aswath damodaran, a finance professor at new york university, recently issued a stark warning: if the current valuation bubble in the ai industry bursts, its systemic impact could far exceed that of the 2000 dot-com bubble. unlike the earlier “.com” boom—characterized by young companies and low capital expenditures—this ai frenzy relies heavily on astronomical hardware investments, ranging from massive data centers to custom-built chip clusters, with vast sums leveraged through debt financing. this means the risks are no longer confined to equity investors; instead, they could propagate through credit networks, reaching the banking system and even the real economy.
at the business model level, damodaran sharply points out that ai is not software in the traditional sense where marginal costs approach zero. each model invocation entails actual computational power consumption and energy expenditure, making its cost structure more akin to pay-per-use streaming services like spotify than subscription-based platforms such as netflix, which spread fixed content costs across a large user base. this inherent nonlinear cost curve renders ai companies’ paths to profitability exceptionally narrow—especially as the industry has already sunk into a price war, with emerging players like deepseek rapidly driving down service rates and further eroding already slim profit margins.
most thought-provokingly, he raises a paradoxical warning: even if all the most ambitious visions for ai were realized—namely, systems reliably replacing half of white-collar jobs—the social costs would be unprecedentedly heavy. the “productivity revolution” narrative underpinning high valuations, when translated into reality, could instead trigger widespread structural unemployment and income disparities. the scholar, renowned for his rigorous valuation framework, concludes by emphasizing: “we’re not debating whether the technology can succeed; we’re assessing whether society is ready for its success.”
