Cognition’s $40B Funding Talks After SpaceX’s $60B Cursor Deal
SpaceX paid $60 billion for Cursor. Days later, Cognition was discussing a $40 billion-plus valuation and publicly saying it was not for sale. That is real leverage.
SpaceX paid $60 billion for Cursor. Days later, Cognition was reportedly discussing a valuation of at least $40 billion and publicly saying it was not for sale.
That is what leverage looks like when the buyer has more money than most countries and still cannot assume it gets the deal.
On August 19, Bloomberg reported that SpaceX had held acquisition discussions with Cognition, the AI coding company behind Devin. The report said those discussions were no longer active, although the companies remained in conversation about possible collaboration, including Cognition using SpaceX computing infrastructure. On August 20, Cognition chief executive Scott Wu publicly denied that Cognition was for sale and said there had been no acquisition talks.
Those two accounts plainly do not line up. Fine. That is not the most important part.
The important part is that SpaceX’s reported interest landed days after it completed its $60 billion all-stock acquisition of Cursor, the coding platform owned by Anysphere. Cognition is reportedly in early funding discussions that could value it at at least $40 billion, after raising $1 billion at a roughly $26 billion valuation in May.
Read that again. In the middle of the biggest AI spending binge in history, a company can raise $1 billion, discuss a valuation north of $40 billion and tell the world it is not for sale—even after its most obvious strategic buyer has just written a $60 billion cheque for its closest category rival.
That is not normal M&A. It is a land grab for the interface between AI models and actual work.
SpaceX bought Cursor because models are not the business
Everyone is obsessed with the big AI labs: the models, the chips, the data centres, the absurd capital expenditure. Fair enough. Those are enormous businesses.
But coding agents are where the money starts touching a customer’s operating budget.
A model that can write, test, fix and maintain software is not merely a clever chatbot. It can sit inside the workflow of a company with thousands of engineers. It sees how work is specified, reviewed, broken down, shipped and measured. That is valuable because software is not just an industry anymore. It is the plumbing under almost every industry.
Cursor had grown into one of the most prominent products in AI-assisted coding, and Reuters reported it had roughly $2.6 billion in annualised business-to-business revenue before the acquisition. SpaceX completed the deal on August 14, folding Cursor into a company already trying to turn its AI operation into a serious rival to OpenAI and Anthropic.
The $60 billion headline gets attention because it is ridiculous money. But the strategic logic is more interesting than the number.
SpaceX did not buy Cursor merely to own another AI product. It bought a product used by developers, a route into enterprise software budgets, a massive stream of high-value coding interactions and a team that knows how to package AI into something people pay for every month.
That is the real asset: distribution wrapped around habit.
A foundation model can improve quickly. Customer behaviour moves far more slowly. Once a developer’s workflow, company codebase, permissions, security checks and team processes are embedded in a tool, changing it becomes annoying, risky and expensive. That is how a useful AI feature becomes a business.
Cognition’s position is exactly why it may not want to sell
Cognition is not sitting there with a slideshow and a dream. Its product, Devin, is designed to handle engineering tasks through an AI agent rather than simply suggest lines of code. Bloomberg reported that Cognition’s annualised revenue run rate was approaching $1 billion as it explored fresh financing.
If those figures hold, Scott Wu has every reason to resist becoming a footnote inside somebody else’s conglomerate.
Founders sell for three broad reasons: they need capital, they need distribution, or they have run out of stomach for the next phase. Cognition appears to have access to capital. It has strong momentum in a market where customers are actively hunting for productivity gains. And a $40 billion-plus prospective valuation gives its shareholders a perfectly rational reason to believe independence could be worth more than a quick exit.
This is what plenty of buyers miss. A high valuation does not automatically make a founder eager to sell. Sometimes it does the opposite.
At $2 billion, an exit can feel life-changing. At $26 billion, with a new round potentially priced at $40 billion or more, the founder starts asking a different question: why sell now if we believe we can be the category owner?
That changes deal dynamics completely. The buyer is no longer offering rescue capital. The buyer is trying to persuade a rival that its independent future is worse than joining the empire.
And that is a much harder pitch.
The overlooked angle: compute may matter more than ownership
The more interesting part of the reported SpaceX-Cognition discussions is not the alleged acquisition. It is the possibility of a computing relationship.
The AI industry has spent two years pretending every important strategic relationship must end in an acquisition. That is lazy thinking.
A company like Cognition needs enormous computing capacity to build and run increasingly capable coding agents. SpaceX wants customers and workloads for the computing infrastructure it is trying to build. A commercial partnership could serve both sides without Cognition surrendering its product, its brand, its independence or its upside.
That may be the better deal for Cognition.
It may also be the better deal for SpaceX. Buying a company is expensive, disruptive and painfully difficult to integrate. Selling it infrastructure, or winning it as a major customer, can generate revenue while keeping the target motivated and moving fast.
There is a blunt operator lesson here: ownership is not always the best commercial outcome.
Too many acquirers see a good company and jump straight to “How do we buy it?” A smarter first question is: “What is the smallest deal that makes us strategically indispensable?”
If SpaceX can become a crucial provider of compute to AI application companies, it does not need to own every application. It can collect a toll from the ones that win.
That is less glamorous than a splashy $60 billion takeover. It may also be a better business.
Why this is a warning for founders chasing an exit
There is a nasty trap in all this for founders reading M&A headlines and fantasising about a strategic buyer arriving in a black car with a nine-figure offer.
Cursor did not become worth $60 billion because it was merely an AI company. It became strategically valuable because it had a product customers used, a rapidly growing revenue base, developer distribution and a position in a critical workflow.
Cognition’s reported funding interest tells the same story. The market is not rewarding AI labels. It is rewarding companies that own a valuable wedge into real work.
If you are building a business, stop trying to look acquirable. That often produces a shallow company with a tidy pitch deck and no real leverage.
Build the thing a buyer cannot easily reproduce:
- Direct customer relationships. - A product embedded in daily behaviour. - Data created through legitimate customer use. - A distribution channel that does not depend entirely on paid ads. - Revenue quality that improves as you scale, rather than costs that explode with every new customer.
Do that, and you may get acquisition interest. More importantly, you will be able to say no to it.
That is where bargaining power lives.
What this means for you
For founders: do not treat an acquisition approach as validation. Treat it as information. Ask what the buyer cannot build, what they need from you, and whether a commercial partnership gets you most of the upside without giving away the company. If the answer is yes, do not sell just because a famous name has called.
For investors: stop valuing AI businesses solely on model quality or growth screenshots. Look for workflow ownership, retention, enterprise penetration and whether the company has a credible path to fund its own next phase. The best deal targets increasingly have the financial strength to remain independent.
For operators: this is your reminder to secure strategic leverage before you need it. Build more than one source of distribution. Build more than one source of infrastructure. And do not let a giant supplier become so essential that it can dictate your future.
SpaceX’s $60 billion Cursor purchase is not proof that every strong AI company should sell. Cognition’s position is the more useful signal.
When you have real revenue, real customers and a credible independent path, the richest buyer in the room does not own you. They have to earn the right to partner with you.