Is Sam Altman Gambling With the U.S. Economy?

Welcome back to In the Loop, TIME’s new twice-weekly newsletter about AI. We’re publishing these editions both as stories on Time.com and as emails. If you’re reading this in your browser, why not subscribe to have the next one delivered straight to your inbox?

Subscribe to In the Loop

What to Know: OpenAI Bets Big

High stakes — Sam Altman is going all in. In an interview following Monday’s deal with AMD—which brought OpenAI’s total deals this year to a value of roughly $1 trillion—Altman said that OpenAI was making a “company-scale bet” that it is now the right time to “spend a lot on infrastructure.”

OpenAI’s CEO is hoping that the more chips the company throws at training AI models, the more useful they’ll become—and the more people will be prepared to pay for them. The deals give OpenAI chips through chipmakers AMD and Nvidia, and data centers through cloud-computing provider Oracle. This will run an eye-popping 20 gigawatts’ worth of computing power, which will be used to train and deploy OpenAI’s models. The first portions of this are due to come online in 2026.

Missing power, uncertain payoffs — One missing piece of the puzzle is how OpenAI will power all of these chips. It would take 20 nuclear reactors to supply 20 gigawatts of power, but the AMD and Nvidia deals include no mention of where all the electricity will come from. The U.S. already faces a severe electricity shortage in the face of growing AI demand.

The more ominous uncertainty is how OpenAI will pay off its increasingly convoluted bets. “It’s very rational to pose that question, because the amount of money being spent here is enormous,” says Richard Shannon, an analyst at Craig-Hallum. Critics worry about the circular nature of the deals: Nvidia invests in OpenAI, which spends the money on Nvidia chips; OpenAI gets a stake in AMD in return for buying AMD’s hardware. Nvidia “makes such huge margins” that they can afford to make these investments, says Shannon.

OpenAI, on the other hand, operated with a loss of $7.8 billion in the first half of 2025. They need a return on their investment in AMD, “or it’s going to be very difficult,” says Shannon.

The company would need to grow its revenue from around $13 billion in annualized revenue to at least $300 billion by 2030 to justify its current spending. If OpenAI’s exponential revenue growth from the last two years holds, it will hit that target in 2028, meaning Altman’s current bets could look genius in retrospect. But if the exponential curves falter, the bets—which now entangle most of the world’s most valuable companies—might cost a lot more than one company, and threaten the entire economy.

Rarer earths — On Thursday, the Chinese government sent a tremor through these plans when it announced export curbs on rare earth metals, which are essential to powering AI systems. “If enforced aggressively, this policy could mean ‘lights out’ for the US AI boom, and likely lead to a recession/economic crisis in the US in the short term,” Dean Ball, a former AI advisor for Trump’s White House, wrote on X.

It will certainly be hard for Altman to achieve exponential revenue growth if he struggles to obtain the materials necessary to power his systems. He and the rest of the AI industry will need some support from President Trump—who later this month will meet with Chinese leader Xi Jinping in South Korea for a crucial conversation about the ongoing trade tensions between the two superpowers.

Who to Know: Nathan Benaich

A snapshot in time — Benaich, a VC at Air Street Capital, co-authors the annual State of AI report, now in its eighth year. The 2025 edition, which was released yesterday, is an exhaustive examination of AI across research, industry, and politics. This year’s edition finds that general AI usage has massively increased. In a survey of 1,000 “highly-educated adult professionals,” 95% use AI in their professional and personal lives. Coding is the biggest use case, followed by content generation and knowledge retrieval. There are a lot of other interesting passages in the report if you have an hour (or ten!).

The rise of AI NIMBYism — Every year, Benaich and his team make a series of predictions for the next calendar year. (They got five out of 10 right in 2024.) One of their 2026 predictions is that “data center NIMBYism”—the idea that community backlash against nearby data center construction will “take the US by storm,” even swaying certain 2026 campaign races.

“There are many reasons to fear or hate AI data centers,” Benaich writes to TIME in an email, calling attention to spikes in electricity rates and watershed stress. He adds that because there are few local or federal restrictions, community backlash has become an “effective recourse” to slow data center buildouts.

However, Benaich says that while many of the community issues with data centers are legitimate, “the US can’t afford to offshore DC development and lose out on the geostrategic benefits. This wave of investment has arguably kept the US out of a recession.”

Bubble? — Benaich’s State of AI report also calls attention to the growing number of “circular AI deals,” as detailed in the first section of this newsletter. However, he waved off the idea that we’re in a dangerous AI bubble. “Unit economics appear sound, and circular investment remains a small portion of total investment,” he wrote to TIME.

AI in Action

Did the world’s biggest pop star use AI tools for her album rollout? Earlier this week, Taylor Swift posted a treasure hunt for fans consisting of a series of short videos. But eagle-eyed fans found traces of what they believe to be generative AI use, including strange finger movements. While Swift has yet to clear up whether the videos were indeed AI-generated, fans responded unhappily, posting the hashtag “#SwiftiesAgainstAI.” The videos in question were later removed from YouTube shorts.

What We’re Reading

Enshittification, Cory Doctorow