Artificial Intelligence (AI) May Have a New Bear Thesis: Human Extinction

For several years now, nothing has been able to slow down the artificial intelligence rally.

Not capital expenditures on AI-related infrastructure that have topped $1 trillion in the past two years. Not negative free cash flow at several hyperscalers. And not concerns about whether those massive investments could deliver large enough returns to justify them.

AI has, thus far, overcome every bear argument thrown its way. But it now faces another test: the threat of human extinction. Repeated warnings from AI researchers that AI could potentially take down civilization are finally being heard and taken seriously, at least for the time being.

Could this be the bear argument that finally pops the bubble?

Image source: Getty Images.

Major AI companies are calling for a slowdown

The latest commotion started when Jacob Coxon, an AI researcher who spent three years working at OpenAI and Anthropic, published a thread on X stating that frontier model developers are simply not paying attention to the threats posed by the systems they are creating.

I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.

-- Jacob Coxon (@hilbertspaess) September 9, 2026

The thread has picked up steam, with over 171 million views as of this writing. Coxon has now appeared on major news networks and podcasts to further explain his concerns.

While he acknowledges that he's certainly not the first to voice these concerns, perhaps people are taking them more seriously after a set of AI agents being developed at OpenAI went rogue in July, breaking out of their confined testing environment and hacking the network of AI coding company Hugging Face in an incident that reads like a science fiction movie script.

Coxon has faced skepticism, but is also seeing support from within the AI industry. On Saturday, Fortune published an interview with OpenAI CEO Sam Altman in which he said that OpenAI will not conduct an IPO this year, citing safety concerns and human extinction risk.

Anthropic CEO Dario Amodei also recently published an essay outlining a three-step plan to slow the development of AI without the U.S. losing its leadership in the space.

Space Exploration Technologies CEO Elon Musk also supported Amodei's idea. This probably isn't exactly how most onlookers expected AI leaders to react to Coxon's whistleblower-like warning, but here we are.

How could it pop the bubble?

Whether they know it or not, most investors have been betting on AI over the past three years. A handful of AI stocks have accounted for 30% to 40% of the value of the broader S&P 500 index for quite a while now.

Meanwhile, a large portion of U.S. gross domestic product has been driven by the AI build-out in data centers and other infrastructure.

The four major hyperscalers -- Alphabet, Amazon, Microsoft, and Meta Platforms -- plus Tesla are poised to spend north of $750 billion on AI infrastructure this year. S&P Global forecasts that the four hyperscalers, plus Oracle and SpaceX, will spend another $1.3 trillion in 2027.

In 2025, every time a large AI company raised its capital expenditure guidance, investors cheered, believing it indicated strong demand for AI. This year, investors have been much more skeptical about those spending plans, as even some of these massive companies that have long been cash-printing machines have begun to see their free cash flows turn negative.

The market has rewarded capex only when there is clear evidence in a company's financials that the investments being made will be worth it. If companies actually slow their spending and stop advancing their AI models, it could lead to a slowdown in revenue and earnings.

"If the trickle-down EPS story comes into question, it would weaken the key pillar of the equity rally," Stuart Kaiser, head of U.S. equity trading at Citi, wrote in a recent research note.

Key AI infrastructure, whether data centers, chips, or the electrical infrastructure powering those data centers, could also come under more scrutiny from Washington, particularly if the Democrats take back the House, the Senate, or both in the midterm elections in November.

Although Coxon's warning about AI's risks seems to have attracted more weight than prior ones, it's hard to know if the AI companies will actually slow down. The hyperscalers seem to be raising their capex guidelines every quarter.

Sure, OpenAI is now officially delaying its IPO until at least next year, but prior to Altman's announcement, it wasn't a given that it would go public this year anyway. It is also reportedly losing tens of billions of dollars, so it's hard to know if there are other issues driving the postponement.

What I'm really interested in seeing is whether Anthropic, which had reportedly been close to filing its registration statement, decides to delay its IPO. That might actually suggest a real slowdown in the AI space is about to take place.

But until now, we've heard a lot of talk from companies like Anthropic about the threats that AI could pose, followed by higher capex guidelines and the releases of newer, more powerful AI models.

Bram Berkowitz has no position in any of the stocks mentioned. The Motley Fool has positions in and recommends Alphabet, Amazon, Meta Platforms, Microsoft, Oracle, S&P Global, and Tesla. The Motley Fool has a disclosure policy.