The week started with Anthropic shipping its most capable model to the public and ended with SpaceX pricing what may be the largest IPO in history. In between, OpenAI filed its own confidential S-1, and Apple rethought what a developer platform looks like in the age of agents. If the AI industry needed a week to announce itself to Wall Street, this was it.
The model
On Monday, Anthropic released Claude Fable 5, making its Mythos-class capability available to general subscribers for the first time. Fable 5 is technically the same model as the restricted Mythos 5, with safety classifiers added that block responses in domains like cybersecurity and biology—triggering in under 5% of sessions on average, though early testers have found edge cases, including classifiers flagging a cancer-genomics coding query as a biosecurity risk. Simon Willison called it impressive on tasks requiring combined reasoning and breadth, and noted the guardrails will need tuning. Pricing is $10 per million input tokens and $50 per million output, with free access for Pro and higher-tier subscribers through June 22 before moving to usage credits.
The money
Anthropic filed a confidential S-1 with the SEC on June 1, days after closing a $65 billion Series H at a $965 billion post-money valuation. OpenAI followed on June 8—Goldman Sachs and Morgan Stanley as underwriters, $852 billion valuation—and similarly declined to commit to timing. Neither has published financials. Both are the first pure-play frontier AI labs to initiate a public offering, and both insist there’s no rush.
SpaceX was less coy. Tonight, shares priced at $135, targeting a $75 billion raise at a $1.75 trillion valuation—the largest IPO in history by any measure, eclipsing Saudi Aramco’s 2019 listing. Trading begins tomorrow on Nasdaq under SPCX. Three companies approaching a combined $3.5 trillion in valuation are going public in the same month. Whatever the next chapter looks like, it has a ticker symbol.
The developer stack
At WWDC 2026, Apple opened its Foundation Models API to developers with fewer than 2 million App Store downloads at no charge, running inference on Private Cloud Compute. A new language model protocol lets Swift apps call Claude, Gemini, or any compliant model through a single API. The Foundation Models framework goes open source this summer. Xcode 27 adds agentic coding support: agents that interact with the iOS Simulator, localize apps, run tests, and pull crashes from Organizer. GitHub, meanwhile, moved Copilot to usage-based billing for all plans on June 1 and shipped a standalone Copilot app for agent workflows.
The billing shift matters. Uber burned its 2026 AI budget in four months and capped employees at $1,500 per month per tool. Supabase raised $500 million at a $10.5 billion valuation this week, crediting Claude Code as the top source of new databases and a 600% year-over-year surge in database creation. The spending is real, the growth is real, and the bills are landing on the same desk. That’s what a maturing industry looks like.