OpenAI Cuts Off Cursor After SpaceX Buyout

ChatGPT

OpenAI is winding down its model supply contract with Cursor after SpaceX's 60 billion dollar acquisition of parent company Anysphere triggered a change-of-control clause, escalating the long-running feud between Sam Altman and Elon Musk.

Staff

August 31, 2026 · 4 min read

OpenAI notified SpaceX on August 28 that it intends to wind down its contract supplying models to Cursor, the popular AI coding editor, with a proposed shutoff date of November 12, 2026. The trigger was SpaceX’s roughly 60 billion dollar acquisition of Cursor’s parent company Anysphere, which closed on August 14 and immediately opened a change-of-control clause in OpenAI’s existing agreement. OpenAI says it’s giving the maximum notice period its contract allows, but the underlying message is unambiguous: it no longer trusts a company under Elon Musk’s control to use its technology within the bounds of its terms of service.

The stated reason, in OpenAI’s own words

In its own announcement, OpenAI framed this explicitly as a change-of-control decision rather than any complaint about Cursor’s product or its developers. “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,” the company wrote, pointing to two specific precedents: Musk allegedly cutting off OpenAI’s paid license to Twitter’s data after acquiring the platform, and Musk’s own sworn admission earlier this year that xAI, now also under the SpaceX umbrella following its own acquisition, had violated OpenAI’s terms of service. OpenAI said it will not supply Cursor with any future models either, including its upcoming Astra release, citing what it called a new level of accountability for ensuring that model is used in accordance with its terms as AI capabilities keep advancing.

How much this actually affects Cursor

Cursor co-founder and CEO Michael Truell responded on X, posting that OpenAI models currently serve about 5 percent of Cursor’s overall user traffic, a relatively small share that suggests the practical disruption to most Cursor users will be limited. Truell added that the two companies are continuing to discuss the decision. Cursor will still offer models from Anthropic, Google, and SpaceXAI after the cutoff, meaning November 12 ends OpenAI’s specific in-editor integration rather than Cursor’s underlying multi-model approach. Anthropic, notably, has already moved to fill part of the gap, publicly promising more Claude computing capacity for Cursor users following OpenAI’s announcement.

The longer feud this sits inside

This decision doesn’t exist in isolation. Musk sued OpenAI and CEO Sam Altman over the company’s conversion to a for-profit structure, a case that reached trial earlier this year with both Altman and Musk testifying. SpaceX has been aggressively consolidating AI and developer tooling assets under Musk’s control: it acquired X and xAI in February, before going public in June, and completed the Anysphere purchase, the largest acquisition of a venture-backed startup on record, on August 14. Just three days before that deal closed, SpaceXAI launched Grok Bot at 120 dollars per seat per month with direct integration into Cursor Premium Teams, a move one analysis characterized as an immediate signal of intent to use Cursor’s existing developer base to drive adoption of SpaceXAI’s own proprietary models rather than external providers like OpenAI.

Musk’s own reaction to OpenAI’s decision was characteristically blunt. In a post on X, he wrote simply: “I couldn’t care less.”

What this signals for developer tools built on rented models

The broader implication reaches well past Cursor and this specific feud. Model access has traditionally functioned as neutral infrastructure, a coding tool integrates whichever models serve users best, largely independent of who owns the tool itself. OpenAI invoking a change-of-control clause specifically because of who now owns Cursor sets a real precedent: model supply can become a lever in competitive and even personal corporate disputes, not just a commercial relationship insulated from who’s on either side of the ownership table. For any developer tool built on model access licensed from a company that could plausibly become a competitor’s parent someday, this is a concrete example of exactly that risk materializing, not a hypothetical one.

Where this leaves developers actually choosing tools

For anyone currently evaluating AI coding tools, the practical lesson isn’t about Cursor specifically, it’s about ownership structure as a genuine, ongoing risk factor. A coding tool that felt neutral yesterday can stop being neutral overnight if it changes hands, and the contract terms that make that possible are rarely visible to the end user choosing which editor to install. Cursor’s 5 percent OpenAI traffic share made this particular transition manageable, but a similar change-of-control clause triggering against a tool where a single model provider accounts for the majority of usage would be a considerably more disruptive migration for the developers caught in the middle.

A rivalry now shaping product decisions directly

What makes this specific case notable is how directly personal animosity is now shaping enterprise product decisions at scale. OpenAI’s own announcement leaned heavily on Musk’s prior conduct with Twitter and xAI as justification, not on anything Cursor or its developers did. That’s a different category of business decision than a typical competitive or strategic move, and it suggests the Altman-Musk conflict, which has already produced years of litigation, competing product launches, and public sniping, has now graduated to actively restructuring which companies can access which foundation models, with real developers absorbing the disruption in between.

See OpenAI’s own announcement for the complete statement.

Up Next

Nvidia Agrees to Buy Hugging Face for $12.9 Billion

Big Tech

Nvidia has reportedly agreed to acquire Hugging Face, the leading open-source AI model hub, for 12.9 billion dollars, in what would be Nvidia's largest acquisition ever and a major vertical move into AI software distribution.

Staff

August 31, 2026 · 5 min read

Nvidia has agreed to acquire Hugging Face, the open-source hub where developers share, download, and collaborate on AI models, for 12.9 billion dollars, according to The Information, a figure other outlets have separately reported as closer to 13 billion once the full deal terms are counted. Neither company has confirmed the deal publicly and no signed contract has been reported yet, but the reporting has since been echoed by Bloomberg, CNBC, Forbes, and Tom’s Hardware, all citing people familiar with the negotiations. If it closes, this would be Nvidia’s largest acquisition ever, more than double its abandoned 40 billion dollar bid for Arm and far beyond the 6.9 billion dollars it paid for Mellanox in 2020.

What Hugging Face actually is, and why that matters here

Hugging Face operates something close to a GitHub for AI models: as of last year the platform hosted more than 2 million models and hundreds of datasets, used by over 13,000 companies to find, test, and deploy open-source AI directly. It’s become genuinely central infrastructure for anyone working with open-weight models, the neutral ground where labs publish releases and developers go to compare and download them regardless of which company built the underlying model. That neutrality is precisely what makes Nvidia’s ownership a meaningfully different proposition than, say, another chipmaker buying a hardware supplier: it would put the primary distribution point for open-source AI development under the same company that already dominates the hardware those models run on.

Nvidia isn’t a stranger to Hugging Face. The company participated in Hugging Face’s Series D funding round back in August 2023, investing 235 million dollars at a valuation of just 4.5 billion dollars at the time. A 12.9 billion dollar price now represents nearly a three-fold jump in under three years, a reflection of just how much more central the open-model ecosystem has become to the broader AI industry since then.

Why Nvidia wants this specifically

Nvidia’s own public framing has emphasized that it doesn’t discriminate between open and closed models, it wants to be the compute layer underneath all of them. Buying the platform where open models actually live extends that ambition directly into software and distribution, not just hardware. There’s also a more concrete commercial angle: Hugging Face already runs paid compute services letting developers run inference or fine-tune models without managing their own GPU clusters, a business Nvidia could inherit and scale rather than having to rebuild from scratch, avoiding the awkward optics of directly relaunching its own DGX Cloud rental service while still gaining a real foothold in the rental compute market, one layer removed from the hyperscalers Nvidia still depends on to actually sell most of its chips.

The timing lines up with a defensive motive too. Nvidia’s AI hardware dominance faces a genuine long-term threat as Anthropic, Google, OpenAI, and other major labs increasingly develop their own custom accelerators, reducing how dependent they are on Nvidia GPUs specifically. Owning the platform where the broader open-source ecosystem discovers, tests, and deploys models gives Nvidia continued relevance and influence over that ecosystem even as individual frontier labs work to reduce their direct hardware dependence on the company.

Part of a broader consolidation wave

Hugging Face reportedly began fielding acquisition interest from another, unnamed suitor before Nvidia entered serious talks, and the deal lands in a month already thick with AI infrastructure consolidation. Stripe recently paid more than 7 billion dollars to acquire OpenRouter, the startup that helps developers route requests across different AI models, itself valued at just 1.3 billion dollars in a Series B round only months earlier. Hugging Face CEO and co-founder Clem Delangue said in June that paying subscribers had doubled during the first half of 2026 and that the company was nearing profitability on its own, momentum that likely factored into the price Nvidia ultimately agreed to.

The regulatory question already forming

European authorities have reportedly already begun examining the proposed acquisition, and Nvidia’s own history gives regulators a clear precedent to draw on. In April 2024, Nvidia agreed to acquire Israeli AI orchestration startup Run:ai; the European Commission reviewed that deal under Article 22(3) of the EU Merger Regulation, a provision that lets the Commission examine a transaction even when it falls below the normal turnover thresholds that would otherwise trigger review, and ultimately cleared it unconditionally in March 2025, with the roughly 700 million dollar purchase closing that December. A deal nearly 20 times larger, putting the dominant distribution hub for open-source AI under the world’s dominant AI chipmaker, is a considerably higher-stakes case for antitrust regulators to weigh, and the vertical integration concern here (hardware plus the software distribution layer riding on top of it) is a more direct one than Run:ai’s narrower orchestration-tooling business presented.

What Hugging Face’s own community stands to lose or gain

Hugging Face’s value has always rested heavily on being perceived as neutral ground, a place labs from OpenAI to Meta to small independent researchers all publish to on equal footing, regardless of who they compete with elsewhere. That perception of neutrality is precisely what a Nvidia acquisition puts at risk. One Forbes analysis of the deal put it directly: Hugging Face will likely lose its vendor independence, aligning more closely with Nvidia’s own ecosystem, a shift that could effectively tie businesses’ model orchestration decisions to a specific hardware vendor even when they never intended to make that choice. Nvidia’s long track record of strong developer tooling, most visibly through CUDA, offers some reassurance that day-to-day service quality would likely remain high. But the deeper structural question, whether competing chipmakers and labs continue trusting a platform now owned by their biggest rival with the same openness they extended to an independent Hugging Face, is a genuinely unresolved one that won’t be answered until well after any deal actually closes.

See TechCrunch’s original reporting for more on how the talks have developed.