马斯克预测超 200 家 AI 云公司将被整合，SpaceX 拟举债 400 亿美元买 GPU

Winners & losers: Is the AI industry bubble closer than ever to bursting? Elon Musk has just predicted that more than 200 neocloud companies, which rent compute power to firms for AI training and inference, will experience "guppy-level death and consolidation." In the same post, the CEO announced that SpaceX has secured $40 billion in debt to fund more GPU purchases, though one report claims the deal hasn't closed.

Writing in an X post yesterday, Musk said lenders had largely stopped providing money to those companies without substantial income outside AI data centers.

He claimed SpaceX could still borrow thanks to its investment-grade status and revenue from Starlink and rocket launches. But for the smaller players depending on the AI boom, his outlook was a lot less optimistic, warning that they face shutdowns and acquisitions.

"Some will make it through, but not many," Musk wrote.

– Elon Musk (@elonmusk) October 25, 2026

Musk was replying to investor Chamath Palihapitiya, who argued that neocloud operators were poorly funded and organized.

Palihapitiya expects unused computing capacity to become a problem in 2027. These companies, which include CoreWeave and Lambda, often borrow heavily to buy the GPUs they rent to customers. Keeping that model going becomes much harder when lenders lose their enthusiasm.

However, Musk could have revealed SpaceX's deal prematurely. Bloomberg reported that the debt package had yet to be finalized, with pricing discussions continuing.

The GPUs are expected to arrive in the first half of 2027, Fox Business reports, so the purchases would have little impact this quarter.

Anyone expecting Musk to snap up struggling rivals might be disappointed. He dismissed that suggestion, saying SpaceX had no intention of buying neoclouds.

Musk isn't exactly renowned for making predictions that turn out to be accurate, but plenty of others share this belief. Gartner expects more than 80% of neocloud providers could stop operating independently by 2030.

The warning comes days after Citadel founder Ken Griffin said investors were "almost certainly" in an AI bubble. Griffin questioned how consistently businesses were getting value from AI, even as infrastructure spending continued. He also argued that a bubble and a lasting technological transformation could happen together, pointing to the survivors of the dot-com crash.

Money isn't the only problem. It was reported in April that between 30% and 50% of US AI data centers planned for 2026 could be delayed or canceled because of power and electrical equipment shortages. With tighter lending posing another obstacle, having customers eager for more compute may not be enough to get a project finished.