Could one of the most prominent tech company leaders be less-than-enthused about the AI economy? In an interview, Amazon CEO Andy Jassy didn’t dismiss the idea that the AI bubble could pop, despite his company’s massive investments in the technology. 

Speaking to CNBC’s Becky Quick at the Davos World Economic Forum, Jassy hedged on whether he thought AI companies’ worth is being inflated by circular deals between model makers and infrastructure providers like his own Amazon Web Services.

Asked point blank by Quick whether high-level AI deals were circular, with the CEO-soothing caveat that “it doesn’t look like that’s what [Amazon] are involved in,” per Quick, Jassy said it’s all about both sides seeing an opportunity to make money. 

“The AI labs are consuming gobs and gobs of compute right now,” Jassy said. “And … they’re trying to find ways to fund that compute.”

The companies that fund that compute are happy to invest in the AI companies themselves if it means an opportunity to provide said compute – which sounds an awful lot like a circular deal despite Jassy not using the term – but he seemed to be okay with it because the AI companies are such impressive outfits.

“These are unusual companies that have been building amazing models for 10-plus years and are reshaping the way people use applications,” Jassy explained. “Companies have opportunities to invest in companies. Those companies are impressive companies and you’re just seeing some of that.”

That’s a lot of words to explain what could seemingly have been answered by saying “yeah, there’s a lot of circular deal making going on right now in the AI space.” 

Asked whether he thinks the current incestuous state of the AI industry could lead to a positive outcome, Jassy simply isn’t too sure.

“I think it could end well,” the Amazon CEO said. However, “anytime you make substantial investments … not all of them are successful. And the hit rate is pretty variable.” 

In other words, impressive companies or not, there seems to be a bubble brewing that’s liable to take a lot of capital with it once it eventually pops. Jassy isn’t even sure there’s a real need for all the compute infrastructure that’s being built to satisfy the AI firms‘ ever-increasing demands, even going as far as to call some high-profile deals hard to make sense of.

AI firms “believe they need quite a bit of compute to train the models the way they believe changes what’s possible,” Jassy opined when asked whether some of OpenAI’s trillions in recent infrastructure deals – one of which was signed with AWS – were realistic.  

“I don’t know the details of all those deals,” Jassy noted. “Sometimes when I get a chance to look at them I have a harder time making sense of them all … but I know that the companies believe they need that much compute.”

When it comes to compute power, Jassy doesn’t seem sure that the industry is going to be able to find enough energy to meet AI’s demands, either.

“There is a power shortage,” Jassy admitted. “I think that it is better than it was 18 months ago and still not as plentiful as we all need.” 

The Amazon CEO said that his company is trying to do “everything we possibly can to help try to enable more power” for its datacenters, including by making investments in next-generation nuclear projects that have yet to materialize.

Don’t assume all those indirect comments suggesting Jassy is unsure about AI mean he’s backing down from investing in the tech, though: It’s full speed ahead when it comes to AI at Amazon. 

Along with continuing to explore ways to expand its own datacenter operations, Jassy said Amazon also sees a future in which Amazon will continue to downsize its workforce thanks to AI innovations. 

“I do think that jobs are going to be impacted by what’s happening with AI over time … if you just think about how AI and agents will enable you to do coding and customer service and research and analytics,” Jassy said. So there’s four categories of Amazon drones that ought to worry about their future employment. 

Jassy echoed the usual language of company leaders facing the need to justify AI layoffs, explaining that some people will still remain in affected positions, and will be able to pawn off busywork on an AI in favor of doing higher-level, more advanced tasks. That, and AI will create jobs that don’t exist yet. 

“Fifteen years ago, there weren’t jobs like cloud architect, and whenever you go through any big technology transformation, you find new jobs,” Jassy said. “I think the same is going to be true here, medium to long term.”

Jassy also opined on President Trump’s tariff regimen, saying that it’s beginning to cause some Amazon vendors to increase their prices, as well as on the company’s relationship with Nvidia. 

“We have a deep relationship with Nvidia. We will for as far as I can foresee,” Jassy explained before adding that customers need better price performance than they can get with Nvidia chips, which is why Amazon pursued its own Trainium architecture. 

Trainium, we note, has benefited extensively from Nvidia architecture. 

“If you’re building a big inference business like we are and you want to have reasonable margins, if you’re not pursuing your own custom AI silicon, you’re going to be structurally disadvantaged,” Jassy said. 

Expect Nvidia to remain in the mix, though – after all, what’s the AI industry without some company-on-company circular investment? ®