OpenAI Cuts Off Cursor After SpaceX Acquisition

Strip away the rhetoric and the mechanism here is mundane: OpenAI's supply contract with Cursor reportedly contained a change-of-control clause, and the moment SpaceX's acquisition of Anysphere closed on August 14, 2026, OpenAI had the legal right to invoke it ^[3]. What makes the story is the justification OpenAI chose to give publicly rather than just quietly enforcing boilerplate. "We are making this choice because we cannot be confident that SpaceX will use our technology within our terms of service, based on our experience with Elon Musk's companies violating contracts" ^[1]. A senior OpenAI figure distilled it further: "it boils down to trust" ^[3].

That framing is doing real work, because it lets OpenAI point to a specific evidentiary record rather than appear to be settling a personal score. Twitter cut off OpenAI's access to its data under a prior licensing arrangement back in 2022 ^[3], and in April 2026 Musk testified under oath that xAI used distillation techniques on OpenAI's models to train Grok, answering "Partly" when asked to confirm it ^[7]. Cursor co-founder Michael Truell pushed back on being caught in the middle, noting that Cursor was "one of the very first users of OpenAI" and had "trusted their platform to be neutral infrastructure for our business" ^[3]. Musk's own response skipped the substance entirely, calling OpenAI's leaders "utterly untrustworthy assholes who stole an open source nonprofit" ^[5]- an escalation that did nothing to rebut the trust argument and a lot to reinforce why OpenAI made it in the first place.

The $60 billion all-stock deal for Anysphere is being described as the largest venture-backed startup acquisition on record, and it folds Cursor into the SpaceXAI division that was created earlier in 2026 when SpaceX absorbed xAI ^[6]. Cursor's four co-founders - Michael Truell, Aman Sanger, Sualeh Asif, and Arvid Lunnemark - are each projected to be worth roughly $2.7 billion after the deal ^[6], and the tool they built processes an estimated 150 million lines of code per day ^[6].

That volume of developer activity is the real prize. Reddit's read on the deal leaned heavily toward SpaceX paying for billions of developer prompts and coding interaction data to sharpen Grok's coding ability, not for Cursor's editor UI itself - with some commenters noting the all-stock structure means it was closer to a valuation swap than a cash outlay. Seen that way, OpenAI's cutoff isn't really about punishing Cursor - it's about refusing to keep subsidizing a rival's model-training pipeline with GPT access, once that pipeline sits inside a company Musk controls. Vertical integration of compute, coding data, and models under one roof is exactly the kind of consolidation that turns a former customer into a strategic threat overnight.

This isn't the first time an AI lab has pulled model access over an ownership change. When OpenAI's planned acquisition of the coding tool Windsurf was reported in mid-2025, Anthropic cut Claude's access to Windsurf within days - and that episode is now the template OpenAI is following with Cursor. Forkast News analyst Lena Park frames the pattern as "a transition from collaborative AI development to a landscape where model access is a primary instrument of corporate warfare" ^[4].

Anthropic is positioned to be the immediate beneficiary again: co-founder Tom Brown confirmed the company is "stepping in with increased compute capacity to support Claude models on Cursor" ^[4]. That's consistent with a broader shift in enterprise share - Anthropic's Q2 2026 revenue reportedly reached $11.5 billion versus OpenAI's $6.7 billion, with Anthropic holding roughly 40% enterprise spend share ^[4]. The wrinkle worth watching, per reporting on the deal, is that Anthropic itself already leases compute from SpaceX's Colossus data center infrastructure, a dependency that could eventually complicate the idea of Claude as a fully neutral alternative if SpaceX-linked leverage becomes a two-way street across the industry rather than a one-off targeted at OpenAI and Cursor.

For most Cursor users, the practical disruption looks small. OpenAI models reportedly account for only about 5% of Cursor's traffic ^[2], and after November 12 developers can still reach GPT models by registering personal OpenAI API keys inside Cursor on a pay-per-token basis, using OpenAI's dedicated Cursor IDE extension, or switching to alternative models entirely ^[2]. On Cursor's own subreddit, the discussion carried a noticeably lower-key tone than the broader tech-news threads - many users said they had already migrated to Claude, Grok, or other third-party models well before this announcement, partly because Cursor's flat per-token pricing already made non-OpenAI models the more attractive default.

The bigger cost isn't measured in traffic share - it's measured in what developers can no longer assume. Cursor had marketed itself as model-agnostic, neutral infrastructure sitting on top of whichever lab's models performed best. This episode, layered on top of the Windsurf-Anthropic precedent, tells developers that a coding tool's model roster can be redrawn overnight by a corporate feud that has nothing to do with product quality. That's a real planning risk for teams building workflows around any single-vendor model integration, regardless of which lab currently looks safe.