BREAKING: OpenAI’s unraveling has begun

Substack · Watch/Read

• Gary Marcus argues that OpenAI’s “unraveling,” which he previously predicted, has begun due to a severe erosion of trust, significant financial headwinds, and a worsening burn rate ahead of its planned IPO.
• Sam Altman’s August 18, 2026 announcement of pausing “frontier RL training” for safety reasons was widely disbelieved, with critics on X (e.g., @Ross__Hendricks) interpreting it as a move to stop “torching cash” to rush an IPO.
• Wall Street Journal reports (Berber Jin) reveal OpenAI’s Q2 2026 losses quadrupled by $3 billion to $12.3 billion, while revenue grew only 18% to $6.7 billion, a “terrible look” that also saw their revenue growth significantly slower than rival Anthropic.
• Marcus warns that if OpenAI collapses, the damage could spread, potentially shattering the “illusion sustaining the entire industry” and leading to widespread failures among hardware companies, “neoclouds,” and other GenAI labs as venture capital dries up.