By Jesus Mesa
Reporter
0
Share
Newsweek is a Trust Project member
See more of our trusted coverage when you search.Prefer Newsweek on Googleto see more of our trusted coverage when you search.
Ed Zitron has done the math on the AI industry. He does not like what he found.
Zitron, a tech analyst and writer behind the popular newsletter Where's Your Ed At and podcast Better Offline, returned to Newsweek's The 1600 podcast for a second appearance with host Carlo Versano, after his first visit became the show's most-watched episode. He came back with the same message and sharper receipts.
"The data centers do not make sense," Zitron said. "No one can actually square the math."
Read More on Politics
Watch Here: The 1600 Podcast Full Episode
The Data Center Time Bomb
Building a modern AI data center costs roughly $44 million per megawatt, about $30 million for servers and GPUs and $14 million or more in construction. A single facility can require hundreds of megawatts, meaning total costs run into the billions before a single paying customer arrives.
Those customers are a problem of their own. "All of your customers are unprofitable AI startups," Zitron said. "So you, the debt holder who has lots of debt that you need to pay off, your revenue source is another company that cannot pay its bills."
The GPUs filling these facilities depreciate over six years at best and are technologically obsolete within one. Data centers take a year or two to build, meaning the hardware inside is often already outdated by the time construction is finished.
Zitron outlined three scenarios he believes will trigger a cascade of failures: a facility that never breaks ground because financing collapses, one that stalls mid-construction, and one that opens only to find no customers left. "All it's going to take for this to fall apart is one or two of those to collapse," he said. "One or two of them."
He pointed to OpenAI as exhibit A. The company has reportedly committed to spending $600 billion on compute by 2030 and owes Oracle $300 billion, with payments due to begin next year. "Where's the money coming from, Sam?" Zitron asked. "Who's got the money? Where's the money going?"
He also noted that $178.5 billion in data center credit deals were done in 2025 alone, with most of that debt rated junk grade. "Everything is relying on a system that doesn't exist anymore," he said.
The Claude Code Illusion
Zitron saved some of his sharpest criticism for Claude Code, Anthropic's buzzy new AI coding assistant, which has generated breathless coverage from outlets including The Atlantic, The Verge and NBC.
The product generates roughly $28 million a month in revenue. For something heralded as the most important tool in technology, Zitron argues that number is a confession. "That is a lot of money to you and me," he said. "But for the product that's allegedly the most important thing that everyone uses that's changing everything, that's it."
Worse, he says the number is built on a fiction. According to research he cited, Claude Code subscribers are allowed to consume between $8 and $13.50 in actual computing costs for every dollar they pay. A user on the $200 per month plan can effectively burn through over $2,700 in real compute costs.
...
"Everyone using Claude Code right now is in an illusory environment," Zitron said. "They are using a product that will not last long term, that they cannot afford to serve."
He compared the dynamic to driving a Hummer with someone else paying for the gas. You would never think about fuel efficiency. You would never worry about the cost of a wrong turn. "You're going to forgive it for a lot of its problems," he said.
When the subsidy ends, users will not gradually pay a little more the way Uber riders adjusted to the end of promotional pricing. They will face bills potentially ten times higher than what they pay today. "You think any of these people are going to go from paying $200 a month to $2,000 a month and be like, 'I love this. This is good'?" he asked.
The Marketing of Moral AI
The conversation turned to Anthropic's recent standoff with the Pentagon, after Defense Secretary Pete Hegseth gave the company a deadline to open Claude for unrestricted military use. Anthropic has said it cannot abide by the Pentagon's ultimatum that its model loosen safety restrictions, lest it be used for potential autonomous killings or mass surveillance.
Zitron was blunt. "Anthropic is doing this for marketing," he said. He argued the concerns about killer robots and a military surveillance dragnet do not reflect what the technology can actually do, and that the company's broader safety messaging is designed to avoid the more tangible harms he believes the industry is already causing.
"AI safety never talks about the economic problems, the bubble, the stealing, the ecological damage," he said. "It always talks about this theoretical future AI."
He pointed to what he sees as a fundamental contradiction at the heart of Anthropic's public image. The company raised $30 billion from 37 different investors while positioning itself as the conscience of the industry. "He said he wouldn't take money from the Saudis," Zitron said of CEO Dario Amodei.
"What did he do? Thirty billion dollar round, 37 different companies."
Nobody Asked for This
Zitron's broadest indictment was simple. The AI industry has no real constituency among ordinary people.
"People don't want this," he said. "Regular people don't want their jobs displaced." He argued the public is approaching the AI boom with a skepticism born of experience. They were promised social media would connect the world. The same voices are now promising a revolution in productivity and human capability.
...
"I think people would want this if it did what they said it did," Zitron said. "Which it doesn't."
He predicted the data center bubble could begin to pop this year, with collapses spreading over months or years, arriving just in time to become a defining issue in the 2026 midterms. The companies responsible, he said, will have little political protection and even less public sympathy.
"If they think people hated bankers," he said, "people hate data centers."
Watch the full conversation between Ed Zitron with Carlo Versano in the video player above or on YouTube.
Subscribe to The 1600 newsletter for more interviews and to receive Carlo's popular daily column. Tap here to have it delivered straight to your inbox.
View 0 Comments
Submit Correction
View Editorial & AI Guidelines