SpaceX Signs the Cursor Deal: $60B Stock Acquisition Closes the Option, Grok V9-Medium Ships Alongside
SpaceX signed a definitive merger agreement on June 16 to acquire Anysphere — the San Francisco company behind Cursor — for $60 billion in SpaceX stock. The deal exercises an option first announced in April 2026. Close is expected in Q3 2026, pending regulatory review.
The option had been structured as a binary: $60 billion for full ownership, or a $10 billion breakup fee for a data and compute partnership. SpaceX chose ownership.
The Context for the Decision
SpaceX went public on June 12 at $135 per share, raising $75 billion in the largest US tech IPO in years. By Tuesday morning its stock was trading above $200 — adding roughly one trillion dollars in market capitalisation in four days. At that price, $60 billion in SpaceX stock became substantially less dilutive than it would have been pre-IPO, which almost certainly accelerated the timeline to close the option.
Cursor had crossed $2 billion in annual recurring revenue by early 2026, the fastest B2B SaaS company to reach that number. More than half the Fortune 500 use it. Over one million developers have it as their primary development environment. The product runs on models from Anthropic, OpenAI, and Cursor’s own Composer line simultaneously, giving users a choice at the point of call.
That model-agnostic design is now the open question.
The Developer Risk
SpaceX’s AI division, built around the assets of the former xAI, lost $6.35 billion in 2025. Every Cursor API call that routes to Anthropic’s Claude or OpenAI’s GPT is revenue leaving SpaceX’s ecosystem. The financial incentive to shift that traffic toward Grok — once Grok is competitive enough to hold users — is direct and large.
No changes to model access have been announced. Cursor continues to support Claude, GPT, and Composer models through at least the Q3 2026 close. No public commitment to maintaining model agnosticism after the merger has been made.
Developers with enterprise contracts built around Claude access inside Cursor have a finite window to track this. If SpaceX announces model defaults after close, migration costs will be higher than if teams plan now.
Grok V9-Medium Ships at the Same Time
The timing of the definitive agreement is not coincidental. Elon Musk announced on May 25 that Grok V9-Medium had completed training at 1.5 trillion parameters — three times the parameter count of V8-small, which has handled all Grok production traffic at approximately 500 billion parameters. Supervised fine-tuning and reinforcement learning completed by early June. As of this week, V9-Medium is rolling out to Tesla’s connected-car fleet, to X, and to developers via the xAI API.
V9-Medium is a coding-specialised model, not a general frontier release. Training incorporated a large corpus of Cursor developer workflow data — actual session traces of how developers interact with an AI coding environment. xAI has described this as a deliberate reconstruction of the training data problems that left V8-small “not built right.”
Current SWE-bench Verified leaders sit at 88.7% (GPT-5.5) and upward of 95% for frontier Anthropic models. The Grok 4 series scored 70-75% on the same benchmark. V9-Medium’s published scores are not yet available. SpaceX has telegraphed that the model will be benchmarked against Claude and GPT-5.5 on SWE-bench Verified, which will provide the first independent read on whether the Cursor training data closed the gap.
What Grok 5 Is Not
V9-Medium is separate from Grok 5, the six-trillion-parameter model still in training on Colossus 2. Prediction markets gave Grok 5 approximately a 12% chance of shipping by June 30. V9-Medium is a production deployment of what is already done; Grok 5 remains a future event.
The Structural Outcome
SpaceX enters the enterprise AI coding market owning the infrastructure (Colossus 1 and 2), the model (Grok, now V9-Medium), the developer interface (Cursor), and the distribution (X, Tesla fleet, developer API). That is a full stack from compute to end-user.
It is the same vertical integration logic that governs every successful platform shift in enterprise software. Whether the stack delivers a model competitive enough to retain Cursor’s user base without the Claude and GPT fallbacks is the question that Q3 2026’s merger close will force.