The 2008 crisis vs the AI bubble: which risks compare?

What were the main causes of the 2008 financial crisis?

The EQOS Asia 2008 back-test lists public data from before the crisis: US house prices up 124% from 1997 to 2006, a price to income ratio of 4.6 against a historical 3.2, and a subprime share of new mortgages at 23.5%, against 8% in 2003 (EQOS Asia 2008 back-test). Prices had left income behind, and lending had moved to weaker borrowers.

The Federal Reserve’s history adds the mechanism: mortgage credit reached high-risk borrowers, loans were repackaged into securities sold to investors, and risk looked small because quality gauges were based on prime mortgages (Federal Reserve History, Subprime Mortgage Crisis). When house prices peaked, losses spread through every holder of the securities.

How did leverage turn a small fall into the 2008 collapse?

The back-test records leverage of 30.7 to 1 at Lehman Brothers in 2007. At that ratio a fall of 3.3% in asset values erases all equity. The actual decline in 2007 and 2008 reached 20%, about 6 times the threshold. Lehman filed for bankruptcy on 15 September 2008 with $619B of debt (EQOS Asia 2008 back-test).

Leverage explains why a modest price fall became a crisis. A firm funded mostly by borrowing has almost no cushion, so lenders pull back at the first loss and force sales that push prices lower. Each step feeds the next. This feedback loop is the central risk to look for in any boom.

What did the 2008 financial crisis cost the economy?

The Federal Reserve’s account records a recession from December 2007 to June 2009, the longest since World War II. Real GDP fell 4.3%, home prices fell about 30% from their mid-2006 peak to mid-2009, and the S&P 500 fell 57% from October 2007 to March 2009 (Federal Reserve History, Great Recession). Unemployment rose from 5% to 10%.

The Fed cut its policy rate from 5.25% in September 2007 to a range of 0 to 0.25% by December 2008. The size of this response shows how deep the shock ran. For the AI bubble, the comparison sets the scale: a financial crisis spreads to jobs and output, while a technology correction can stay inside one sector.

Which 2008 risks compare with the AI bubble?

One risk matches clearly: prices above fundamentals. The EQOS Asia AI study records AI valuations up ×4 in 18 months and 95% of AI projects failing in production (EQOS Asia AI study, April 2026). A second risk to test is opacity. In 2008 few holders knew what the securities contained, and in AI few buyers can explain what a model does. The study puts AI explainability at 0.25 out of 1.

A third shared risk is dependence on a few firms. In 2008 a handful of banks sat at the centre of the network. In AI, a handful of chip makers, cloud providers, and model builders sit at the centre of the chain. When many customers depend on the same few suppliers, one failure travels far.

Which 2008 risks differ from the AI bubble, and which need checking?

The largest difference sits in leverage and in what is financed. The 2008 crisis ran through banks that held 30 to 1 leverage against mortgages. The AI build-out finances productive assets such as data centres, and the answer on leverage depends on the mix of debt and equity behind them. That mix is the number to find, and the study records about $200B a year invested without giving the split.

A second difference is the instrument. In 2008 the securities were complex and rated AAA in volume, with 80% of CDO tranches rated AAA in the back-test. AI financing may be simpler, or it may hide risk in contracts between suppliers and customers. Reading those contracts is the work that separates a 2008 pattern from a normal build-out.

What do the EQOS Asia scores say about 2008 and the AI bubble?

The EQOS Asia 2008 back-test scores the global financial system of 2006 and 2007 at φ(S) = 0.234, with 6 of 9 dimensions in the critical zone, about 2 years before Lehman. The AI study scores the enterprise AI market of 2026 at φ(S) = 0.337, in the fragile zone, with no dimension critical.

The scores say that the AI market is less damaged than finance was before 2008, and that it is still fragile. Communication sits at 0.305, just above the 0.3 critical line. A small shock, such as a public hallucination or a data leak, could push that dimension below the line. This is a reading of structure, and it gives no date.

How can you test whether the AI bubble carries 2008 style leverage?

Ask four questions of any AI exposure you hold. How much of the funding is debt? Who lends, and can they withdraw quickly? Do customers and suppliers owe money to each other in circles? How many months of obligations does cash cover? Answers that show high debt, fast withdrawal, circular dependence, and thin cash echo 2008.

The EQOS Asia method page shows how such answers combine into a recomputable score, the Λ simulator gives a φ(S) estimate in 10 minutes, and a 30-minute pre-diagnostic applies the scoring to a decision in Singapore, ASEAN, or Hong Kong. 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