SpaceX is exploring an unusual new source of training material for artificial intelligence: data left behind by failed startups.
Teams inside SpaceXAI have discussed acquiring customer and operational information from troubled or defunct companies, according to people familiar with the matter. The talks remain informal and may not result in a deal.
The idea comes as AI companies increasingly search beyond the public internet for proprietary datasets that can improve models and autonomous agents.
Meanwhile, SPCX stock continued higher Thursday, extending a 5.2% gain from the previous session as investors remain focused on SpaceX's rapidly growing AI business.
SpaceX Looks Beyond the Internet for AI Data
Failed startups can hold something increasingly valuable to AI developers: years of real-world business information.
That could include customer interactions, operational records and other proprietary material unavailable through conventional web training.
SpaceX's interest resembles Google's reported $10 million offer for business data from Spirit Airlines after the airline wound down operations. That proposal also raised questions over how customer and employee information should be handled when corporate data is repurposed for AI.
For SpaceX, acquiring specialized datasets could complement its enormous investment in computing infrastructure.
The company has quickly transformed AI compute into a major business. CFO Bret Johnsen recently disclosed another hosting agreement worth roughly $1.1 billion per month, or about $13 billion in annual recurring revenue, beginning in December.
Johnsen said SpaceX believes it remains on track to reach $100 billion in ARR, with AI compute contracts increasingly contributing to that target.
Coinpaper recently examined SpaceX's fast-growing AI data-center business as reliability problems raised questions about how quickly the company can expand capacity.
SPCX Stock Becomes an AI Trade
The new data strategy adds another dimension to the SPCX stock story.
SpaceX had around 1.4 gigawatts of AI computing capacity at the end of the second quarter, while expectations for its AI revenue have risen sharply. Analysts cited by Barron's now project about $67 billion in AI revenue for 2027, compared with an earlier estimate of $38 billion.
The strategy also illustrates how the AI race is changing. Computing power remains critical, but access to unique training data is becoming another competitive advantage.
Failed startups could therefore have an unexpected second life: even when their businesses disappear, their accumulated data may remain valuable for training increasingly capable AI systems.
For SpaceX, combining proprietary data with its rapidly expanding compute infrastructure could deepen its AI ambitions, and give investors another reason to increasingly view SPCX stock as an AI infrastructure play as much as a space company.
