OpenAI, the firm behind the revolutionary ChatGPT, is making a financial bet so large it’s shaking Silicon Valley. The core of the issue is a staggering $1.4 trillion question: how can a loss-making startup afford such a massive spending commitment on computing power over the next eight years? This figure eclipses its current annual revenues, which CEO Sam Altman claims will end the year above $20 billion, creating a chasm that has investors increasingly nervous about a potential AI bubble.
The sheer cost of “compute”—the AI infrastructure like chips and servers needed to train and run models—is the crux of the problem. This spending gap became the tense backdrop for a recent podcast exchange between Altman and one of his key investors, Brad Gerstner. When Gerstner called the $1.4 trillion cost versus $13 billion in revenue a question “hanging over the market,” Altman snapped, correcting his revenue figure and curtly offering to find a buyer for Gerstner’s shares, finishing with “enough.”
The company’s financial messaging stumbled again when its new CFO, Sarah Friar, suggested the U.S. government could potentially underwrite chip spending. This sparked immediate backlash and comparisons to the 2008 bank bailouts. Friar and Altman both scrambled to clarify, with Altman posting on X that OpenAI “do[es] not have or want government guarantees” and that taxpayers should not bail out bad business decisions, suggesting instead that the government build its own AI or support chip manufacturing.
Altman’s core defense is one of overwhelming future demand. He argues that with 800 million weekly users and 1 million business customers, the risk of not having enough computing power is far greater than the risk of having too much. He is betting that growing demand for paid ChatGPT versions, new hardware, and “huge value” from AI-driven scientific research will cover the costs, projecting revenue in the “hundreds of billions” by 2030.
This, however, remains a monumental gamble. Skeptics like Carl Benedikt Frey of Oxford University point to U.S. Census Bureau data showing a recent decline in AI adoption among large companies, suggesting the initial hype may be fading. Frey is doubtful OpenAI can reach $100 billion in revenue by 2027 without new breakthroughs. Altman acknowledges the risk, conceding that if he’s wrong, “the market—not the government—will deal with it.”