AI bubble soft landing: what a gentle deflation looks like

What does an AI bubble soft landing mean in plain words?

A soft landing is a gentle deflation. Prices move sideways or fall slowly for several quarters, revenue grows, and the gap between valuation and earnings narrows without a crisis. No large group of holders is forced to sell, no major lender fails, and funding keeps flowing at lower prices. Banks, insurers, and pension funds keep their positions, which keeps the financial system calm.

The name comes from aviation: the plane descends without damage. In markets, the descent is measured in years. The AI bubble soft landing keeps the technology and most of its leading firms, protects the jobs and contracts tied to them, and it removes the excess in valuations through time and earnings, not through a sudden fall. A reader can think of it as a long, planned descent in which every stage has a checkpoint.

Which conditions allow an AI bubble soft landing?

Four conditions matter. First, earnings growth: customers must pay for AI products at a growing rate. Second, funding from customers, not only from new investment rounds. Third, low leverage, so that a price decline does not trigger margin calls. Fourth, orderly supply of shares and chips, so that no wave of sales lands in one quarter. Lock-up expiries and large share sales are the usual sources of such waves, and both appear on public calendars.

Each condition can be observed. You can read revenue growth in company reports, funding sources in deal announcements, leverage in credit data, and supply in listing calendars. When three of the four hold, a soft landing becomes the most likely path. When two fail, the odds shift toward a harder ending.

What are the early marks of an AI bubble soft landing?

Look for price indexes that go flat for two or three quarters while reported AI revenue keeps rising. Look for funding rounds closed at stable or slightly lower prices, with no cancelled rounds. Look for mergers among weaker firms, and for projects that move from pilot to production at a better rate.

A quiet news cycle is a mark too. Soft landings produce few headlines, because nothing breaks. Readers who expect drama may miss the signal. A monthly note that records revenue growth, round sizes, and the count of pilots that reach production gives you a plain picture of the landing. Share the note with your board each quarter, so that the same facts frame every discussion.

Who gains and who loses in an AI bubble soft landing?

Firms with paying customers and disciplined costs gain, because time works for them: revenue catches up with their valuations. Buyers of AI services gain too, as suppliers compete on price and quality. Patient long-term holders gain when earnings close the gap instead of prices falling to meet it. Suppliers of useful tools, such as data quality and monitoring services, also gain, since every firm needs them.

Firms that depend on constant new funding lose, since flat prices make new rounds harder. Projects without a business case lose their budgets. The EQOS Asia AI study shows why: 95% of AI projects fail in production, and ROI not demonstrated explains 22% of those failures (EQOS Asia AI study, April 2026).

How does the EQOS Asia study describe the route to a soft landing?

The study scores the enterprise AI market at φ(S) = 0.337 and gives 4 corrections that raise a company’s own score: measure data quality before deploying, require explainability, audit the gap between AI strategy and the field, and fund retraining at 2 to 5% of AI investment. After the 4 corrections, φ(S) reaches 0.401.

The predicted failure rate drops from 94% to 82%, which the study converts into about $80 billion of recoverable value per year for the world market. The route to a soft landing is therefore a set of company decisions added together. Each firm that applies them moves the market’s average a little toward the soft branch. The four steps cost little compared with the sums lost on failed projects.

What should a business do during an AI bubble soft landing?

Use the quiet period to fix foundations. Audit the signal and noise in your data before any new deployment, since a score under 0.5 means AI amplifies noise. Prefer models that explain their decisions for critical choices, and keep large language models for assistance. Retire pilots that never reached production, and move their budgets to projects with a named owner and a measurable return.

Keep spending tied to revenue, and write a retraining plan for every role that automation changes. These steps make your results stronger in a soft landing and your firm more resilient in a harder ending. This page is general information, not investment advice.

How can an AI bubble soft landing turn into a break?

A soft landing is a branch, and events can move the market off it. The study places Communication at 0.305, only 0.005 above the stop threshold of 0.3. A public incident, such as a biased decision or a data leak, could push it across and pull Trust down with it.

Watch that dimension and the four conditions each month. To test your own case, the Λ simulator returns a score in about 10 minutes, and the EQOS Asia method page explains the calculation. For Singapore, ASEAN, or Hong Kong decisions, the pricing page lists the diagnostic formats.

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