Interest rates as a trigger for the AI bubble to burst
The essentials in 30 seconds
- As of October 2026, interest rates can trigger an AI bubble to burst through two channels: valuation and borrowing cost.
- A rise from 4% to 5% cuts the present value of a profit due in 20 years by about 17%.
- The EQOS Asia AI study places the enterprise AI market at 0.337, a fragile zone (April 2026).
- Two channels deserve a watch: the Fed rate path and the yen carry trade in Asia.
- A three-step stress test shows how much a rate rise costs your AI exposure.
Interest rates are the slowest trigger of an AI bubble burst and the widest. They sit within the wider question of what would trigger the AI bubble to burst. This page explains the two channels through which rates act, gives a worked example with round numbers, and points to the two deeper pages on the Fed and on the yen. Everything here is general information as of October 2026.
On this page
Compare the EQOS Asia offers and request a 30-minute pre-diagnostic.
How do interest rates act on an AI bubble?
A share price is the present value of the profits a company is expected to earn. Interest rates set the discount applied to those profits. When rates rise, each future dollar is worth less today, and the price falls even if the forecast does not change. The reverse also holds, which is why falling rates helped lift valuations in the first place. A bubble that grew in a period of cheap money therefore carries a rate risk built into its price.
Rates also change behaviour. Safe assets pay more, so investors ask more from risky ones. Money moves out of the stories that need patience and into cash that pays now. A bubble depends on investors accepting low returns today for large returns later, and higher rates make that trade harder to accept.
Why do AI stocks react more to interest rates than other shares?
AI valuations lean on profits that arrive late. Consider an illustrative profit of 100 due in 20 years. At a 4% discount rate it is worth about 45.6 today. At 5% it is worth about 37.7. That is a drop of about 17% from a one point rise. A profit due in 10 years falls only about 9%, from 67.6 to 61.4.
The lesson is about timing. The later the profit, the larger the loss from a rate rise. A company that earns today feels little, while a company that promises earnings in a decade feels a lot. This example uses round numbers to show the mechanism and is not a forecast of any price.
How does the cost of borrowing reach AI data centre spending?
Data centres, chips, and power contracts are often paid for with debt or with leases. Higher rates raise the payment on each loan. A project that cleared its return target at one rate can miss it at another. Lenders then ask for more proof of revenue, and the weakest projects stop first.
This channel works on spending rather than on prices. The EQOS Asia AI study records about $200 billion invested each year in AI research and deployment, with 95% of projects failing in production (EQOS Asia AI study, April 2026). Costlier money makes that gap harder to carry, because each failed project now costs more. Projects that already earn revenue keep access to credit, while projects that only promise revenue wait longer for it.
What does the Fed rate path mean for the AI bubble?
The United States sets the reference price of money for most AI companies, since their shares and debts trade in dollars. Markets react to the expected path of rates as much as to each decision. A surprise matters more than a move that everyone expected.
The page on the Fed rate path and the AI bubble explains how to read expectations and which surprises matter. Because the Fed meets on a public calendar, this channel gives the clearest warning of the two. You can mark the dates in advance and review your plan after each one.
How does the yen carry trade link Asian funding to the AI bubble?
A carry trade borrows in a cheap currency and invests in assets that pay more. When the cheap currency strengthens or its rates rise, the trade loses money, and investors sell assets to repay. Those assets can include AI shares, so a move in Japan can reach a portfolio in Singapore or Hong Kong.
The page on the yen carry trade unwind and the AI bubble: the Asian channel covers how this works for Asian readers. This channel is faster and less visible than the Fed path, because positions are private. Currency moves are the public clue that deserves a quick check.
Which interest rate signals matter most for the AI bubble?
Three signals carry most of the information. The first is the direction of long-term yields, which set the discount rate. The second is the gap between short and long rates, which shows what markets expect. The third is credit spreads, which show how much extra lenders charge risky borrowers. A rise in all three together is a stronger warning than one alone.
Check these signals on a fixed day each month and record the readings. A single reading says little, while a trend over 3 months says more. Pair the rate signals with company signals such as order books and guidance, because rates explain the pressure and earnings show whether the business can bear it.
How can you test your AI bubble exposure against a rate rise?
Run a three-step stress test. First, list your AI commitments and the year each is expected to pay back. Second, reduce each payback by the discount effect of a one point rise in rates. Third, mark every commitment that no longer clears your target. Long-dated items fail first, so they deserve the earliest review. Short-dated items, such as a software subscription renewed each year, can wait for the next quarter.
The EQOS Asia Λ simulator lets you add a rate shock to your own case and see how the score moves. For a decision tied to large AI spending, a session with the EQOS Asia team applies the method to your numbers. This page is general information, not investment advice.
See also in this guide
Frequently asked questions
Do higher interest rates burst an AI bubble?
Higher rates lower the present value of future profits and raise borrowing costs, which can start a fall in a fragile market. They rarely act alone, and the EQOS Asia AI study places the enterprise AI market in the fragile zone at 0.337, so the pressure counts.
Why are long-dated profits hit hardest?
A profit far in the future is divided by a larger number when rates rise. In the worked example, a profit due in 20 years loses about 17% of its value from a one point rise, while one due in 10 years loses about 9%.
Which rate channel is easiest to watch?
The Fed rate path, because decisions follow a public calendar and markets publish their expectations. The yen carry trade is faster and less visible, so a currency move is the clue to check. Credit spreads add a third public signal.
What should a company do before rates move?
Run the stress test: list AI commitments with their payback years, apply a one point rise, and mark the ones that no longer clear the target. Stage the long-dated items and review the plan after each central bank meeting.
Test your knowledge
1. What does a rate rise do to the present value of a profit due in 20 years, in the worked example?
- Cuts it by about 17%
- Raises it by about 17%
- Leaves it unchanged
Show the answer
Cuts it by about 17%.
2. Which profits fall most when rates rise?
- Profits earned today
- The latest profits
- Profits due furthest in the future
Show the answer
Profits due furthest in the future.
3. Which channel follows a public calendar?
- The yen carry trade
- Private positions
- The Fed rate path
Show the answer
The Fed rate path.
4. What does a carry trade borrow in?
- A strong currency
- A cheap currency
- Gold
Show the answer
A cheap currency.
5. What is the third rate signal in this guide?
- Credit spreads
- Share volume
- Office rents
Show the answer
Credit spreads.
About the author
Stéphane Jambu
President, EQOS Forecast Asia
Stéphane Jambu leads EQOS Forecast Asia and works as a consultant in e-commerce, search visibility, and AI. LinkedIn profile
Sources: EQOS Asia AI study (eqos-asia.com/case-studies/ia-2026.html), EQOS Asia method page (eqos-asia.com/method.html), consulted on 22 October 2026. Page updated on 22 October 2026.