AI bubble checklist: ten tests, one verdict

How do the ten tests of the AI bubble checklist work?

Each test asks one question and returns one colour. Green means the boom reading holds, amber means the evidence is mixed, and red means the bubble reading holds. Write the source, the number, and the date beside each colour. A test without a dated number stays blank, which keeps the final count honest. The ten tests cover price, spending, funding, revenue per user, deployment, concentration, financing, asset life, speed, and structure.

The list serves one purpose: to replace a feeling with a record. Run all ten in one sitting, because mixed dates blur the picture. Save the sheet, and next quarter run the same ten again. The change between two sheets tells you more than any single verdict, since a rising count of reds is an early warning that one reading cannot give.

Which three tests check AI prices, spending, and funding?

Test 1 compares price with earnings: green when prices follow profits, red when prices run far ahead of them. Test 2 compares spending with revenue: green when revenue from AI buyers covers the build-out, red when it covers a small part. Test 3 traces the funding source. Green means customers pay for the growth. Red means new investors pay for the growth, which is the classic mark of a bubble.

Use dated sources for all 3: company filings, central bank reports, and the EQOS Asia AI study, which counts about $200B a year invested in AI research and deployment (EQOS Asia AI study, April 2026). Tests 2 and 3 usually decide the verdict, because they show whether the money loop closes. If customers repay the spending, a fall in prices stays a correction. If they leave a gap, the price has to close it.

How do revenue per user and project success test the AI bubble?

Test 4 measures revenue per user: green when each paying user brings more revenue each quarter, red when growth comes only from more free users. Test 5 measures deployment: green when most pilots reach production, red when most stall. The EQOS Asia AI study cites MIT 2025 on 95% of generative-AI projects failing in production, so test 5 reads red for the enterprise market as of October 2026.

Test 5 deserves a local version. Count the pilots in your own organisation and the share that reached production last year. The study ranks the causes of failure: insufficient or biased data at 42%, no alignment with strategy at 31%, and team resistance at 24%. Fixing the top cause moves your own sheet from red toward amber, whatever colour the market shows.

Which balance sheet tests belong in an AI bubble checklist?

Test 6 measures concentration: green when market gains spread across many companies, red when a handful of AI names carry the index. Test 7 reads financing: green when companies fund AI from cash flow, red when debt and complex structures carry the build-out. Test 8 checks asset life: green when chips and data centres keep their value through the payback period, red when they lose value faster than they repay.

These 3 tests change fastest, so date each reading to the day. Annual reports, bond filings, and depreciation notes hold the numbers. Each one rewards a careful reading of a few pages instead of a long study. With limited time, run tests 7 and 8 first, since financing and asset life describe whether the build-out can repay itself.

How do speed and structure complete the AI bubble checklist?

Test 9 measures speed: green when prices rise in step with earnings, red when they multiply faster than any revenue could follow. The EQOS Asia AI study records valuations up ×4 in 18 months, so test 9 reads red. Test 10 reads structure through the φ(S) score, which the study sets at 0.337 for the enterprise AI market, in the fragile zone where a shock can tip everything.

Recompute test 10 for your own case on the Λ simulator, free and anonymous, in about 10 minutes (EQOS Asia AI study, April 2026). The study sets 0.5 as the viability threshold. A score below it reads red, a score near it reads amber, and a score above it reads green. The market score and your own score can differ, and the difference is useful information for your decision.

How do you turn ten AI bubble tests into one verdict?

Count the reds. With 0 to 3 reds, the boom reading leads, and plans can continue with a quarterly review. With 4 to 6 reds, the evidence is mixed, and staged commitments make sense. With 7 to 10 reds, the bubble reading leads, and exposure deserves a size check. This scoring rule belongs to this page, and the thresholds are choices you can adjust and record.

As of October 2026, a first pass with the study figures gives red on tests 5 and 9 and a fragile reading on test 10. The other seven need current filings. Fill them in before you read the count. A partial sheet shows direction, and a complete sheet gives the verdict. The verdict is a probability reading and leaves the timing of any fall open.

What should you do with the AI bubble checklist result?

Link each colour to a decision. A red on funding calls for staged purchases. A red on asset life calls for shorter contracts. A red on deployment calls for a pilot review before the next budget. Decisions attached to colours turn the sheet into a plan. Share the sheet with your board or investment committee, so that the same facts frame every discussion.

EQOS Asia can score your case with the same method. If your decision involves AI spending in Singapore, ASEAN, or Hong Kong, a 30-minute pre-diagnostic returns a score and a ranked set of options, and the method page shows how the score is built. This page is general information, not investment advice.

Frequently asked questions

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