Kelsey Piper at The Argument has a blunt read on Ed Zitron, the tech columnist who's become one of AI's loudest skeptics: his case has fallen apart and he knows it. Zitron spent 2024 and 2025 arguing AI was a bubble because companies weren't adopting it, costs weren't falling, and the technology had peaked. Then companies adopted it. Costs plummeted. The technology got much better. So Zitron shifted to calling OpenAI and Anthropic outright frauds, comparing them to FTX and Enron. Piper sees this shift as telling. "This is basically an admission that he can't make the case in terms of the economics anymore," she writes.
The numbers back her up. GPT-4-level intelligence now costs roughly 1/1000th of what it did when GPT-4 launched. API costs fell from around $60 per million tokens in early 2024 to under $6 for equivalent capability by 2026. Mixture of Experts architectures, better quantization, and specialized hardware made that possible. Competition between OpenAI, Anthropic, Google, and Meta's open-source LLaMA models kept the pressure on. About 30% of Fortune 500 companies now have enterprise AI deals. Though adoption has surged, this rapid integration is creating massive backlogs of unreviewed code, while more than half of Americans use chatbots weekly.
Zitron's fraud allegation deserves scrutiny. He claims OpenAI might be counting free tokens as revenue, which would be securities fraud if true. But he offers no evidence, just suspicion. That's a serious accusation to lob without proof. Piper points out you can still make a credible bear case on AI valuations as investors have largely ghosted OpenAI, though Zitron occasionally gestures at those arguments but never makes them in detail. He's too busy calling people liars.