AI bubble could spark 2008-like crisis, Gary Marcus warns

A warning about AI’s financial foundations

The failure of a leading artificial intelligence company could send financial stress beyond the technology sector and potentially produce a crisis resembling 2008, AI critic Gary Marcus has warned. For India, where enterprise adoption of AI must be balanced with financial prudence, the warning raises a question about the durability of the businesses supplying the technology.

Speaking to Economic Times Markets during a visit to India this week, the American cognitive scientist and entrepreneur questioned whether OpenAI could justify its spending. He suggested that the company might become an AI-sector counterpart to WeWork, eventually being acquired at a steep discount rather than completing its hoped-for public listing.

These are Marcus’s forecasts, not established outcomes. The supplied report does not demonstrate that OpenAI faces imminent failure, identify a bank suffering AI-related losses or establish that a Microsoft acquisition is under discussion.

OpenAI, Microsoft and the WeWork comparison

Marcus floated a scenario in which Microsoft could acquire OpenAI for $200 billion, against a possible $1 trillion valuation sought through an initial public offering. Neither figure is presented as an agreed transaction price. His argument is that spending commitments could become difficult to defend if investors lose confidence in the business.

The report also says OpenAI has delayed its planned IPO until next year while targeting a $1 trillion valuation. It does not supply a company announcement or a detailed listing timetable, so that remains a reported plan rather than a confirmed market event.

The WeWork analogy draws on a documented reversal. Once valued at $47 billion, the office-sharing company abandoned its planned 2019 IPO amid concerns about losses and governance, and filed for bankruptcy in 2023. Marcus uses that history to illustrate how private-market enthusiasm can unravel when a business faces closer scrutiny.

The comparison does not establish that OpenAI has the same financial weaknesses or will follow the same path. Its relevance is narrower: a large valuation alone cannot guarantee a successful listing or a sustainable enterprise.

How a sector setback might spread

Marcus argued that even one major failure in the AI ecosystem could produce international contagion. The report links that concern to views among some unnamed US policymakers that borrowing for AI investment has contributed to rising bond yields.

However, it gives no figures for banks’ exposure, no breakdown of borrowing by AI companies and no estimate of losses that lenders could face. Those omissions matter when assessing the comparison with 2008: a sharp technology-sector correction and a systemic banking crisis are different propositions.

The warning therefore identifies a possible transmission route—investment financed by borrowing—but does not establish its scale. Evidence about creditors, financing terms and concentrations of exposure would be needed to judge whether trouble at a particular company could threaten the wider financial system.

Marcus also challenges AI safety claims

His criticism extended beyond valuations. Marcus accused Anthropic of inconsistency for warning about AI risks and advocating a slower pace while, according to his account, pursuing a $2 trillion IPO. The supplied text does not independently establish that listing plan or include a response from Anthropic.

He also advocated recalling autonomous AI agents, describing them as unsafe, and argued that personal agents would intensify the risks. He distinguished between human extinction, which he considered unlikely, and catastrophe or dystopian outcomes, which he viewed as more probable. These are his risk assessments; the report supplies no quantified supporting analysis.

Marcus is a professor emeritus of psychology and neural science at New York University. He founded Geometric Intelligence, acquired by Uber in 2016, and later co-founded Robust.AI. In 2023, he testified on AI regulation before the US Senate alongside OpenAI chief executive Sam Altman.

Analysis: India’s test is resilience, not hype

For Indian businesses, the practical lesson is to distinguish a useful technology from the financial durability of its supplier. Marcus’s warning is not evidence that Indian banks are exposed to an impending AI crash; the report identifies no such exposure.

A proportionate response would preserve competition and innovation while examining supplier dependence, contractual protections and the ability to switch services. That approach avoids both unquestioning enthusiasm and restrictions based only on alarming predictions.

What matters next is evidence: confirmed listing disclosures, financing commitments, spending obligations and clearer information about lender exposure. Until those emerge, the proposed fire-sale and banking-crisis scenarios should be treated as warnings to examine—not outcomes to assume.