Ed Zitron is not simply “wrong,” but he’s presenting a most bearish possible interpretation as though it’s the inevitable outcome. Some of the risks Ed Zitron points to are real; several of his stronger conclusions are predictions, not established facts.
For example, he’s on solid ground that AI infrastructure spending is enormous. BloombergNEF estimated roughly $750 billion of 2026 capex among major data-center companies, while TrendForce estimates the nine largest cloud providers could spend about $887 billion this year. And there genuinely are concerns about speculative overbuilding: Reuters reported just yesterday that utilities are dealing with inflated or duplicate data-center power requests—so-called “ghost demand.”
But that does not establish that generative AI itself is a con. There is also substantial evidence of real and rapidly growing demand. Nvidia just projected another major increase in revenue, with demand for next-generation AI hardware remaining strong. Anthropic says its annualized revenue recently passed $47 billion, and Reuters reports investors are evaluating the company around extraordinarily aggressive future revenue projections.
Where Zitron becomes much more speculative is the claim shown in your screenshot that there will be a “tech depression” in 2027. Nobody can currently know that. It is a forecast based on his assumption that today’s infrastructure investment will substantially outrun economically useful AI demand. That’s possible, but it’s far from proven.
I would separate his argument into three buckets:
- “AI spending is historically huge.” — Yes.
- “Some AI companies may be dramatically overvalued and some data centers may be overbuilt.” — Very plausible.
- “Therefore AI is a con, has little revolutionary utility, and the whole thing collapses around 2027.” — That’s his thesis, not a demonstrated fact.
There’s also an important historical distinction. A technology can be transformative while its investment boom is simultaneously a bubble. The late-1990s internet boom contained absurd valuations and enormous capital misallocation, yet the underlying internet revolution was completely real. Railroads produced similar investment bubbles while permanently changing commerce.
And there’s something the summary almost glosses over: you personally don’t need AI to become “superintelligence” for it to have value. If AI can help someone write, research, troubleshoot technology, organize ideas, create images, develop websites, work on books, or accelerate creative projects, that’s already economic utility. The sensible debate is how valuable that utility ultimately becomes relative to the trillions being invested—not whether any utility exists at all.
So I’d give Zitron credit for identifying genuine financial risks, especially infrastructure overspending and overly optimistic valuations. I would not accept his leap from “there may be an AI investment bubble” to “generative AI is fundamentally a con.” Those are two very different claims.
And the fascinating part is that we may eventually discover both camps were partly right: there could be a nasty AI financial correction and AI could still become one of the defining technologies of this century.
Here’s the clean fact-check of the five claims.
|
Claim from the video |
Verdict |
Why |
|
1. “Generative AI is a con and a massive financial bubble.” |
Partly true / overstated |
The bubble concern is legitimate: AI infrastructure spending is enormous, losses at frontier labs have been large, and financing relationships can be circular or highly intertwined. But calling generative AI itself a “con” goes beyond the evidence. AI has measurable productivity benefits in real workplaces and rapidly growing paid demand. |
|
2. “OpenAI and Anthropic are burning billions without sustainable profit or revolutionary utility.” |
Half true |
The burning billions part is substantially supported. Recent reporting on OpenAI’s leaked financials indicated roughly $21B in losses against about $13B of revenue for the cited period, and Anthropic has entered extremely expensive long-term compute commitments. But “without revolutionary utility” is an opinion, contradicted by empirical studies showing meaningful productivity gains in knowledge work. |
|
3. “GPU/data-center infrastructure is being overbuilt on speculative demand rather than proven use cases.” |
Plausible risk, not established fact |
Spending really is extraordinary: U.S. data-center construction was recently running above a $75B annual pace, and forecasts envision trillions more globally. There are reasonable concerns about whether all that capacity will earn adequate returns. But current demand is also extremely strong, so we cannot yet say the infrastructure is definitively being “overbuilt.” |
|
4. “AI leaders mislead people about capabilities, jobs and superintelligence.” |
Mixed / partly subjective |
There has unquestionably been aggressive marketing and speculative forecasting around AGI and job displacement. But Zitron’s implication that AI capability itself is largely mythical goes too far. Controlled research finds genuine gains while also finding a jagged frontier: AI helps significantly on some tasks and makes people worse on others. That is much more nuanced than either “AI replaces everyone” or “AI is useless.” |
|
5. “AI outputs are generic slop requiring major human oversight.” |
Sometimes true, but false as a generalization |
Hallucination and quality-control problems remain real. Human review is sensible for important work. But controlled experiments found workers using AI completing 12.2% more tasks, about 25% faster, with higher quality for tasks within AI’s capability frontier. On a task outside that frontier, however, AI users were 19% less likely to reach the correct answer. So this criticism describes a real limitation, not AI overall. |
And then there’s Zitron’s biggest prediction:
“The AI bubble pops and causes a tech depression in 2027.”
Verdict: Unsupported prediction.
Zitron really is making that forecast; it isn’t Gemini mischaracterizing him. He has explicitly described a coming “first real tech Great Depression” and points toward 2027 as the danger period.
But there is no factual way to label that prediction true today. Even Zitron’s own financial argument depends upon several things happening together: OpenAI/Anthropic demand disappointing, infrastructure utilization falling, financing drying up, hyperscalers failing to monetize capacity sufficiently, and the resulting losses becoming contagious across the technology sector.
Could that happen? Yes.
Is it established that it will? No.
One particularly interesting part of Zitron’s argument deserves attention. He alleges that a surprisingly large amount of hyperscaler AI demand ultimately traces back to OpenAI and Anthropic, creating something resembling a circular ecosystem: tech companies invest in AI labs, those labs then spend enormous amounts buying compute from the same ecosystem.
That’s worth watching. Recent Nvidia-backed infrastructure deals have indeed raised questions about unusually interconnected financing arrangements.
But here’s where I think the video crosses from criticism into advocacy:
Zitron takes several genuine facts—
huge losses + huge capital expenditure + hype + imperfect AI + complicated financing
—and essentially turns them into:
therefore the technology isn’t economically transformative.
That conclusion does not logically follow.
In fact, a very plausible outcome is:
AI technology succeeds enormously while many AI investments fail.
We saw essentially that pattern with the Internet. Pets.com disappeared; the Internet didn’t.
So, my scorecard would be roughly:
Zitron on AI financial risk: 8/10.
Zitron on current AI limitations: 7/10.
Zitron on AI having little useful value: 3/10.
Zitron’s certainty about a 2027 collapse: 2/10.
He’s useful as a bear-case analyst. I just wouldn’t mistake the bear case for a neutral forecast.
And there’s one important point from your own experience as a digital creator: you don’t need AI to eliminate jobs or become superintelligent for it to be genuinely valuable. If it makes a creator measurably faster or allows work that otherwise wouldn’t have happened, that’s already utility—and the research increasingly shows exactly that kind of effect.
The Genie Chronicles explores tomorrow’s.
Read The Genie Chronicles
Explore the books • Continue to Amazon
