Musk predicted that more than 200 “neoclouds” — newer companies that rent out AI computing power — will consolidate, with only a few surviving.
The episode shows how costly the AI computing race has become. Building large clusters of AI chips can require tens of billions of dollars. Companies with diverse revenue and strong credit can still borrow; firms whose entire business depends on AI infrastructure face a tougher time as lenders become more selective.
Palihapitiya had earlier said many neoclouds appear poorly funded and badly run. He argued that Amazon, Microsoft and Google are moving slowly on deals and prefer to keep computing power for their own use. Delays in building data centers, he said, could leave many chips and half-finished projects sitting idle or failing to earn enough to cover their financing costs in 2027.
Lately, financial markets have become increasingly worried about the payoff from massive AI spending….
