Key Takeaways

  • Cognition's CEO stated the startup is not for sale and has held no acquisition talks with SpaceX.
  • The disputed report reflects growing competition for AI coding agents, talent, and computing capacity.
  • A commercial infrastructure partnership could remain possible even without a takeover.

Cognition's chief executive has pushed back on claims that SpaceX attempted to acquire the AI coding startup, drawing a clear line between a potential commercial relationship and a change of ownership.

In a post on X, the CEO characterized the acquisition information as inaccurate. He stated that Cognition is not for sale and that the startup and SpaceX had not engaged in acquisition negotiations.

The denial followed a Bloomberg report, cited by TechCrunch, that SpaceX had explored buying Cognition while expanding its artificial intelligence operations. The report also indicated that deal discussions were no longer active, although the two entities could still be considering ways to work together.

That distinction matters. Cognition could purchase access to computing infrastructure from SpaceX without surrendering its independence, much as AI developers routinely obtain capacity from cloud and infrastructure providers. SpaceX is starting to sell computing power to external AI businesses, including Anthropic, according to the source material. The chief executive did not specifically address that part of the report.

Advanced coding agents require far more than a strong user interface and a capable model. Training, testing, and operating them at enterprise scale depend on large pools of accelerators, storage, networking, and reliable power. Infrastructure access can therefore become almost as strategically significant as model performance.

IDC estimates that global AI infrastructure spending reached about $153 billion in 2024 and more than doubled to $318 billion in 2025. Its 2026 forecast expects spending to exceed $487 billion to $497 billion, with the market projected to pass $1 trillion around 2029. The United States accounted for roughly 59% of early AI infrastructure spending.

In this competitive environment, vendors such as Cognition, OpenAI, Anthropic, and Google are building AI coding agents and foundation models that require immense GPU and cloud capacity. The broader cloud infrastructure market is projected to reach roughly $462 billion by 2029, representing about 83% of total compute and storage spending, with a 17.8% compound annual growth rate from 2024 to 2029, according to IDC data.

For enterprise buyers, ownership is only part of the story. AI coding agents can access source code, repositories, development environments, and internal documentation. Procurement teams will likely scrutinize data handling, model oversight, access controls, auditability, and vendor continuity. Relevant frameworks guiding trustworthy system design include the NIST AI Risk Management Framework, along with emerging ISO/IEC standards that address lifecycle controls. The Gartner view that generative AI consulting and implementation demand is rising alongside agent and copilot adoption reinforces that deployment involves substantial process work, not merely a software subscription.

For now, Cognition’s position is direct: no sale and no acquisition talks. A compute partnership with SpaceX remains a separate possibility, based on the reporting. In an AI market where infrastructure suppliers increasingly overlap with model developers and application vendors, that kind of limited relationship may prove more practical than another megadeal.