Anthropic is preparing to file for a U.S. IPO after confidentially submitting a draft Form S-1 to the Securities and Exchange Commission earlier this year, while OpenAI remains at an earlier stage of its offering process. Anthropic was valued at nearly $1 trillion, or about $965 billion, in a $65 billion Series H round, and market participants say it could raise an amount comparable to or greater than SpaceX’s $86.2 billion record offering. The anticipated deal is concentrating investor attention on AI infrastructure, aerospace and defense and could pressure other issuers during a crowded September-October window shaped by weaker IPO returns, Federal Reserve policy, inflation and the U.S. midterm elections. Prediction markets assign a 94.5% probability to an Anthropic listing by Dec. 31, 2026, but only 6.5% to a listing by Sept. 30, 2026, down from 8%. At the same time, Anthropic is easing safeguards in updated Fable and Mythos models and introducing enterprise monitoring that does not store customer data, while OpenAI prepares a broad Astra release but limits its most powerful cybersecurity capabilities to trusted testers after classifying the model as reaching a critical cybersecurity threshold. The contrasting safety and access messages reflect both companies’ effort to demonstrate high-growth businesses to Wall Street while reassuring global regulators and the public that their models will not create unacceptable risks.