OpenAI plans to wind down its contract supplying models directly to Cursor after SpaceX acquired Anysphere, the company behind the AI coding editor. OpenAI has proposed November 12, 2026, as the cutoff date, so this is a planned change—not a completed shutdown. The company says it is acting because it cannot be confident that SpaceX will use its technology within OpenAI’s terms of service.
That decision puts a major customer relationship under strain just after a reported $60 billion all-stock acquisition. It also exposes a larger risk for developers: access to a hosted AI model can depend not only on technical performance, but on who owns the platform using it.
OpenAI has chosen a future cutoff, not an immediate shutdown
OpenAI announced on August 28, 2026, that it intends to wind down its model-supply contract with Cursor and will not provide future models to Cursor under the planned arrangement. The proposed end date is November 12, 2026.
In practical terms, developers should not read the announcement as proof that Cursor has already lost OpenAI access. It establishes OpenAI’s decision and its proposed timetable, while the final outcome of the transition remains unsettled.
The immediate question is whether Cursor will be barely affected because OpenAI models reportedly handled only a small share of its traffic. The answer is not that simple. Traffic share does not by itself measure revenue, token consumption, the importance of particular customers, or the value of a model in a product’s workflow.
Why SpaceX’s acquisition changed the relationship
SpaceX announced a reported acquisition of Anysphere on June 16, 2026, in an all-stock transaction valued at approximately $60 billion. Anysphere is Cursor’s parent company, making the ownership change central to OpenAI’s decision.
OpenAI’s stated rationale is compliance and trust. The company said it could not be confident that SpaceX would use its technology within its terms of service, citing its experience with Elon Musk’s companies and contract violations. OpenAI has not publicly provided an itemized explanation of the specific terms it believes are at issue.
That is distinct from the wider strategic theories surrounding the breakup. The timing also places OpenAI and Cursor in a more competitive relationship in AI-assisted software development. Reporting has discussed possible concerns about model distillation and competitive pressure, but OpenAI has not confirmed those as the formal reason for ending the contract.
The distinction matters. OpenAI’s compliance concern is a stated reason. Competitive tension and model-distillation risk are additional reported possibilities, not established explanations that replace the company’s public account.
The money behind the breakup
The financial context makes the decision more consequential than a routine supplier change. Cursor was reported to be one of OpenAI’s five largest customers by revenue at the start of 2026. It was also reportedly projected to generate more than $1 billion in annualized revenue for OpenAI by spring 2026.
Those figures describe a reported customer ranking and a projection, not a disclosed realized loss. They do, however, explain why the planned wind-down stands out: OpenAI appears willing to give up a commercially significant relationship after the change in ownership.
The reported acquisition value and revenue projection measure different things. The approximately $60 billion figure refers to the reported SpaceX–Anysphere transaction. The more-than-$1-billion figure refers to projected annualized revenue for OpenAI from Cursor. Neither number shows how much money OpenAI will ultimately lose, because the final financial impact has not been disclosed.
The 5% traffic dispute
Michael Truell, Cursor’s co-founder and CEO, said OpenAI models served about 5% of Cursor user traffic. That figure was presented to emphasize Cursor’s ability to rely on other model providers.
OpenAI executive Thibault Sottiaux challenged that interpretation, saying token usage is not a proxy for revenue or value created and asking Truell to share the calculation behind the figure. The two claims are not necessarily contradictory: a model can represent a small share of user traffic while still being commercially important, depending on how usage, pricing, and customer value are measured.
So does the 5% figure mean Cursor users will barely notice the change? No—not by itself. It describes one traffic measure attributed to Truell, but it does not settle the business importance of OpenAI models or predict the final effect on developers.
What Cursor users may see next
The available transition path is not final. Cursor has used multiple model providers, and Anthropic was reported to be increasing the Claude compute available to Cursor. That gives Cursor another potential source of capacity, but it does not establish the complete replacement plan or guarantee that every workflow will behave the same way.
OpenAI’s announcement does establish that the planned wind-down concerns direct contract-based supply and future models. It does not establish that all possible ways of accessing OpenAI models inside Cursor will work after the proposed date, so developers should avoid treating unofficial workarounds as a confirmed product path.
For teams that depend on Cursor, the sensible operational question is not simply whether OpenAI represents 5% of traffic. It is which models each project uses, whether those models are interchangeable for its coding tasks, and how much disruption a provider change would cause. Those are the dependencies worth mapping before November 12, 2026.
The wider platform-risk lesson
Cursor’s situation illustrates the trade-off behind multi-model developer platforms. Connecting several providers can reduce dependence on any one model, but it does not eliminate dependence on contracts, usage policies, pricing decisions, or changes in corporate ownership.
A change of control can alter the risk calculation overnight. Before the acquisition, OpenAI and Cursor were model supplier and customer. After SpaceX acquired Anysphere, OpenAI treated the new ownership structure as a reason to reconsider whether continued access fit its terms and trust requirements.
That matters beyond Cursor. Developers building on hosted AI services should distinguish between a product’s interface and the providers underneath it. A polished coding editor may make model switching feel invisible, but the commercial relationship behind the interface can still determine which models remain available.
The bottom line is straightforward: OpenAI has announced a planned wind-down of Cursor’s direct model supply, with November 12, 2026, proposed as the cutoff. The move follows SpaceX’s reported acquisition of Anysphere and comes despite Cursor’s reported importance as an OpenAI customer. Until the companies announce a different arrangement, the practical consequence is uncertainty for developers—and a sharp reminder that model access is also a question of ownership and contracts.
