<\/span><\/h2>\nJohn J. Hardy, Chief Macro Strategist at Saxo Bank, notes that SpaceX made a successful stock-market debut at the end of the second quarter. The listing made Elon Musk, at least on paper, the world\u2019s first trillionaire after the company floated a small portion of its shares and raised an impressive $87.5 billion.<\/p>\n
SpaceX fits, in a somewhat unconventional way, into the AI narrative and the belief that future capacity for building additional data centres on Earth could be constrained. Under this vision, part of AI infrastructure could eventually be built in space, where low temperatures and a constant supply of solar energy could create a vast market for AI services worth tens of trillions of dollars.<\/p>\n
For now, however, the company is losing billions on its terrestrial data centres and renting computing capacity to other AI companies, such as Anthropic, apparently without finding a more profitable use for its own infrastructure.<\/p>\n
Leaving aside the profitable launch business and Starlink services, the challenges facing SpaceX\u2019s still very earthbound AI segment are visible elsewhere as well. The frantic pace of current and planned spending by hyperscalers \u2014 the largest builders of data centres, including Google, Meta, Amazon, Microsoft and Oracle \u2014 has not only continued but accelerated in recent months.<\/p>\n
The market consequences are becoming increasingly visible.<\/p>\n
<\/span>Pressure on hyperscalers<\/span><\/h3>\nThe first challenge concerns hyperscalers themselves, as free cash flow has fallen below zero in some cases. These companies have other sources of income: Google and Amazon, for example, run cloud businesses and benefit from strong demand for components such as chips used in their own and third-party data centres.<\/p>\n
Yet the scale of investment has become so large that free cash flow has turned sharply negative for some companies, particularly Amazon and Oracle.<\/p>\n