PATRICK T. FALLON

Blackstone Group (NYSE: BX) president and chief operating officer Jon Gray said on Tuesday (November 3) that spending on AI models by the firm's investee companies had risen 20-fold over the past year, as businesses saw financial benefits from applications including coding, customer interaction and content creation. He also said demand for computing power was running ahead of supply. Reports seen so far have not disclosed the base figure behind the 20-fold increase or how many companies it covered.

Gray said at the same event that AI infrastructure required enormous amounts of capital, while debt investors faced concentration risks across the sector and in individual companies, creating a need for more equity financing. He expected most large technology companies to raise funds again. Emmanuel Roman, chief executive of Pacific Investment Management Company (Pimco), warned during the same discussion that AI financing could reach a scale the market would struggle to absorb. Gray said some scepticism in the market was healthy, adding that some AI investments would fail. The key, he said, was to invest in a way that best protected against downside risks.

US technology stocks fell sharply in October ahead of the summit, while Alphabet's latest quarterly results renewed questions over whether its huge AI spending could generate returns. Morgan Stanley chief executive Ted Pick, who also attended the event, estimated that AI build-out could require US$10 trillion to US$15 trillion over the next decade. Goldman Sachs chief executive David Solomon said demand for computing power still exceeded supply, but that the long-term value could flow to companies developing AI applications.

Documents on Oracle's (NYSE: ORCL) fundraising plans emerged a day before Gray's remarks. The company filed two documents with US regulators on Monday (November 2), setting out plans to issue debt and new shares to raise a combined US$20 billion (about HK$156 billion, based on an exchange rate of roughly HK$7.80 to the US dollar). The offering comprises US$10 billion of common stock and US$10 billion of senior notes. Oracle said the proceeds would be used for general corporate purposes, which could include capital expenditure, debt repayment, acquisitions, dividends and share buy-backs; it did not specify how much would be used for AI infrastructure. Oracle said last week that capital expenditure would peak in the current and next financial years, and pledged to raise no more than an additional US$20 billion for AI before the 2030 financial year. The stock has fallen about 30% year to date. As at August 31, notes payable and other borrowings stood at US$125.34 billion. On Wednesday, the chief financial officer said free cash flow was expected to turn positive in the second half of the 2028 financial year, sending the shares up about 11% that day.

On the demand side, Palantir (Nasdaq: PLTR) reported third-quarter results on Monday. Revenue rose 96% year on year to US$2.313 billion (about HK$18 billion), with US commercial revenue reaching US$984 million and exceeding US government revenue of US$953 million for the first time. GAAP net profit was US$1.128 billion, up about 137% from US$476 million a year earlier.

As for the scale of computing procurement, reports citing a draft of Anthropic's IPO prospectus said the company had committed to spending US$111 billion on Google cloud services, US$110 billion on Amazon Web Services and US$31 billion on Microsoft, regardless of actual usage. It had also allocated more than US$160 billion to rent Google chips, taking total infrastructure costs to more than US$600 billion in the future. These figures came from reports about the draft prospectus and do not represent formally announced agreements.