OpenAI’s rise has become one of the defining stories of the global artificial intelligence boom. Over the past year, the company behind ChatGPT has signed massive agreements with chip manufacturers, cloud providers, and data center developers, locking in the infrastructure needed to power increasingly advanced AI systems. The scale of those commitments, however, has stirred an uneasy debate: if the company’s ambitions outpace its finances, could OpenAI one day be viewed as too important to let fail?
Jason Furman, a Harvard economist and former senior White House adviser, dismisses that idea outright.
“Definitely not,” Furman said when asked whether OpenAI could ever merit a government rescue. “But if they do [go bankrupt], they’re not banks. They’re not too big to fail.”
Furman, who advises OpenAI part-time on labor issues, has spent decades studying economic cycles and financial crises. His resume includes stints during the Clinton administration as officials prepared for a potential dot-com crash and later as a key architect of the Obama administration’s response to the 2008 financial crisis. From that vantage point, he argues that the AI boom—despite its size and hype—does not pose the kind of systemic threat that would justify taxpayer intervention.
Why the AI Boom Looks More Like the Dot-Com Era
According to Furman, comparisons between today’s AI surge and the housing bubble that preceded the Great Recession are misplaced. Instead, he sees stronger parallels with the technology boom of the late 1990s.
“When the dot-com bubble burst, we had a shallow recession,” he said, noting that the fallout was limited and that long-term productivity ultimately improved. Some of today’s most influential technology companies were born or reshaped during that period.
The housing crash, by contrast, was devastating because mortgage-backed securities were deeply embedded in the financial system and widely believed to be safe. When those assumptions collapsed, banks, money market funds, and insurers were all dragged into the crisis.
“There’s not any really strong analogy of a bank run financial crisis that comes out of this bubble bursting,” Furman said of AI.
Technology stocks, he added, represent a much smaller portion of household wealth than housing once did. That limits the so-called “wealth effect” that can amplify economic downturns.
Valuations, Not AI Itself, Are the Bigger Risk
Furman is careful to separate excitement about AI’s capabilities from concern about how markets are pricing the sector. While he does not rule out the possibility of a technology bubble, he is more troubled by financial expectations baked into current valuations.
To justify those prices, he said, companies must show that the technology continues to improve and that those improvements can be turned into sustainable profits.
One challenge is that faster or more powerful systems do not always lead to proportional economic gains. “Every time your microchip in your computer gets two times as fast, you don’t write Word documents two times as fast,” Furman said, suggesting that AI could produce excess capacity without delivering matching productivity growth.
The second challenge lies in monetization. Investors are betting that AI companies can create products people will pay for and defend those offerings against cheaper competitors. “It’s not like I’m sure at all that there’s not an AI technology bubble,” Furman said. “But it’s the valuations I’m much more worried about.”
What Happens If OpenAI Stumbles?
Speculation has grown around OpenAI’s long-term obligations to data centers and energy infrastructure, raising questions about whether a pullback could ripple through the broader economy. Furman said he sees little evidence that such a collapse is imminent.
“I have no reason whatsoever to think OpenAI or any other company in this business is going to go bankrupt,” he said.
Even if a major AI firm did fail, he argued, the consequences would likely be manageable. Slower data center construction could reduce jobs and investment, possibly contributing to a mild recession. But those effects would be limited, and policymakers would still have tools to respond.
“It wouldn’t be a good thing,” Furman said, “but it does not feel like it would be catastrophically or memorably bad to happen to the macroeconomy.”
Government Support Raises Red Flags
Where Furman grows more uneasy is at the idea of government involvement. OpenAI briefly fueled speculation last month when Chief Financial Officer Sarah Friar referred to a possible role for the U.S. government in “backstopping” financing for AI infrastructure. Friar and CEO Sam Altman later clarified that she had misspoken and that the company is not seeking a bailout.
Nonetheless, OpenAI has lobbied to expand a 35% tax credit currently aimed at chip manufacturing to include AI data centers, servers, and power infrastructure. Separately, the Trump administration’s decision earlier this year to take a 10% stake in Intel underscored Washington’s growing willingness to intervene in strategic technology sectors.
“I worry if anyone talks about helping or taking equity stakes that also implies that if things turn south, the government would be there with a rescue or a bailout,” Furman said. “Government should have no money involved in this at all.”
Job Loss Fears May Be Overstated—for Now
Despite frequent warnings from AI developers about mass job displacement, Furman believes the labor market impact remains limited. He likened today’s predictions to earlier claims that automation would eliminate radiologists or truck drivers—claims that failed to materialize.
“Right now, it is an aide to existing actual human workers,” he said, stressing that adoption tends to be gradual and uneven across industries.
He also cautioned against reading too much into short-term economic data. Fluctuations in productivity or employment, he said, are noisy and do not yet show clear signs of AI-driven disruption.
“I really am reasonably confident we are not seeing it yet,” Furman said. “People keep overreacting.”
