Skip to content
NewsFunding & M&A

SpaceX sought to acquire Cognition days after closing Cursor; CEO Scott Wu denied talks on X

· by Pondero Newsdesk

The short version

Bloomberg reported August 19 that SpaceX approached Cognition AI about a potential acquisition less than a week after closing its $60 billion Cursor deal. CEO Scott Wu denied the report on X, saying the company is not for sale. Deal talks are over; a compute partnership is under discussion.

SpaceX sought to acquire Cognition days after closing Cursor; CEO Scott Wu denied talks on X

Four days after SpaceX officially closed its $60 billion purchase of Cursor, the company approached Cognition AI about a second acquisition, Bloomberg reported August 19. Cognition CEO Scott Wu denied the report within hours on X, writing that the company "is not for sale" and that "the two companies haven't been in talks." The deal discussions are no longer active, Bloomberg and TechCrunch both confirmed. What is active: early talks over a compute partnership that would give Cognition access to SpaceX's GPU infrastructure, a detail Wu did not address in his denial.

What happened

SpaceX closed its Cursor acquisition on August 14, per TechCrunch. The deal, first announced as an option agreement in April, gave SpaceX a $60 billion all-stock takeover of the AI code editor and access to its developer user base. Cursor immediately began describing its GPU access: the startup said it would run on "the largest fleet of GPUs in the world" as part of SpaceX's infrastructure.

The Bloomberg story, citing people familiar with the matter, said SpaceX moved on Cognition while the Cursor ink was still fresh. Cognition did not engage with the approach, one of Bloomberg's sources said, per TechCrunch's coverage. Neither SpaceX nor Cognition responded to requests for comment from reporters.

Wu's X post called the story "inaccurate" and stated Cognition "is not for sale," adding that the two companies "haven't been in talks." His statement left one specific Bloomberg claim unaddressed: the reported discussions about Cognition using SpaceX's computing capacity. TechCrunch flagged that gap. Wu did not follow up.

Why Cognition was the target

Cursor's enterprise client list was thin at the time of acquisition. Cognition's Devin, by contrast, already had paying customers at Mercedes-Benz, Citi, and Goldman Sachs, per TechCrunch. Those three names represent the regulated-industry, large-contract segment that SpaceX's AI business is trying to reach quickly.

The product positioning also differs. Cursor accelerates individual coding sessions inside an IDE. Devin accepts engineering tasks, executes multi-step work autonomously, and returns completed output. SpaceX owning both the developer tool and a production-grade autonomous coding agent would cover more of the enterprise software stack than either acquisition alone.

Musk told SpaceX employees last week that in "about four or five years, AI will be 99% of the value" of the company, per a Yahoo Finance report on the internal message. Anthropic's growth, driven heavily by Claude Code enterprise adoption, is the benchmark SpaceX has to chase. This month, Cursor and SpaceX jointly released Grok 4.6, which the companies said scores higher on coding and multi-step agentic task benchmarks, per TechCrunch. Adding Devin would have extended that stack further up the automation ladder.

The compute offer and what it signals

SpaceX's reported fallback proposal, giving Cognition access to its inference infrastructure, reveals the real card SpaceX is holding: GPU capacity at scale. SpaceX has been renting out computing capacity to third parties, including Anthropic and Google, while it builds toward its broader data-center ambitions, per TechCrunch.

For an AI coding agent that runs continuous inference loops on large enterprise tasks, access to that capacity at competitive rates is a meaningful operating advantage. Devin's workload profile, long-running autonomous tasks for enterprise clients, consumes substantially more inference than a code autocomplete tool. A compute deal that reduces Cognition's GPU spend could improve margins without a full acquisition.

The reported partnership remains early and unconfirmed by either party. Its existence in the Bloomberg story, and Wu's choice not to deny it specifically, makes it the most actionable detail in the episode for anyone watching Cognition's cost structure and customer commitments.

Cognition's independent position

Cognition raised $1 billion in May 2026 at a $25 billion post-money valuation, per TechCrunch. Its fundraising trajectory was already climbing before the Bloomberg story broke: a separate Bloomberg report from August 12 said Cognition entered early discussions with investors for a new round at a valuation of at least $40 billion. That report appeared before the acquisition story, and the acquisition episode has not visibly derailed the fundraise.

Cognition also expanded its product surface area in 2025 through acquisition. After Google DeepMind acqui-hired Windsurf's CEO and research team in a $2.4 billion talent-and-licensing deal, Cognition acquired the remaining Windsurf assets. It then laid off 30 employees and offered buyouts to the 200 remaining Windsurf staff, per TechCrunch. The result was a broader coding product portfolio at the cost of a difficult integration. Cognition now sits as one of the few large independent AI coding companies that has not been acquired by a major model provider.

SpaceX's Cursor acquisition moved from option agreement to closed deal in roughly four months. A similar timeline for Cognition is now off the table, with the CEO's public denial eliminating the path to a friendly deal. SpaceX has one more card to play: it is already a potential compute supplier. Any future approach would likely start from that relationship rather than from a clean acquisition offer.

What to watch next

Two events will determine how this story develops. First, the Cognition fundraise: if the company closes a round above $40 billion in September as the Bloomberg report indicated, the jump from a $25 billion post-money valuation would signal that staying independent was the correct call financially. Second, a formal compute partnership announcement. The GPU access deal was reported but not confirmed. If it closes, it would mark a new type of relationship between SpaceX and the independent coding-agent market: not acquisition, but infrastructure dependency.

The broader AI coding consolidation wave is moving fast. Cursor is inside SpaceX. Windsurf's engineering core is inside Google DeepMind. Cognition is the last sizable independent in the autonomous coding agent space. How it finances and scales over the next quarter will answer whether independent AI-first companies can hold ground against model providers with their own hardware.

Sources