Cognition CEO Scott Wu is pushing back hard on a report that Elon Musk’s SpaceX tried to buy his AI coding startup. According to TechCrunch AI, Bloomberg reported Wednesday that SpaceX made a run at Cognition as it scrambles to catch OpenAI, Anthropic, and Google in the AI race. Wu shot the story down within hours, writing on X that it was inaccurate, that Cognition “is not for sale,” and that the two companies never held talks.
What stands out here is the timing. This denial lands just days after SpaceX closed its $60 billion acquisition of Cursor, another AI coding startup. So even if the Cognition talks didn’t happen the way Bloomberg framed them, the direction of travel is obvious: SpaceX is buying its way into AI coding, fast.
Why SpaceX is chasing coding startups
SpaceX absorbed Musk’s AI company xAI earlier this year, then went public in June in an IPO that pushed its market cap to nearly $2.3 trillion at its peak. The company has sold investors on big AI ambitions, including data centers in space. But TechCrunch AI notes the xAI business is still early and trailing rivals, and it keeps tripping over Grok’s controversies, from last year’s “MechaHitler” incident to this year’s nonconsensual imagery scandals.
Musk recently told SpaceX staff that in “four or five years, AI will be 99% of the value” of the company. That’s a huge claim, and it only works if SpaceX pulls in far more AI revenue than it does today.
AI-assisted coding is the clearest path to that revenue right now. Anthropic’s rapid growth, driven heavily by Claude Code, is the proof point everyone in the industry is watching. Buying Cursor was step one. This month, Cursor and SpaceX jointly shipped Grok 4.6, a model that scores higher on coding and multi-step agentic benchmarks.
What Cognition would have added
Cognition builds Devin, an AI coding agent, and it comes with an enterprise roster that includes Mercedes-Benz, Citi, and Goldman Sachs. For SpaceX, that customer base is the prize. Enterprise coding contracts are sticky and lucrative, exactly the kind of revenue Musk needs to back up his 99% claim.
Here’s the wrinkle Wu didn’t touch. Bloomberg reports the acquisition talks are dead, but the two companies are still discussing a partnership, possibly with Cognition renting SpaceX’s computing capacity. SpaceX already sells that capacity to other AI players, including Anthropic, until it needs the compute itself. Wu denied the sale. He said nothing about a compute deal.
The bigger picture for the industry
Cognition is one of the last large independent AI coding startups that hasn’t been swallowed by a major model maker. It raised $1 billion in late May at a $25 billion post-money valuation, and TechCrunch AI reports it’s now in early talks for a new round at a $40 billion valuation. That’s a fast markup, and it signals investors still see standalone value here.
The company knows the consolidation game well. Last year it scooped up the remaining assets of rival Windsurf after Google DeepMind acqui-hired that startup’s CEO and top researchers in a $2.4 billion talent-and-licensing deal. Cognition then cut 30 employees and offered buyouts to the 200 Windsurf staffers, with those who stayed facing 80-plus-hour weeks and six days in the office. That wartime intensity would fit right in at a Musk company, given Musk says he works up to 120 hours a week.
A few things to watch from here:
- Consolidation is accelerating. Cursor is gone. Windsurf is gone. Cognition is now the biggest independent left standing, which makes it a target whether or not this specific report was accurate.
- Compute is leverage. Even without a sale, SpaceX’s capacity gives it a way to wire itself into rivals’ operations.
- Enterprise is the battleground. Whoever locks in the big banks and automakers wins the monetization race.
Cognition says it’s staying independent. In a market where the giants keep writing ten- and eleven-figure checks, that stance gets harder to hold every quarter.