The $60 Billion Divorce: Why OpenAI Just Cut Off One of AI Coding's Biggest Names

Two weeks ago, Elon Musk's SpaceX closed a $60 billion all-stock deal to buy Anysphere, the company behind Cursor — arguably the most successful AI coding assistant on the market, with roughly $4 billion in annualized revenue and a user base spanning solo developers to enterprise engineering teams.

It should have been a straightforward story about a rocket company making an aggressive move into software. Instead, it became something messier: OpenAI announced late last Friday that it will stop supplying its AI models to Cursor entirely, with existing access shutting off on November 12, 2026, and its next flagship model, internally known as Astra, never arriving on the platform at all.

What makes this story worth paying attention to isn't just the size of the numbers. It's what OpenAI actually said when it pulled the plug — and what that says about how fragile the AI supply chain really is underneath the products millions of developers use every day.

The Official Reason Isn't About Cursor at All

Normally, when a company ends a partnership, it points to product friction, competitive concerns, or a contract dispute over pricing. OpenAI did something unusual: it said, in plain language, that it does not trust Elon Musk's companies to honor its terms of service.

The company cited two specific prior incidents as its evidence:

• A terms-of-service breach it says occurred when SpaceX acquired Twitter.

• A separate violation by xAI that Musk reportedly acknowledged under oath earlier this year during ongoing litigation.

In other words, OpenAI isn't arguing that Cursor did anything wrong. It's arguing that the moment Cursor's parent company became part of Musk's corporate universe — sitting alongside X and xAI — it inherited a trust problem that OpenAI wasn't willing to carry forward.

A Feud a Decade in the Making

To understand why OpenAI reacted this sharply, it helps to rewind to 2015, when Musk co-founded OpenAI alongside Sam Altman as a nonprofit research lab and helped fund its early work. He left the board in 2018 after a mix of internal disagreements over the company's direction and a falling out that included OpenAI hiring away talent he'd wanted for a Tesla AI effort. He also stopped the donations he'd previously committed.

By 2024, the relationship had deteriorated into open litigation, with Musk suing OpenAI, Altman, and President Greg Brockman over the company's shift from nonprofit to a commercial structure — a case that reached trial earlier this year.

Musk's public response to the Cursor cutoff fit that history. Posting on X, he dismissed the decision, called Altman and Brockman untrustworthy in blunt terms, and repeated his long-standing claim that OpenAI's founders had taken control of what was supposed to remain an open, nonprofit effort. Whatever the merits of that fight, the practical effect landed on a third party: the developers who build software inside Cursor every day.

How Much Does This Actually Hurt Cursor?

Here's where the story gets more interesting than a simple corporate feud. Cursor CEO Michael Truell moved quickly to downplay the impact, stating that OpenAI's models account for only around 5% of Cursor's total user traffic — a strikingly small number for a company of OpenAI's stature, and one that says a lot about how the AI coding tools market has evolved.

Why Cursor might genuinely absorb this hit smoothly:

• Grok 4.5 Integration: Cursor already ships Grok 4.5, a model it co-developed with SpaceX's AI division back in July, available across every pricing tier. That partnership gave Cursor a frontier-level in-house option well before this dispute began.

• In-House Proprietary Model: Cursor has been building its own coding model, Composer 2.5, based on an open-source checkpoint — a direct hedge against depending on any single outside lab.

• Anthropic Expansion: Rival lab Anthropic moved almost immediately after OpenAI's announcement to offer Cursor expanded support and higher usage limits for Claude, giving developers a fast alternative.

The Bigger Picture: Cursor is widely regarded as the most valuable software company ever built primarily on top of someone else's AI models, and this episode is effectively forcing it to finish a transition toward being a multi-vendor, increasingly self-sufficient platform faster than it might have otherwise.

The Real Story: Who Controls the AI Supply Chain?

Zoom out, and this dispute is really a preview of a bigger shift happening across the AI industry. For the past few years, the intelligence layer inside AI-powered software has functioned almost like plumbing — invisible infrastructure that application companies plugged into without much friction. This episode makes clear that model access can just as easily become a lever, pulled the moment a corporate relationship turns adversarial.

SpaceX's ambitions make this dynamic even sharper. The company is reportedly building out Colossus, a compute cluster aiming for the equivalent of roughly one million Nvidia H100-class chips, as part of a strategy to control both the compute layer and, now, an application layer through Cursor — the same vertical-integration playbook SpaceX used years ago to push launch costs down from roughly $10,000 per kilogram toward the $100 range.

Owning the rocket and the launch pad worked for SpaceX once already. Musk appears to be betting the same logic applies to AI infrastructure and the software built on top of it.

4 Key Lessons for Developers and SaaS Founders

If you're a developer, founder, or engineering lead relying on any AI-powered tool — coding assistant or otherwise — this episode is worth treating as a warning rather than an isolated drama:

1. Multi-vendor architecture isn't optional anymore. Any enterprise software product built on a single AI model provider now carries real business-continuity risk that has nothing to do with the quality of that provider's technology.

2. Corporate ownership changes can silently change your tool's capabilities. An acquisition on one side of a SaaS app you rely on can quietly cut off the AI model powering it on the other side, with a notice period as short as contract terms allow.

3. Watch where the leverage sits. As more application companies get acquired by firms that also control cloud infrastructure and compute, expect more disputes where AI API access becomes the bargaining chip rather than subscription price or product features.

4. Budget for switching costs. If your team's tooling, prompts, or workflow automation are tightly coupled to one AI model's quirks, factor the cost of migrating to a different large language model into your software budget now, not after your provider makes the decision for you.

There's also a timing detail worth noting: OpenAI itself is reportedly eyeing an IPO next year, and Anthropic has held early conversations with bankers about a potential offering that could value it as high as $2 trillion. Every one of these companies is under pressure to show investors that its models are indispensable, not interchangeable — which makes a very public, principle-driven cutoff like this one as much a signal to Wall Street as it is a message to SpaceX.

Final Thoughts

On paper, this is a story about two Silicon Valley billionaires who have disliked each other for the better part of a decade finally letting that dislike show up in a contract termination notice.

Key Takeaway: The model powering your favorite tool is a business relationship first, and a technical choice second.

Cursor appears well-positioned to absorb this particular hit. The next company caught in the middle of a feud like this one might not be.