A 5.5% weekly drop in the Nasdaq-100 through Oct. 16 has sharpened attention on AI infrastructure debt, while earlier memos from Howard Marks, co-founder and co-chair of Oaktree Capital Management, detail financing and return risks.
The U.S. Bureau of Labor Statistics (BLS) said consumer prices rose 3.7% over the 12 months through September, compared with 3.4% in August. The index for final demand rose 1.3% from August to September, a monthly producer-price change.
In a Dec. 9, 2025, memo, Marks flagged debt-funded expansion and circular financing among AI companies. He questioned whether chips and data centers could generate enough earnings to support the borrowing over time. He warned that debt magnifies the risk if capacity is built beyond customer needs. The memo also listed high valuations and fear of missing out among potential bubble signals. Marks said the evidence did not establish that an AI bubble had formed.
In a Feb. 26, 2026, memo, Marks said investors still need to determine how much AI-related revenue ultimately comes from end users.
In his December memo, Marks said established companies active in AI already had revenue, profits and cash flow, unlike many firms during the internet boom. Marks wrote in December 2025 that the price-to-earnings multiples of Nvidia, Microsoft, Alphabet, Amazon and Meta were below those Microsoft, Cisco and Oracle reached during the internet boom.
Marks cited the late-1990s telecom buildout as an example of overbuilding and circular financing. He has also written that technology booms can leave lasting infrastructure even when investors lose money.
Marks estimated it may take about 10 years to determine whether AI infrastructure earns returns that justify its cost, as an earlier EconoTimes report on his 10-year assessment horizon noted. He has treated AI adoption and the economics of its infrastructure as separate questions. End-user revenue and adequate returns on infrastructure spending are measures he identified for assessing the investment case.
