Why do experts disagree about the AI bubble?

Why do experts disagree about the AI bubble across time horizons?

Time horizon explains the largest share of the disagreement. An investor who measures 12 months sees stretched prices and heavy spending. A researcher who measures 20 years sees a general-purpose technology that reshapes industries. Both can be right, because a bubble and a lasting technology can coexist, as the railways and the internet showed. When two experts quote different horizons, ask each one for the date range behind the claim before you compare conclusions.

Horizon also changes what counts as a loss. A fund that reports each quarter treats a sharp price drop as a failure. A company that plans for 10 years treats the same drop as a chance to buy hardware and talent at lower cost. The market move is identical, and the verdict differs because the mission differs. Write down the horizon of every source you read, starting with your own.

How does the definition of an AI bubble change the verdict?

Definitions split the camp next. Some experts define a bubble as prices above fundamental value. Others define it as a crash that has already happened. A third group defines it by credit: a bubble exists when borrowed money funds the buying. Under the first definition, many experts say yes today. Under the second, every expert waits for hindsight. The word carries a different test for each speaker, so the argument often stays about words.

Settle the definition before the debate. The plain-English definition of the AI bubble used in this series reads: AI prices and spending that run ahead of the revenue behind them. With that sentence on the table, the question becomes measurable. You compare spending with revenue and read the gap. Experts who accept a shared definition usually narrow their disagreement to a few numbers.

Which AI bubble numbers do bears and bulls each quote?

Data differs even when the headline is shared. The EQOS Asia AI study records valuations up ×4 in 18 months, about $200B a year invested, and 95% of projects failing in production (MIT 2025, cited in the EQOS Asia AI study, April 2026). The same study projects +40% productivity gains and 300 million jobs that change. A bear quotes the first group of figures. A bull quotes the second. Both quote the same page.

Each figure measures a different layer. Valuation measures what investors pay. Project failure measures what companies manage to deploy. Productivity measures what the technology reaches in the best case. Mixing layers produces confident claims that nobody can compare. Line up the figures by layer, and the disagreement shrinks to one link between layers: how fast deployment catches up with price.

How do roles and incentives shape what experts say about the AI bubble?

Roles shape emphasis. An AI lab raises capital, so its leaders describe a long runway. A short seller earns when prices fall, so the research describes weakness. A bank advises both sides. This describes how work tilts attention, and every source has a position. Reading arguments instead of reputations keeps your view independent. The guide on who says AI is a bubble names the voices and sets out the reasoning behind each one.

Test any claim with 3 questions. Who gains if you believe it? Which number supports it, with which date? What result would change the speaker’s mind? A source that states its own disproof deserves more weight than a source that offers none. Apply the same 3 questions to your own view, and you gain a calibrated starting point for every later decision.

Which bear arguments about the AI bubble carry the most weight?

The bear side rests on 3 arguments: spending that outruns revenue, projects that stall before production, and valuations that rise faster than earnings. Each argument points to a gap that must close, and the bear expects the closing to come as lower prices. The detail sits in the case for an AI bubble, with the strongest version of each argument and the data that would weaken it.

Bears hold evidence you can check. The EQOS Asia AI study records 95% of projects failing in production and a φ(S) score of 0.337, in the fragile zone where a shock can tip the system. A bear reads those figures as a sign that demand lags the build-out. The strongest bear case accepts that the technology is real and argues about price and pace.

How do bulls answer the AI bubble warnings with earnings?

The bull side rests on 3 arguments as well: productivity gains of +40%, real demand from businesses, and a technology that changes 300 million jobs (EQOS Asia AI study, April 2026). Each argument points to revenue that arrives later and larger. The case against an AI bubble sets out the strongest version of each argument, together with the data that would weaken it.

Bulls hold evidence you can check too. Adoption grows, and each year of use teaches companies where AI pays. A bull reads the 95% failure rate as the normal learning phase of a new tool: early projects fail, and later projects use the lessons. The strongest bull case accepts the price risk and argues that earnings will grow into it.

How can you weigh expert views on the AI bubble yourself?

A shared method narrows the gap. Take the 5 sources above: horizon, definition, data layer, position, and evidence. Score each side on all 5, then compare. Two experts who share a horizon and a definition usually disagree on one number, and that number can be tracked each quarter. The AI bubble checklist turns the comparison into ten tests with one verdict.

EQOS Asia adds a measurement to the debate. The φ(S) score combines many signals into one value that anyone can recompute on the Λ simulator. If your decision involves AI spending in Singapore, ASEAN, or Hong Kong, a 30-minute pre-diagnostic applies the score to your case. 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