SpaceX just spent $60 billion buying one AI coding startup. Bloomberg says it tried for a second.
Sources told Bloomberg that SpaceX approached Cognition, the startup behind the AI coding agent Devin, about a potential acquisition. Cognition CEO Scott Wu shot the report down on X, calling it inaccurate and stating flatly that Cognition "is not for sale."
A Denial With a Catch
The two accounts aren't as far apart as they first sound. Wu says the companies weren't in talks.
Bloomberg's sources say something slightly different: SpaceX made an approach, and Cognition simply didn't engage with it. Both versions agree no deal is happening right now.
What both sides also agree on is that the relationship isn't over. Bloomberg reports ongoing discussions about Cognition potentially using SpaceX's computing capacity.
That's the same capacity SpaceX currently sells to other AI companies including Anthropic. Wu didn't address that specific detail in his denial.
The Pattern SpaceX Is Following
This is SpaceX's second attempted AI coding acquisition in a matter of weeks. The first one, Cursor, actually closed, for $60 billion, just last week.
SpaceX and Cursor didn't wait for the ink to dry before collaborating. This month the two companies jointly released Grok 4.6, a model that scores higher on coding and complex agentic task benchmarks.
Cognition would have added something Cursor doesn't have: an enterprise customer base that includes Mercedes-Benz, Citi and Goldman Sachs, plus Devin itself as a second coding agent product.
Cognition isn't hurting for options either way. The company raised $1 billion in late May at a $25 billion valuation, and Bloomberg reports it's already in early talks for a new round at $40 billion.
Why AI Coding, Specifically
Musk has been direct about where he thinks the value is. He told SpaceX employees last week that AI will be "99% of the value" of the company within four or five years.
Getting there requires real AI revenue, not just ambition. AI-assisted coding has become one of the clearest ways any AI company actually makes money right now, and Anthropic's growth through Claude Code is the proof point everyone in the industry points to.
xAI, meanwhile, remains early-stage and behind rivals like OpenAI, Anthropic and Google. It's also been dealing with reputational problems, from last year's "MechaHitler" controversy to this year's fight with Minnesota over Grok's role in generating nonconsensual sexual imagery, a legal battle MAIN has covered as it plays out in federal court.
Buying up coding startups is a faster path to enterprise AI revenue than waiting for Grok's reputation to catch up.
Cognition also has its own track record of consolidating rivals. Last year it acquired the remaining assets of Windsurf after Google DeepMind acqui-hired that startup's CEO and top researchers in a $2.4 billion deal. The integration wasn't gentle: Cognition laid off 30 employees and pushed the roughly 200 Windsurf staff who stayed onto more than 80-hour workweeks and six days a week in the office, a pace not unlike the one Musk himself has described keeping.
What This Means for Miami
Cognition's client list, Citi and Goldman Sachs among them, includes major financial institutions with substantial Miami operations and decision-makers.
If Cognition eventually does end up inside SpaceX's orbit, whether through acquisition or a compute partnership, Miami's finance sector should expect its AI coding vendor relationships to keep consolidating around a small number of increasingly Musk-controlled AI products.
That consolidation cuts both ways for local enterprise buyers. Fewer vendors can mean more leverage in negotiations for well-resourced clients, but it also means more concentrated risk if reputational issues at any one company, xAI's ongoing legal troubles included, start affecting products built on shared infrastructure.
