AI Companies Seek Massive Funding Amid Rising Infrastructure Costs

Technology companies have intensified efforts to secure financing for an estimated $10 trillion in AI-related spending in coming years. The companies need the money for data centers, computer chips, technology and equipment. AI companies have borrowed hundreds of billions of dollars, issued additional shares and developed financing methods that may carry new risks.

Anthropic, which makes the Claude chatbot, began preparations on Monday for an initial public offering expected to raise about $100 billion and become the largest ever. Anthropic's IPO document warned that the proceeds would cover only an initial portion of the company's funding needs. Anthropic said the company anticipates substantial and growing investments for the foreseeable future.

Meta announced on Monday that the company would sell up to $95 billion in new stock. SpaceX finalized a complex $40 billion borrowing deal for AI chips and equipment less than five months after completing the largest IPO to date. Raymond James Investment Management chief market strategist Matt Orton urged AI companies to obtain funding while available, saying every company would need more.

Columbia University finance professor Stijn Van Nieuwerburgh estimated in a draft research paper last month that building AI infrastructure would cost more than $10 trillion through 2032. Van Nieuwerburgh estimated that average annual AI spending, relative to U.S. economic output, exceeds the combined national commitments to household electrification, interstate highways and 1990s internet infrastructure. Van Nieuwerburgh and other experts say AI companies are searching across financial markets to meet those costs.

The U.S. government also faces higher borrowing costs because AI companies are competing for investors' money. Treasury Secretary Scott Bessent described short-term competition for capital in a CNBC interview in August. Investors are becoming more selective about financing AI companies. Stock and bond prices have fluctuated sharply in recent weeks, particularly among neocloud companies that rent data center computing capacity to leading AI developers. Investors have rejected several neocloud IPOs or required higher interest rates on loans.

Many neocloud companies depend on debt or other external funding to equip expensive data centers before receiving the hoped-for rental revenue. Gartner identified the mismatch between immediate spending and later revenue. Gartner predicted that 80% of neocloud companies could fail or be forced into sales in coming years.

Moody's chief credit officer Atsi Sheth said greater investor selectivity is healthy and reduces the likelihood that AI investment becomes a bubble. Sheth warned that systemic problems arise when investors mistakenly believe risky investments are safe. Blackstone's Jas Khaira said last month that the source of all the financing AI companies require remains uncertain.

Microsoft's revenue and profits are growing through sales of AI software to businesses seeking productivity gains. Microsoft nevertheless announced voluntary buyout offers on Monday for about 60,000 employees, one-quarter of the global workforce, while AI spending substantially exceeds the current financial returns. A Microsoft spokeswoman said only a fraction of eligible employees are expected to accept the offers.

AI has increased Meta's advertising sales across the company's social networks. However, Meta is beginning to experience cash outflows because of AI spending. Meta acknowledged in a securities filing last week that the company needed to secure AI financing before investors exhausted available funds. Meta's filing said funding demands from Google, SpaceX, Anthropic and OpenAI had made raising capital increasingly difficult.

The global economy and individual finances are increasingly exposed to the outcome and rising costs of AI investment. Businesses across the AI industry, Wall Street, office leasing and paint sales increasingly depend on AI spending. That dependence rests on expectations that AI will transform work and economic life while generating growth, corporate profits and wealth for decades.

Stock markets advanced on Monday. Orton and some other investors said Anthropic's disclosure of rapidly growing revenue from a relatively broad customer base eased concerns about an AI bubble. Orton expressed optimism about AI's eventual financial returns. Orton described Anthropic's IPO as a test of investors' willingness to tolerate uncertainty about when those returns will arrive. Orton said substantial gains could be available broadly, but markets question how quickly those gains will materialize.