Elon Musk has officially announced the transfer of SpaceX engineering data for the next phase of training his AI model, Grok. According to the billionaire, this information will be directly integrated into the new architecture, which already boasts 2 trillion parameters — nearly double the 1.5 trillion of the V9 version on which the public release of Grok 4.5 is based.

It is important to note that data subject to U.S. export controls (specifically, the International Traffic in Arms Regulations — ITAR) will not be included in the training set. However, the model will still gain access to critically important information about rocket and satellite developments. This is not just a technical step, but a strategic move: SpaceX has accumulated a unique dataset over recent years that no competitor in the AI field possesses.

Scale of Ambitions: From IPO to AI Integration

In parallel, SpaceX continues to expand its AI projects. Following the successful public offering of SPCX shares and the influx of fresh capital, the company has already invested $60 billion in developer tools, including the acquisition of Cursor's parent company. Musk has set a goal to drive SpaceX's revenue to one trillion dollars by 2030, and AI integration is a key element of this strategy.

xAI, for its part, is actively expanding Grok's functionality. Recently, Grok for Excel was announced: users will be able to build financial models, analyze market data, and work with charts based on Grok 4.5. This shows that the xAI ecosystem is built on the same model that will soon begin training on SpaceX engineering data.

Developer Opinions Diverge

The community's reaction to the new version of Grok is mixed. Some developers, who have been simultaneously testing Grok Build and OpenAI's Codex, note that Grok Build already outperforms Claude on complex programming tasks. However, earlier reviewers pointed to contradictory results from Grok 4.5, despite Musk's claims about its "Opus-class" performance.

Cryptalist Analytical Conclusion: The transfer of SpaceX engineering data is not just a model improvement, but the creation of an exclusive competitive advantage. Neither OpenAI nor Anthropic have access to such datasets on rocket and satellite technologies. The only question is whether the new 2-trillion-parameter model can truly surpass Claude and Codex in real-world tests, not just in claims. The AI market is entering a new phase where data becomes the primary weapon.