Image by DL314 Lin on Unsplash
Last week, Ars Technica hosted a strikingly candid live chat with Ed Zitron, tech critic and host of the podcast Better Offline. The big question: Is generative AI a trillion-dollar transformation or just a $50 billion illusion?
Let’s talk about it.
“Everyone’s Acting Like It’s Something It Isn’t”
Right out the gate, Zitron didn’t hold back. When asked why he’s so critical of AI, he said it’s being sold as a panacea—one that will fuel the next wave of software, hardware, and compute growth.
But, Zitron said, “It’s a $50 billion revenue industry masquerading as a trillion-dollar one.”
He pointed to staggering financial losses, especially at OpenAI. The company reportedly burned through $9.7 billion in the first half of 2025 alone. That’s not exactly the hallmark of a hyper-profitable sector. Meanwhile, companies installing the latest AI chips, like Oracle’s investment in Nvidia’s Blackwell GPUs, are reportedly losing money due to huge energy costs.
Can the AI Use Case Justify the Hype?
The conversation wasn’t all doom and gloom. There was an honest exchange about AI’s small but meaningful wins.
Image by Ryan Quintal on Unsplash
For example, the host shared how chat-based AI tools help people overcome cognitive hurdles, like memory loss or focus issues—something especially tangible for those dealing with brain fog after long COVID. These tools act like “framework translators,” mapping fuzzy ideas into clearer paths.
Zitron didn’t deny this — he just didn’t see it as a trillion-dollar breakthrough.
“That’s cool,” he said. “I’m glad that helps you. But it’s not a trillion-dollar use case.”
He even admitted he tries these tools himself. Still, he came away frustrated. “AI is all issues. It’s all filler, no killer.”
The Money Loop Behind the Curtain
Zitron believes there’s more at play than overhyped promises — he sees a circular flow of investment that’s propping everything up.
Take this loop he described:
- OpenAI inks a $300 billion deal with Oracle.
- Nvidia is the main supplier of GPUs.
- CoreWeave, funded by Nvidia, becomes its own biggest customer.
- Then CoreWeave uses those GPUs as collateral to raise debt and buy even more Nvidia chips.
“It’s castles on sand,” Zitron said, questioning how sustainable all this really is — especially as costs rise while utility lags behind.
Bigger, Hotter, Hungrier Data Centers
Here’s where things get truly wild: OpenAI has plans to build Stargate, a data center complex in Texas requiring 10 gigawatts of power. That’s the output of 10 nuclear reactors.
By comparison? The town of Abilene, Texas — Stargate’s future home — has only 350 megawatts of generating capacity and a 200-megawatt substation.
Image by Leif Christoph Gottwald on Unsplash
“It’s not like Red Alert 2,” Zitron joked. “You don’t just build a power station and it happens.” He argues this kind of infrastructure fantasy is one of many promises made without enough real-world grounding.
The Timer Is Ticking
Zitron predicts that the AI market bubble could burst within the next 18 months. But he’s not expecting a single dramatic collapse like Bear Stearns.
Instead, he sees a cascade:
- One startup runs out of funding.
- Others panic.
- Investors freeze.
- Fundraising becomes impossible.
And at the heart of it all? Nvidia.
Zitron says Nvidia’s stock makes up about 7–8% of the S&P 500’s value. If its hypergrowth slows, the entire AI market could unravel.
“Nvidia’s growth is why the bubble is inflated,” he said. “If their growth goes down, the bubble will burst.”
Pushing Back — and Finding Middle Ground
To his credit, the host tried to look at the long-term arc. Computing costs have historically come down. What’s power-hungry today might be cheap tomorrow.
Think about it: in the 1950s, the Air Force’s SAGE computer system used two megawatts for a few thousand operations per second. Today, your phone is millions of times more efficient.
Zitron wasn’t convinced. He argued that, right now, costs are actually going up. Even newer AI systems like Cerebras or Grok may offer speed, but not savings.
And while AI assistants might get smarter, they still struggle with consistency and simple commands — things you’d expect before putting them fully into operating systems.
The Sam Altman Factor
It’s impossible to talk about the AI boom without mentioning OpenAI CEO Sam Altman. At the end of the talk, Zitron was asked if he could say anything nice about him.
His response? Direct as ever:
“Sam Altman is going to be the reason the markets take a crap. Sam Altman has lied to everyone.”
But he did give credit where it was due: “He’s really good at making people say ‘Yes.’”
So, Is AI a Bubble or Not?
That’s the big question.
Zitron’s view is clear. He thinks tech is chasing a growth high that it can’t sustain, with generative AI as its latest fix. His worry? When reality hits, it won’t just be AI firms that suffer — Silicon Valley and the broader market might take a hit too.
That doesn’t mean AI lacks real utility. Plenty of people find it useful, even life-changing. The tools are improving and evolving. But usefulness doesn’t always mean profitability — and definitely not sustainability at trillion-dollar levels.
The next 18 months might be reckoning time.
Stay tuned. The bubble may not pop loudly. But it could deflate quietly, bit by bit.
Image by Stephen Harlan on Unsplash
Keywords: AI bubble, Ed Zitron, OpenAI, Nvidia, AI infrastructure, AI power costs, generative AI market, Sam Altman, CoreWeave, data centers, AI economics, tech investment, AI criticism, Ars Technica interview
