Has OpenAI become too big to fail ?

Its CFO floated the idea of a government backstop to help finance its data centers

It took just one comment to set off alarm bells — a symbol of the growing fears of a bubble forming around generative artificial intelligence. Last week, Sarah Friar, OpenAI’s Chief Financial Officer, floated the idea that the US government might guarantee the massive debt the ChatGPT maker is preparing to take on in order to finance the enormous investments planned to multiply its computing power in the coming years.

The executive quickly backtracked, saying she had misspoken. “We don’t want government guarantees,” added Sam Altman. While OpenAI’s CEO insists that “taxpayers should not bail out companies that make bad business decisions,” the episode raises a deeper question: has the AI giant become too big to fail ? In other words, would Washington be forced to step in if OpenAI ran into trouble, just as it did to rescue the banks in 2008 ?

Eliminating risk

In recent months, OpenAI has committed to spending $1.4 trillion over eight years to acquire Nvidia GPUs and use cloud services from Oracle and Microsoft — an astronomical figure compared with the $13 billion in revenue expected this year. To fund these colossal projects, the startup won’t be able to rely solely on new fundraising rounds or even an IPO. It will have to take on substantial debt from banks and investment funds, Sarah Friar acknowledged.

In that context, a government guarantee from Washington would be a game changer: it would reduce or even eliminate risk for lenders. The benefits would be twofold. OpenAI could raise much larger sums — likely in the hundreds of billions of dollars, an unprecedented scale — while also securing significantly lower interest rates, Sarah Friar explained at a Wall Street Journal conference.

“A certain risk”

To sustain such high levels of borrowing, Sam Altman is betting on a sharp rise in revenue: $100 billion by 2027 and “hundreds of billions” by 2030. He’s banking on new lines of business, including an AI-focused cloud platform, robotics, and even a new device intended to replace smartphones. Still, the CEO admits there’s “a certain risk”, like the possibility of failing to meet obligations to partners or creditors.

The company’s financial equation doesn’t depend solely on revenue growth. Another key factor is the lifespan, and thus amortization, of the GPUs it buys, before they become too outdated to train or run state-of-the-art AI models. “If we can use them for six or seven years, financing is much easier, its CFO explained. But if that timeframe shortens, things get a lot more complicated.”

The flagbearer of AI

Following Sarah Friar’s comments, investor David Sacks, now advising Donald Trump on AI policy, declared that Washington would not bail out a struggling player. “The U.S. has at least five major companies developing frontier models. If one fails, others will take its place,” he said.

But OpenAI’s weight far exceeds that of its rivals. The company has become the symbolic standard-bearer of an industry whose success has justified staggering levels of investment. An OpenAI collapse would send a brutal signal: that these astronomical outlays were disconnected from AI’s true commercial potential.

Such a scenario could have a devastating impact—not only on the US economy, whose recent growth depends heavily on AI infrastructure spending, but also on stock markets supercharged by the performance of tech giants, starting with Nvidia. Would that be enough to justify a government rescue of OpenAI ?

Photo: Wall Street Journal (screenshot)