Musk's AI Unit Considers Buying Data from Bankrupt Startups to Train Grok, Sources Say

Elon Musk's artificial intelligence company, SpaceXAI, has internally discussed acquiring data from distressed or defunct startups, a strategy aimed at sourcing high-quality training material for its AI models at a relatively low cost, according to people familiar with the matter. The discussions are reportedly in an informal stage and may not result in any transactions.

SpaceX did not respond to a request for comment.

A New Frontier for AI Training Data

This potential move signals a strategic expansion in how AI firms source the vast amounts of information needed to train large language models. According to the sources, SpaceXAI, formerly known as xAI, is particularly interested in corporate operational data and customer information to improve the performance of its models, including its flagship AI, Grok.

Until now, the company has primarily relied on data from Musk's social media platform X and information curated by an internal team of specialized "AI tutors." The consideration of external data from failed businesses suggests that as AI models grow in complexity, the demand for diverse, high-quality, and proprietary datasets is outstripping readily available sources. The ability to legally acquire structured, real-world business data is becoming a key competitive advantage in the AI sector.

Parallels and Potential Pitfalls

Ad

The strategy mirrors a similar move by Google earlier this year, which reportedly offered $10 million to purchase commercial data from the bankrupt budget carrier Spirit Airlines for AI training. That proposal, however, drew scrutiny over data privacy, including objections from some of the airline's former employees.

Acquiring data from bankrupt companies could provide a cost-effective source of information, but it also carries significant legal and ethical risks. Issues surrounding customer privacy, consent, and the authorized scope of data use could pose substantial hurdles, as highlighted by the controversy around the Spirit Airlines deal.

A Multi-Pronged Data Strategy

While exploring external sources, SpaceXAI continues to leverage the vast data ecosystem within Musk's corporate empire. Musk recently stated in an internal SpaceX meeting that the company plans to use all of its internal information to train Grok, telling employees directly, "It will also be trained using your data."

This suggests a dual approach: leveraging internal data from companies like SpaceX and X while simultaneously seeking to acquire unique external datasets. The race for AI supremacy is increasingly defined not just by computing power and model architecture, but by access to proprietary data that can give a model a decisive edge in performance and capability.