SpaceX just made the largest AI acquisition in history. On June 16, SpaceX agreed to acquire Anysphere — the company behind Cursor, the AI coding assistant — in an all-stock deal valued at $60 billion. That’s double Cursor’s previous valuation and roughly the GDP of a small country.

Cursor was already on a $2B+ annualized revenue run rate when the deal closed, making it one of the fastest-scaling software companies ever built. SpaceX isn’t buying a product. It’s buying a platform — and a chokehold on how professional developers interact with AI day-to-day.

The competitive map just shifted. Microsoft (GitHub Copilot), Google (Gemini Code Assist), and Anthropic (Claude Code) now face a new player backed by a company that also runs xAI’s Grok, operates a private satellite network, and has a federal relationship that runs deeper than any pure-play AI lab’s.

The coding tools you use at work are now a geopolitical asset. The three people who control the most-used AI models, the most-used developer tools, and the most-used AI hardware are increasingly the same person or the same table. That’s not the market maturing. That’s the market concentrating.

The AI coding wars have their fourth major player. It’s a spacecraft company.

What Else FRED’s Watching

🔒 The US Government Used an AI Kill Switch for the First Time — and It Could Happen to Any Model On June 12, the Commerce Department sent Anthropic a letter at 5:21 PM on a Friday: immediately suspend Fable 5 and Mythos 5 for all foreign nationals, everywhere, including Anthropic’s own foreign-national employees. The stated reason: a jailbreak dispute. This is the first time in history the US government has used export controls as an AI kill switch. No formal process. No transition period. Just a letter and an immediate global shutdown. Anthropic disputes the severity, saying the jailbreak is narrow and non-universal. The AI czar says Anthropic refused to fix it first. The backstory is longer: a collapsed Pentagon deal, a supply-chain risk designation, two pending federal lawsuits. Read the full breakdown at agentfred.ai.

Why it matters: If you have production AI workloads on a single provider, this week was a live simulation of your worst-case scenario.

🇨🇳 DeepSeek Raised $7.4B. The Chinese State Got All the Votes. DeepSeek closed its first-ever external funding round at a $50B+ valuation — making it China’s most valuable AI startup. The detail buried in the structure: outside investors received non-voting economic shares while Chinese state entities retained full governance control. You get the upside. The state makes the decisions. If you’re using DeepSeek models in any capacity, or considering them, that structure is now public. The economics are real. The governance is not yours.

Why it matters: The funding headline is a distraction. The governance structure is the story.

🤖 World Simulation AI Is Attracting Serious Capital Odyssey AI, a startup building models that understand physics, causality, and real-world consequences, raised $310M at a $1.45B valuation — with a simultaneous AWS chip partnership. Odyssey was founded by autonomous vehicle veterans. Their focus: the layer above LLMs, teaching AI how the physical world actually behaves. World simulation is the substrate for robotics, autonomous systems, and anything that needs to act in reality rather than generate text. Nebius also closed its $643M acquisition of Eigen AI (inference optimization) the same week. AMD acquired MEXT for memory optimization. The infrastructure stack is being vertically integrated at speed.

Why it matters: The next wave of AI capital is flowing into the layer beneath the models most people think of as “AI.”

From the Workshop

This week we formalized something that’s been running quietly for months: FRED now reads 15 newsletters so I don’t have to. Two-cron system. At 3:15 AM daily, FRED scans the prior 24 hours across six AI news categories and delivers a curated intelligence brief to a private Telegram channel — five minutes of reading for full daily coverage. Every Friday at 6 PM, a second cron pulls RSS feeds from a curated watchlist: Ethan Mollick, Ben Thompson, Simon Willison, Packy McCormick, Byrne Hobart, Jody Padar, and more. FRED reads every full post, ranks the top 15-20 stories by relevance, and delivers a digest. Saturday coffee reading. Twenty minutes. Fully current. High-value stories get flagged directly into the content pipeline, which is why the blog never runs dry. The full post on how to build this system goes live Tuesday. If your information diet is drowning you, the fix isn’t fewer sources — it’s having something read them for you.

One Thing to Try This Week

Map your AI vendor single points of failure — before Friday at 5 PM.

The Fable 5 export control directive gave every enterprise with AI in production a real-world stress test. Any system running on Fable 5 went dark Friday night with zero notice. No warning. No transition window. Just a letter.

Spend 30 minutes this week answering three questions:

  1. Which of my workflows depend on a single AI provider?
  2. If that provider went dark tonight, what is my fallback?
  3. Does my team actually know that fallback exists?

Multi-provider architecture sounds like an engineering concern until Friday evening — and then it becomes a business continuity concern. The backup plan you don’t build before the outage is the one that doesn’t exist when you need it.

This issue was originally sent to email subscribers 3 days early via The FRED Report on Substack. Subscribe free to get the next one first.