Aswath Damodaran, a finance professor at NYU, warns that an AI industry collapse would inflict deeper damage than the dot-com bubble burst because companies are financing massive physical infrastructure rather than building lightweight software.
During the dot-com crash, investors lost money on intangible assets. AI companies are different. They're pouring billions into data centers, chips, and hardware. This debt-financed infrastructure creates real financial obligations that persist even if the industry implodes. When companies default, creditors seize hard assets worth far less than their purchase price, triggering cascading losses across financial systems.
Damodaran identifies a second problem that exists independent of any crash scenario. Even if AI succeeds commercially, its core business model replaces human workers at scale. The economic disruption of mass job displacement remains unresolved. No consensus exists on how societies absorb that shock, whether through retraining programs, universal income, or other mechanisms.
The comparison to 2000 is instructive. Dot-com companies burned through venture capital on websites and software that evaporated overnight. The physical losses were minimal. AI companies face different economics. OpenAI burns through billions annually on compute. Microsoft and Google are constructing new data center complexes. Meta is building specialized AI chips. These expenditures represent real capital commitments that can't be instantly written off.
If AI hype deflates and adoption stalls, investors face a reckoning. Hundreds of billions in infrastructure sits idle. Debt obligations remain. Banks and venture funds holding those assets take losses. The ripple effects spread to traditional financial markets, particularly if major tech companies overlevered themselves to fund AI buildouts.
Damodaran's analysis cuts through startup optimism. He acknowledges AI's transformative potential while highlighting a blind spot in current valuations. Market pricing assumes either success or manageable failure.