There is no clear answer to when, or whether, the AI market will suffer a sharp downturn. But comments from executives at major investment firms highlight growing concern about the risks behind the industry’s rapid expansion and its reliance on debt.
At the 2026 Global Financial Leaders’ Investment Summit in Hong Kong, PIMCO CEO Emmanuel Roman warned that the AI market could experience a downturn followed by further damage. He also pointed to uncertainty about how the sector’s pricing and economics will develop.
The cost of building AI infrastructure has put pressure on debt markets, while major technology companies continue to seek financing. Meta’s recent bond offering drew $58 billion in orders against an expected size of $15 billion to $20 billion, according to Bloomberg. Oracle said it would need to issue another $20 billion in bonds, Alphabet is selling €6 billion, and Microsoft recently raised $25 billion through its first bond sale in a decade.
Roman estimated that AI infrastructure could require $2 trillion in financing over the next five years, compared with an overall US bond market of about $12 trillion. PIMCO has also been involved in financing the AI sector, including an early chip-backed deal for CoreWeave and purchases of Meta and SpaceX debt.
Apollo Global Management president Jim Zelter also urged caution. He expects demand for computing power to grow and said AI is improving quickly, but stressed that the sector’s future pricing and economics remain uncertain.
Investor appetite has not disappeared: demand for Meta’s bonds was strong, and Wall Street banks have also arranged $63 billion in debt financing for Anthropic. At the same time, the source reports that China’s AI market has already been described as having experienced a bubble burst.
Roman identified US interest-rate increases, war and a eurozone crisis as possible triggers for a market shift. The competing signals leave the timing and scale of any downturn uncertain, even as more investors voice concerns about the cost and sustainability of the AI buildout.
Sources: The Verge.