Credit Markets Are Quietly Pricing an AI Borrowing Risk Tech Stocks Ignore
For most of the last decade, the largest technology companies barely needed lenders. Software cost little to copy, cash piled up, and debt was a minor line on the balance sheet. AI has changed that. The four largest cloud providers were projected to spend more than $200 billion on capital projects in 2025 alone: data centres, specialized chips and the power to run them. A growing share of that is borrowed, and trading in credit default swaps on big tech names has jumped as a result.
A credit default swap works like an insurance policy on a company's debt. The buyer pays a regular fee, and if the company fails to pay its bondholders, the seller covers the loss. The surge in this trading tells you something a share price can't: the people who get paid only when money arrives on time are asking whether AI revenue will show up before the loan payments do. That timing question applies to every capital-heavy bet you hold, lend to or rely on, well beyond a handful of Silicon Valley names.
Credit is paid by the calendar
A shareholder is paid by how big the eventual profit becomes. If AI pays off in 2031 instead of 2028, a patient shareholder can still do well. A bondholder has a fixed schedule. Interest is due every six months, and principal is due on a set date. A three-year delay in revenue helps the bondholder not at all, and it can hurt badly.
So the two markets can look at the same company and reach different moods. Equity prices the size of the prize. Credit prices whether the bills get paid along the way. When protection buying rises while stocks stay strong, the useful reading is that credit holders see a gap between when the spending happens and when the cash comes back.
Asset-heavy borrowers deserve a cash-flow review
Tech used to carry debt profiles closer to a consulting firm. AI pushes it toward a utility or a manufacturer: big physical assets, long build times, equipment that ages fast. Borrowing costs matter far more in that model. The Bank of Canada has noted that higher interest rates have already raised debt-servicing costs for Canadian businesses, and short-term rates in 2026, though off their peak, sit well above the near-zero levels of the 2020-2021 tech boom. A data centre financed today carries a heavier interest bill than one financed five years ago, which shortens how long a company can wait for profits.
The practical step is to look at any asset-heavy borrower in your book the way a lender would. How many years of interest does current cash flow cover? When do the large maturities land? What happens if revenue arrives two years late?
Protection is a hedge you can copy
Keep the signal in proportion. Microsoft, Alphabet and their peers hold enormous cash reserves, and an actual default remains unlikely. Buying insurance often means professionals are being careful during a period of fast change, and the cost to insure big tech debt is still well below that of troubled industries. If AI does automate expensive work at scale, the debt could be repaid faster than sceptics expect.
What's worth borrowing is the habit. Credit desks separate "this business will win" from "this business can pay on schedule," and they hedge the second even when they believe the first.
That separation matters close to home too. Edmonton has built a real AI research base through Amii, and local hiring, venture funding and office leasing all lean on the health of the wider tech sector. A landlord with tech tenants, a lender to growth firms, or an investor in a fund heavy with AI exposure faces the same question the swap market is asking: if the payoff runs late, does the cash still cover the payments?
Run that test before you add exposure. If the answer depends on everything arriving on time, you've found where your own protection belongs.
For most of the last decade, the largest technology companies barely needed lenders. Software cost little to copy, cash piled up, and debt was a minor line on the balance sheet. AI has changed that. The four largest cloud providers were projected to spend more than $200 billion on capital projects in 2025 alone: data centres, specialized chips and the power to run them. A growing share of that is borrowed, and trading in credit default swaps on big tech names has jumped as a result.
A credit default swap works like an insurance policy on a company's debt. The buyer pays a regular fee, and if the company fails to pay its bondholders, the seller covers the loss. The surge in this trading tells you something a share price can't: the people who get paid only when money arrives on time are asking whether AI revenue will show up before the loan payments do. That timing question applies to every capital-heavy bet you hold, lend to or rely on, well beyond a handful of Silicon Valley names.
Credit is paid by the calendar
A shareholder is paid by how big the eventual profit becomes. If AI pays off in 2031 instead of 2028, a patient shareholder can still do well. A bondholder has a fixed schedule. Interest is due every six months, and principal is due on a set date. A three-year delay in revenue helps the bondholder not at all, and it can hurt badly.
So the two markets can look at the same company and reach different moods. Equity prices the size of the prize. Credit prices whether the bills get paid along the way. When protection buying rises while stocks stay strong, the useful reading is that credit holders see a gap between when the spending happens and when the cash comes back.
Asset-heavy borrowers deserve a cash-flow review
Tech used to carry debt profiles closer to a consulting firm. AI pushes it toward a utility or a manufacturer: big physical assets, long build times, equipment that ages fast. Borrowing costs matter far more in that model. The Bank of Canada has noted that higher interest rates have already raised debt-servicing costs for Canadian businesses, and short-term rates in 2026, though off their peak, sit well above the near-zero levels of the 2020-2021 tech boom. A data centre financed today carries a heavier interest bill than one financed five years ago, which shortens how long a company can wait for profits.
The practical step is to look at any asset-heavy borrower in your book the way a lender would. How many years of interest does current cash flow cover? When do the large maturities land? What happens if revenue arrives two years late?
Protection is a hedge you can copy
Keep the signal in proportion. Microsoft, Alphabet and their peers hold enormous cash reserves, and an actual default remains unlikely. Buying insurance often means professionals are being careful during a period of fast change, and the cost to insure big tech debt is still well below that of troubled industries. If AI does automate expensive work at scale, the debt could be repaid faster than sceptics expect.
What's worth borrowing is the habit. Credit desks separate "this business will win" from "this business can pay on schedule," and they hedge the second even when they believe the first.
That separation matters close to home too. Edmonton has built a real AI research base through Amii, and local hiring, venture funding and office leasing all lean on the health of the wider tech sector. A landlord with tech tenants, a lender to growth firms, or an investor in a fund heavy with AI exposure faces the same question the swap market is asking: if the payoff runs late, does the cash still cover the payments?
Run that test before you add exposure. If the answer depends on everything arriving on time, you've found where your own protection belongs.
Read Next
A Tesla and SpaceX Merger Solves Musk's Problems and Hands Shareholders His Risks
Prem Watsa's IDBI Bet Tests What Foreign Capital Can Do in Indian Banking
When Insiders Lose Faith in Housing: 7 Assumptions Every Buyer Should Stress-Test
Prices Are Down 20%: 7 Underwriting Checks Before Your Next Rental Build