Marcus cautions that the OpenAI IPO might jeopardize the AI supply chain.

      TL;DR Gary Marcus cautions that OpenAI’s difficulties with its IPO could trigger a ripple effect throughout the AI supply chain, affecting Nvidia, Oracle, and CoreWeave. With enterprise AI investment already declining and OpenAI losing $3.7 billion quarterly, the AI researcher perceives a credit event risk that hasn't been appropriately factored in.

      Marcus has long been warning that the AI sector is built on unstable foundations. He now identifies a particular sequence of consequences: if OpenAI’s IPO performs poorly, it will impact much more than just one company. “The valuations of these companies depend heavily on the expectation that OpenAI will have a huge demand for chips and data centers,” Marcus told Business Insider. “They will likely face issues if they fail to secure public funding.”

      The dependency chain

      Marcus’s argument is based on the observation that Nvidia, Oracle, and CoreWeave have greatly benefited from OpenAI’s demand for computing power. Last year, OpenAI spent $34 billion and is projected to incur around $27 billion in losses by 2026, according to its financial reports.

      CoreWeave, which has recently become a part of the Nasdaq-100, earns a substantial portion of its income from OpenAI workloads, mainly through Microsoft. If OpenAI were to reduce spending following a poor public offering, those providing its infrastructure would lose a key client.

      OpenAI is currently contemplating significant price reductions to compete against Anthropic, which would further distance it from profitability. The company does not anticipate achieving positive cash flow until 2030.

      The WeWork parallel

      Marcus has frequently compared OpenAI to WeWork, the co-working firm that saw its stock plunge 99% before declaring bankruptcy in 2023. “WeWork was valued at an incomprehensible level in relation to its core metrics, seeming more focused on appearance than reality,” he commented.

      “I often perceive OpenAI similarly, recognizing that they lack a substantial technical advantage, and as competitors close the gap, price wars will arise, making profitability difficult.” Anthropic’s confidential IPO filing at a $965 billion valuation, exceeding OpenAI’s own $852 billion, highlights the competitive pressures Marcus refers to.

      The tokenmaxxing hangover

      Marcus’s warning comes at a time when the AI investment surge is already revealing weaknesses. Companies like Uber, Meta, and Amazon have begun to restrict AI usage among employees after finding that high token usage failed to yield tangible returns.

      “The decline of tokenmaxxing is forcing OpenAI to think seriously about slashing costs,” Marcus noted. “While this could help retain users, it drives them even further from profitability.” OpenAI’s revenue for Q1 2026 was $5.7 billion, yet it faced a cash burn of $3.7 billion. The company is also dealing with an investigation by 42 state attorneys general and a lawsuit from Florida, adding legal uncertainty to its pre-IPO situation.

      The blast radius

      What worries Marcus the most is the systemic risk involved. If OpenAI’s IPO struggles, lenders supporting AI infrastructure might start to question the creditworthiness of the entire asset class.

      “Nobody really knows the extent of the impact, as we cannot ascertain how serious the consequences would be for lenders,” he said. For investors in Nvidia, Oracle, and CoreWeave, the pressing concern shifts from whether OpenAI will go public to what will happen to all related entities if the IPO underperforms.