OpenAI notified SpaceX on 28 August that it intends to wind down the contract that puts its models inside Cursor, with a proposed shutoff date of 12 November. Cursor is the coding assistant that SpaceX bought in an all stock deal worth roughly $83 billion and folded into its AI division earlier this month.
The two companies had worked together for nearly four years. For the engineering teams that standardised on Cursor over the past two years, plenty of them in Melbourne and Sydney, this is a supplier decision taken well above their heads, and the ten week countdown has already started.
What OpenAI actually announced
OpenAI set out its reasoning in a short company post. The custom agreement with Cursor contains a change of control clause that gives OpenAI a limited window to cancel after an ownership change, and SpaceX completed its purchase of Anysphere, the company behind Cursor, earlier in August. OpenAI says it is giving the maximum notice the contract allows, according to its announcement.
The date is not the only part that matters. OpenAI also confirmed it will not ship future models to Cursor in the meantime, and named Astra, a model it has not yet released, as one of them. Cursor users are on a frozen menu between now and November rather than a shrinking one.
Why OpenAI says it walked away
The stated reason is trust rather than money. OpenAI wrote that it cannot be confident SpaceX will use its technology within its terms of service, based on its experience with Elon Musk’s companies violating contracts. It pointed to two specific episodes rather than leaving the claim general.
The first is Twitter, now part of SpaceX, which OpenAI says broke the terms of its contract after Musk acquired it. The second is xAI, also folded into SpaceX, which Musk admitted under oath earlier this year had violated OpenAI’s terms of service. Musk sued OpenAI over its restructuring and a federal jury ruled against him this year, finding he had filed too late, as Reuters reported.
Michael Truell, a Cursor founder who is now a SpaceX executive, said his team was speaking with OpenAI to resolve the issue and noted that Cursor was one of OpenAI’s earliest customers. Neither side has described what a narrower arrangement would look like.
What Cursor users actually lose
Cursor is not going dark. It serves models from Anthropic, Google and xAI alongside OpenAI’s, so the practical effect is the loss of one supplier from a menu of four. Teams that already route most of their work through Claude or Gemini inside Cursor may barely notice the change.
The teams that will notice are the ones that tuned their workflow to the GPT line. Prompts, custom rules and evaluation suites built around one model family do not port across for free, and anyone who has run that migration knows it costs more than a change of setting. Our comparison of Claude Code and Codex covers where each one holds up.
There is a pricing question underneath it. Cursor stopped showing dollar costs on its usage page this month, which makes it harder for a finance team to model what a forced switch does to the monthly bill before it happens.

The alternatives are already crowded
Anyone forced to look around will find the coding assistant market busier than it was a year ago. GitHub Copilot remains the default inside Microsoft shops, Claude Code and Codex both run in the terminal rather than the editor, and the gap between them is now measured in agent behaviour rather than autocomplete quality.
The honest answer for most teams is that no single tool wins on every task, which is why our Cursor and GitHub Copilot comparison ends on workflow fit rather than a score. What changed this week is that tool choice and model choice are no longer the same decision.
Why Australian businesses should care
The lesson here is not really about Cursor. It is that the model inside your tooling can be withdrawn by a third party you hold no contract with, for reasons that have nothing to do with you or your usage. Cursor did not lose access because of anything its customers did. It lost it because its owner changed.
For Australian teams that means treating model access the way you already treat any other single supplier. Know which model family each workflow depends on, keep a tested fallback, and write down the switching cost before you need it rather than during a scramble. That belongs in the same conversation as any other AI strategy decision, next to budget and data handling.
Ten weeks is enough time to do this properly if it starts now. Pull the list of repositories and teams that actually call OpenAI models through Cursor, run a fortnight of real work against Claude or Gemini in the same editor, and record where the output quality or the token bill moves. The teams that leave it to late October will be making the call on vibes instead of numbers.
Model supply is now a procurement question
Nothing here is settled. OpenAI called 12 November a proposed shutoff date and Truell says the two sides are talking, so a narrower deal remains possible. Australian teams should plan as though it lands anyway, because the notice period is the whole of the warning they get.
The wider signal is harder to unwind. A frontier lab has shown it will cut a major distribution partner over who owns it, which turns model supply from a technical choice into a procurement one. The teams that come out of this well will be the ones who already knew what their second option was.
