What would happen if a $2 trillion AI bubble burst?

Market value is not GDP

If an AI company was valued at $100 billion and is later valued at $50 billion, the economy has not lost $50 billion of factories, code or output. Investors have changed what they are willing to pay for expected future earnings.

The first hit would be a repricing of AI-related stocks, infrastructure companies and privately held start-ups. If those losses spread to households, retirement savings and investment funds, they can weaken confidence and risk appetite. A Futu News report summarising IMF analysis said a correction in technology-stock valuations could lead to losses in household wealth.

Private companies are another pressure point. The World Economic Forum notes that during a bubble, new firms able to present themselves as AI-related can raise money cheaply and help drive real-economy investment. Once valuations fall, the next funding round becomes harder, hiring slows and projects that looked fundable at boom prices may be shelved.

The main GDP channel is an investment air pocket

The biggest direct effect on GDP would come from capital expenditure, or capex. If companies delay data centers, cancel server orders, trim software budgets or pause power-infrastructure projects, the hit moves through construction, suppliers, utilities and technology services.

That matters because the AI buildout has already become part of the growth story. The IMF says the IT investment boom has been concentrated in the United States but has also generated spillovers, especially for Asian technology exports. The demand chain includes servers, data centers, software and power infrastructure.

Data center construction itself boosts GDP while it is happening and creates jobs. But the World Economic Forum warns that in a bubble, the economic output from those investments can disappoint once the projects are finished. In plain English: building the site helps the economy today, but if future usage or profits fall short, the next wave of spending can vanish quickly.

A Futu News account of IMF analysis said a decline in AI investment could reduce global economic growth by about 0.4 percentage points. That is not a simple conversion of a $2 trillion market loss into GDP. It is the effect of weaker investment, lower asset prices and slower activity feeding through the economy.

The real fork in the road is credit

A stock-market correction is painful. A credit crunch is more dangerous. A credit crunch means lenders become reluctant to refinance old debts or make new loans, even to borrowers outside the original boom.

If the AI boom is mostly equity-funded, losses fall mainly on shareholders, venture investors and employees whose compensation depends on stock or options. That can be harsh, but it is less likely to threaten the wider financial system.

If, however, large AI infrastructure projects are debt-funded, the problem becomes cash flow. Can data centers, energy projects and AI supply-chain companies generate enough revenue to service debts, refinance loans and justify leases?

Oliver Wyman cites a J.P. Morgan estimate that more than $6 trillion in funding will be needed through 2030 for AI-related data centers, energy projects and the AI supply chain. The same analysis says an increasing share of that investment is debt-financed, with much of it in off-balance-sheet vehicles away from the cash-rich technology giants.

That is the tripwire. If collateral values fall, refinancing becomes harder and lenders take losses, financing conditions can tighten for companies that have little to do with AI. At that point, an AI correction becomes a broader economic problem.

Three possible versions of an AI bust

Jobs would weaken first in tech, then around it

Employment effects would probably begin with AI start-ups, software companies and infrastructure suppliers that hired for rapid growth. If funding becomes scarce, hiring freezes and layoffs are the obvious adjustment.

The second round would hit the surrounding economy: data center construction, electrical work, power infrastructure, equipment suppliers and professional services tied to the AI buildout. The logic is straightforward. AI demand has supported servers, data centers, software and power infrastructure, and building data centers is itself job-creating economic activity. If that spending stops, the jobs tied to it come under pressure.

A bubble burst would not mean AI is over

A valuation crash would not prove that AI is useless. It would prove that some expectations, prices or investment plans were too aggressive.

The bullish case still has substance. The World Economic Forum has cited research suggesting AI could potentially tackle $4.5 trillion worth of work across the United States. But potential work is not the same as realized profit, higher productivity or cash flow that can pay for expensive infrastructure.

After a bust, the market would likely separate useful AI from uneconomic AI. Projects that save money, raise revenue or improve productivity would keep going. Projects dependent mainly on cheap capital and optimistic valuations would be cut.

What to watch if the AI boom turns

Stock indices alone will not tell the full story. The more important signals are:

• Big technology companies’ AI capex guidance
• Data center and power-infrastructure delays or cancellations
• Server and AI supply-chain order trends
• Venture funding conditions for AI start-ups
• Credit spreads on corporate debt, private credit and infrastructure financing
• Signs that AI adoption is actually improving margins, revenue or productivity

The bottom line: a $2 trillion AI valuation crash would not automatically subtract $2 trillion from GDP. But because AI spending is now tied to infrastructure, jobs, supply chains and financing, the damage could become macroeconomic if investment freezes and credit tightens. The crucial questions are whether AI can deliver enough real value to justify the buildout, and whether the debt behind that buildout can survive a sharp repricing.