Credit feeds the Boom of the AI ​​- and the fears of a bubble

JPMorgan Chase & Co. and Mitsubishi UFJ Financial Group direct the sale of a loan of more than $ 22 billion to support the plan of Vantage data centers to build a massive campus of data centers, people who are aware of the case said this week. Meta Platforms Inc., Facebook’s parent, obtains $ 29 billion in Pacific Investment Management Co. and Blue Owl Capital Inc. for a massive data center in Rural Louisiana, Bloomberg reported this month.

And many more of these offers arrive. Openai alone believes that he will need billions of dollars over time to spend on the infrastructure required to develop and manage artificial intelligence services.

At the same time, the main players in the industry recognize that there is probably the pain to come for AI investors. The CEO of Openai, Sam Altman, said this week that he saw parallels between the current frenzy of investment in artificial intelligence and the Dot-Com bubble in the late 1990s. Discussing the assessments of the startups, he said: “Someone will be burned.” And an initiative of the Massachusetts Institute of Technology published a report indicating that 95% of generative AI projects in the business world have not generated any profit.

In total, it is enough to make the observers of nervous credit.

“It is natural that credit investors rethinned in the early 2000s when telecommunications companies undoubtedly surfed and excessive and we saw significant withdrawals from these assets,” said Daniel Sorid, head of the American investment class credit strategy at Citigroup. “Thus, the AI ​​boom certainly raises medium -term questions around sustainability.”

The early construction of the infrastructure necessary to train and feed the most advanced AI models has been largely funded by AI companies themselves, including technology giants like Google and Meta Platforms Inc. Alphabet Inc.

The exhibition here presents itself in many forms and sizes, with various degrees of risk. Many large technological companies – IA Hyperscalers – have paid new infrastructure with gold -plated business debt, which is probably sure due to existing cash flows that guarantee debt, according to a recent analysis by Bloomberg Intelligence.

Much of the debt financing now comes from private credit markets.

“Private credit financing of artificial intelligence takes place at around $ 50 billion per quarter, in the low -end, in the last three quarters. Even without taking into account the mega offers from Meta and Vantage, they already provide two to three times what public procurement provides, “said Matthew Mish, head of credit strategy at UBS.

And many new IT centers are funded through titles backed by mortgage claims, linked not to a business entity, but to payments generated by complexes. The amount of CMB supported by AI infrastructure is already up 30%, to $ 15.6 billion, compared to the total of the year entirely in 2024, JPMorgan Chase & Co. estimated this month.

Sorid and a colleague from Citi published a report on August 8 by focusing on specific risks for public service companies that stimulated the loan to build the electrical infrastructure necessary to supply swallowed data centers. They and other analysts share a common concern to spend so much money at the moment, before AI projects show their ability to generate long -term income.

“Data center transactions are tenor funds from 20 to 30 years old for a technology that we don’t even know what they will look like in five years,” said Ruth Yang, a world leader in private market analysis at S&P Global Ratings. “We are conservative in our assessment of cash flow flows in the long term because we do not know what they will look like, there is no historical basis.”

Stress has started to appear in the rise of payment loans in kind to private credit attached to technology, noted UBS Group. In the second quarter, PIK income in BDC has reached the highest level since 2020, climbing to 6%, according to UBS.

But it is unlikely that the fire hose does not stop anytime soon.

“Direct lenders are constantly collecting capital, and he must go somewhere,” said John Medina, vice-president of the World Moody World Project and the infrastructure financing team. “They see these hyperscalers, with this need for massive capital, as the next long -term infrastructure asset.”