Ed Zitron: The AI Bubble is Bleeding Cash, Here Are The Receipts

**Jack Farley** (0:00)
Got a very special conversation. I am speaking to one of the most prolific skeptics about AI.
I'm joined today by Ed Zitron, author of Where's Your Ed at Newsletter and the Better Offline Podcast. Ed, welcome to Monetary Matters.

**Ed Zitron** (0:14)
Thanks for having me.

**Jack Farley** (0:15)
Ed, how about you start off and lay out your view of artificial intelligence and what the return on investment is for the enormous sums that are currently being spent.

**Ed Zitron** (0:25)
So trillion plus dollars in, the fact we're still debating the ROI kind of says everything. I think if this was a real industry with the kind of TAM that they've been selling us on for the past four years, we wouldn't have that debate. There wouldn't be one. The fact that we're having it, the fact that you have people to this day in the year of our Lord 2026 saying, well, AI is real.
I've never heard something real where someone has had to insist upon that. Economically speaking, the vast majority of AI companies barely make more than $100 million a year in revenue. Anthropocene OpenAI account for I think 89 percent of all the top. On top of that, they're all horribly unprofitable. They lose billions of dollars. The only way that they can reconcile those is by doing some, what I would call wacky accounting.
In my reporting from earlier in the week, OpenAI spent $34 billion to make $13.07 billion.
Their net loss is just a little under $21 billion. They have, of course, with the FT given comment that suggests that they only lost $8 billion. I think that's laughable. I think that that was a booster key jingling. I think that that was put in there because they have to find a way to reconcile with the fact that they told CNBC they lost $8 billion at the beginning of the year. When it's very clear they didn't, like put aside whatever fanciful quotes there are. That they lost $21 billion.

**Jack Farley** (1:57)
How much, by the way, has Uber or Amazon lost when they were losing tons of money and before they became the giant behemoths? And how does that compare to the, as you say, $21 billion that OpenAI lost last year?

**Ed Zitron** (2:10)
Okay, so Uber, I believe, burned like $32 billion.
They're now kind of like messy, gap profitable. A lot of that was sales and marketing, and a lot of that was subsidizing rides, but the level of subsidies they did were just completely different. The Amazon Web Services example is the really egregious one. Between 2003 and 2017, I think it is, normalized for inflation, Amazon Web Services maybe 53, 55 billion dollars total. And that's all of Amazon's capex, not just AWS. Just to give you some context, if OpenAI closes all the money they've been promised this year, they will have raised $122 billion in the last six months. If Anthropic closes all the money they've been promised, they will have raised $95 billion in the last six months. OpenAI raised $40 billion last year, Anthropic raised $16.5 billion.
The answer is incomparable. Amazon's CapEx, for its retail operation on top of AWS, was less than half of what OpenAI has raised in one round this year. There's no real comparison. People making this comparison are looking for a means of rationalizing the irrational and of finding a way to not reconcile with the fact that the two largest companies, pretty much the only demand for AI compute between Anthropic and OpenAI.
Those two companies don't make sense. They don't make economic sense. The only way to make them make economic sense is to just ignore your lying eyes. It sucks because I think that the tech industry has been poisoned with a very specific kind of ideology and I think the irony is, so many of them are atheists, stern, hard rationalists, but they have a quasi-religious attachment to artificial intelligence. Every minor sign is proof that the great prophecy is true, and every little point that could possibly attack a skeptic is considered unilateral proof that they're wrong. And it sucks because they're trying to dump these companies on retail investors who are going to be the victims of this theological hype cycle.

**Jack Farley** (4:22)
I think you're exactly right. I mean, it was a senior executive co-founder of Google who said that they would rather go bankrupt than lose an AI. They're committed.

**Ed Zitron** (4:30)
Yeah. And all of that, I think, comes from this theory called the rock-com bubble, which is tech is out of ideas.
When I say that, I don't mean they literally have none. I mean, they have no hyper-growth ideas. They don't have a new AWS. They don't have a new iPhone or a new smartphone. They don't have an ex-cloud computing. And AI was meant to be the panacea. It was meant to be AI models, as in via the API were meant to be the thing that created the next generation of startups, that created the next generation of enterprise bolt-ons so that the software industry could grow. It was meant to be the future of consumer software. And it was even meant to be the next AWS in the form of AI GPUs, which is why they've sunk so much money into it. I think that the hyperscalers, Meta, put aside, I don't think Meta really has a strategy. I think it's just like spend money until Mark gets bored. But Microsoft, Google, Amazon, I think they saw this as an opportunity to create the next generation because they don't have anything else. There is no other business line that's shown anything close to the possibility of having the growth. Just talking pure growth. The worst thing is that while they've seen some growth from AI, it's mostly because Anthropic and OpenAI spend an alarming amount of money on compute, which means that they're basically just feeding their money back into themselves and then feeding their own cash flows into NVIDIA or Broadcom or one of the Taiwanese ODM, so Hon Hai, Foxconn, the companies that build the servers in Taiwan. So it's rough. It's very rough because for me to be wrong, there needs to be hundreds of billions of dollars of AI compute demand. Put aside OpenAI and Anthropic, hundreds of billions of dollars because there are over 100 gigawatts of data centers planned, 12.5 to 15 million per megawatt.