Analysis
SpaceX approached AI coding startup Cognition AI Inc. about a potential acquisition, Bloomberg reported Wednesday, in what would have been the company's second major AI-startup takeover in under a week. Cognition didn't engage with the approach, and CEO Scott Wu publicly disputed the report, saying the story was inaccurate and that Cognition "is not for sale," adding that the two companies haven't been in talks, according to TechCrunch's write-up of his response.
The approach came just five days after SpaceX completed its $60 billion all-stock acquisition of Cursor-maker Anysphere on June 16, and it follows SpaceX's earlier all-stock merger absorbing Elon Musk's xAI in February 2026. Together those two deals mean SpaceX has now built, bought, or attempted to buy three of the highest-profile names in AI coding and model development inside a single year -- a pace of AI-adjacent dealmaking that is unusual for a company whose core business is rockets and satellites.
## What Cognition is and why it's a target Cognition, founded in 2023, makes Devin, an AI coding agent marketed as capable of autonomously handling software engineering tasks rather than just suggesting code completions. The company has partnered with large industrial customers including Mercedes-Benz Group and GE Aerospace, giving it a foothold in enterprise engineering workflows that Cursor's more consumer/developer-focused product doesn't have. Cognition was valued at $26 billion in a May 2026 funding round and is already in early talks for a new round that could value it above $40 billion, according to earlier TechCrunch coverage -- meaning any acquisition approach would have to clear a bar well north of $26 billion just to match where new investors are already pricing the company.
## Reading the denial A specific, on-the-record denial from a startup CEO -- not a "declined to comment," but an explicit "not for sale" -- is a stronger public rebuttal than most acquisition-rumor stories draw, and it puts SpaceX in the position of having a Bloomberg-sourced report about its own dealmaking activity directly contradicted by the other side. That tension doesn't necessarily mean Bloomberg's sourcing was wrong; early-stage acquisition feelers are frequently characterized differently by the two sides involved, with an acquirer's outreach team describing preliminary conversations as "talks" while the target's leadership, especially one mid-fundraise at a rising valuation, has every incentive to shut that framing down publicly and fast.
The report also notes ongoing discussions between the two companies about working together in some other capacity, including potentially routing Cognition's compute needs through SpaceX's infrastructure -- which suggests whatever conversation did happen hasn't ended, it's just not an acquisition conversation anymore, at least not publicly.
## What the Cursor precedent tells us SpaceX's completed $60 billion Cursor deal is the clearest precedent here: Anysphere's investors and leadership evidently found SpaceX's stock-swap terms attractive enough to sell a fast-growing, richly valued AI coding company rather than continue independently toward its own eventual exit. Cognition, mid-raise at a $40 billion-plus target valuation with enterprise contracts SpaceX doesn't have anywhere else in its portfolio, may simply have concluded it doesn't need the same kind of exit Cursor took -- particularly with new capital reportedly available at a valuation acquisition talks would have needed to beat.
What's missing from the coverage so far is any confirmed price SpaceX would have offered, meaning it's not possible to say whether Cognition turned down a rich premium or a lowball approach -- Wu's denial addresses whether talks happened at all, not what was on the table.
