Summary: Bloomberg reported on May 21, 2026 that Cursor reached a $3 billion annualized sales rate while its acquisition by SpaceX was being assembled. The report also counted more than 3,000 customers paying at least $100,000 annually. The article says AI coding procurement is becoming enterprise infrastructure purchasing, with vendor survival, model supply, and acquisition risk now part of total cost.
News from Bloomberg dated May 21, 2026: the AI coding tool Cursor has reached a $3 billion annualized sales rate, and the figure lands while the SpaceX acquisition is still being put together. The same report counts more than 3,000 customers paying Cursor at least $100,000 a year. Roll the calendar back three months and the same outlet was logging its recurring revenue “doubling in three months to $2 billion.”
Read together, these figures form the sharpest thermometer available for the AI coding market. Quarter-over-quarter growth of fifty percent, enterprise contracts in the six figures, and a transaction about to redraw the cap table all point in one direction: the category has left pure tool-versus-tool competition behind and entered a phase defined by consolidation at scale.
$2B to $3B, One Quarter Apart
The timeline lines up cleanly. Annualized revenue stood at $2 billion in February 2026; by May 2026 it read $3 billion. Bloomberg’s March piece had already revealed an internal forecast at the company putting annualized revenue past $6 billion by the end of 2026. Measured from the May figure, reaching that mark means doubling once more across the seven months left on the clock — a demanding ask, though one in keeping with the tempo of the previous two quarters.
Just as telling is what the growth looks like. One quarter covered the entire move from $2 billion to $3 billion: the percentage rate eased from “doubling” down to fifty percent as the larger base took effect, yet the absolute gain still lands in the billion-dollars-per-quarter range. Software history offers very few tools that held that kind of pace at this size.
One qualifier belongs here: an annualized sales rate is a run-rate construct — today’s contracted revenue stretched across twelve months — and it can correct downward exactly as fast as it climbed. The signal lives in the cadence, not in any single print: three straight quarters of acceleration, every landmark filed by the same outlet, and a business that was already at scale when the year began.
3,000 Accounts, Each Worth at Least $100K
The number that rewards close reading is the customer mix: over 3,000 customers each pay $100,000 or more annually. Count every one of them at the bare floor and the cohort still yields at least $300 million a year — around a tenth of the total, with the true share certain to run higher, since plenty of those contracts sit far past the $100,000 line.
That structure carries two messages. First, the growth engine has changed place: individual developer subscriptions have given way to enterprise contracts — larger checks and stickier renewals, paired with a completely different set of demands around governance, security, and procurement. Second, the money companies set aside for AI coding is turning from experiment budgets into recurring line items — $100K is hardening into the basic unit of an enterprise deal in this category, and the product’s identity shifts with it: no longer a per-developer subscription that happened to grow, but line-of-business infrastructure bought the way data warehouses and observability stacks are.
How the SpaceX Deal Redraws the Game
By folding the $3 billion milestone and the SpaceX transaction into a single headline, Bloomberg made its point: the acquisition is already baked into how the market prices Cursor. Three consequences follow for the developer-tools industry. First, AI coding has grown into a business large enough that the Musk camp chose to acquire rather than assemble. Second, once control changes hands, Cursor’s model supply chain steps into new contractual territory — it leans on several model vendors, and vendors have never been comfortable about customers bought by their rivals. Third, rivals such as Claude Code, GitHub Copilot, and Codex will no longer be facing a standalone startup but a group with compute and capital behind it. On the buyer side, running several tools at once is already the norm: Virgin Atlantic has publicly described shipping faster with Codex — large organizations pick portfolios, not a single wager.
The Takeaways for Developers and Product Teams
Three working conclusions. One: buying AI coding tools is turning into a contract-negotiation exercise — security review, data boundaries, and vendor terms become the default checklist, and the teams behind these tools should expect procurement to comb through them. Two: supply-chain risk now belongs on the shortlist of selection criteria — an acquisition, a model cutoff, or a terms change can sever a team’s workflow without warning, and holding at least two workable tools is the minimum hedge. Three: the leader widens the gap by a billion dollars of fresh revenue every quarter, which will push median tools and in-house builds out of realistic range within a year; when choosing, price vendor survival and acquisition risk into total cost of ownership.