Cursor, the AI coding editor that defined last year's wave, lost half its market after Anthropic launched Claude Code.
SpaceX agreed to acquire Cursor, the AI coding editor, for roughly $60 billion in stock, and disclosed the deal in an SEC filing on June 16, 2026, just days after its own blockbuster IPO. One number makes the $60 billion price hard to justify: Cursor's share of the AI coding tools market fell from about 41% in June 2025 to roughly 26% by May 2026, per a market estimate attributed to an unnamed third-party analytics source cited by Leiphone; the underlying dataset and its methodology are not publicly identified in the source. Yet the same period saw Cursor's annualized revenue climb to about $4 billion, up from roughly $2 billion in March 2026 and about $1 billion across late 2025. The deal only makes sense once you understand what SpaceX is actually buying, and why the share collapse is the condition that made it possible rather than evidence against it.
The 41-to-26 collapse is a named story. Anthropic launched Claude Code, its own native AI coding assistant, in February 2025, and it became the default tool for a generation of developers who had previously defaulted to Cursor. That single launch, plus a wave of rivals and OpenAI's own coding surface, cut Cursor's dominance in half within a year. It also exposed the structural vulnerability Cursor had built into its product: every completion the editor serves is a call to either Anthropic's Claude or OpenAI's GPT. Cursor does not own the model. It owns the interface, the workflow, and the developer relationships.
That is the part that matters for SpaceX. Musk's xAI unit, which runs the Grok models, has no market-validated AI-coding product. SpaceX's IPO materials position the merged entity around what the company calls Macrohard, an autonomous agentic-AI platform, with Cursor described as one of its real-world commercial anchors. To make Macrohard credible to public-market investors, SpaceX needs a real revenue line and a real proof point that agentic AI is shipping in production, not just in demos. Cursor delivers both: about $4 billion in annualized revenue, growing fast, plus a daily stream of high-quality developer-interaction signal that is exactly the kind of post-training data the industry is currently fighting over.
The interaction data is the quieter half of the story. Every Cursor session is a record of what a working developer accepts, rejects, edits, and asks an agent to fix. That signal, the difference between a generated completion a human commits and one they throw away, is the closest thing the AI industry has to a real-world reward model. OpenAI, Anthropic, and Google all need it. None of them can collect it at Cursor's scale without owning Cursor. Buying Cursor gives SpaceX a self-replenishing pipeline of exactly the data its competitors want, and a reason to keep Cursor's interface open even as it integrates the signal into Grok.
The market-share slide is what made the price reachable. Cursor's late-2025 valuation of roughly $30 billion assumed continued dominance. By the spring of 2026, with Anthropic's Claude Code and others eating share, the asking price had compressed to a range where a strategic buyer, one who valued data and integration more than current market share, could justify a premium. Forbes and AP describe the deal as all-stock, Reuters confirms the $60 billion headline consideration, and The New York Times had reported on the talks weeks earlier. Inside Cursor, the story had been written for months: in March 2026, Cursor's product and engineering leads, Andrew Milich and Jason Ginsberg, joined xAI, reporting directly to Musk, a move Leiphone reads as early scouting for exactly this combination.
The risk is real and sits on top of the same dependency that created the opportunity. Anthropic publicly cut API supply to Windsurf during the 2025 OpenAI/Windsurf talks, an event Cursor's own team treats as a precedent for what could happen to them. If Anthropic decides to prioritize Claude Code, Cursor's product experience degrades overnight. SpaceX's answer is twofold. The Macrohard thesis presumes enough in-house compute to substitute or re-route model supply, and Cursor had already begun building its own Composer model family in 2025 to reduce that exact dependency. Whether that works, whether SpaceX can turn Cursor from a 150-person company dependent on two external model labs into an integrated agentic-AI anchor inside Macrohard, is the open question the deal now bets on.
Cursor has roughly 150 people, per reporting cited in the Leiphone feature. The valuation moved from about $2.5 billion in early 2025 to roughly $30 billion by late 2025 to about $50 billion by spring 2026 to the $60 billion deal price in June 2026, according to Sacra's tracking. That trajectory does not look like the curve of a company losing market share by accident. It looks like the curve of a company becoming strategic infrastructure at the exact moment its dominance stopped being defensible on its own.
The thing to watch next is not the closing paperwork. It is whether Anthropic continues to serve Cursor at scale once the deal is in Musk's hands, and whether Cursor's daily interaction data starts showing up inside Grok's next training run. Both questions resolve inside the next twelve months.