Summary
Sabine Hossenfelder argues that the biggest threat to AI progress may not be regulation or public backlash, but the fact that too few people are paying for it. She suggests many skeptics tried a chatbot once, years ago, and never saw how much the models have improved, while viral fail clips shape the public impression.
Drawing on the economic idea of an “experience good,” she explains why free-tier users may never encounter the capable models, and why one economist warns of a demand bottleneck that could stall further investment. The numbers are stark: ChatGPT has 900 million weekly users but about 50 million paying subscribers, roughly one in 18, and only 2 to 3% of American households pay for AI.
The video also reviews the finances of the leading labs. OpenAI’s quarterly revenue rose from $5.7 billion to $6.7 billion while its operating loss grew from $9.3 billion to $12.3 billion, and the company projects spending $278 billion more than it takes in through 2030. Anthropic, heading toward a record IPO, reported a large first-half loss that is mostly a paper effect of investor share revaluations. Ramp data shows flat business token spending. Hossenfelder concludes that a lack of compelling everyday use cases, with agents still cumbersome and unreliable, is holding back adoption.
📺 Source: Sabine Hossenfelder · Published November 11, 2026
🏷️ Format: Opinion Editorial
