.
But the scale is becoming difficult to ignore.
JPMorgan expects AI and technology-related borrowing to remain a major source of investment-grade bond issuance, while Goldman Sachs estimates that debt could finance roughly one-third of hyperscaler capital expenditure this year and an even larger share in 2027.
The key question is whether investor demand can keep pace with the supply.
If AI revenues and productivity gains eventually justify today’s enormous infrastructure spending, the additional debt may prove manageable.
If returns arrive more slowly, however, investors could become less willing to finance ever-larger amounts of long-term AI infrastructure.
That would not necessarily end the AI boom.
It could simply force companies to become more selective about where they spend.
The biggest change is that AI is no longer being financed solely by optimism about future technology.
It is increasingly being financed by the bond market.
And bond investors, unlike equity investors, eventually demand proof that the money being borrowed today can generate enough cash tomorrow.
