As U.S. Treasury yields rose to their highest level since 2007 this week, companies that rely on debt financing are facing higher borrowing costs. This means that infrastructure construction, which has already reached a historical high of AI, will become even more expensive.

In June, JPMorgan Chase estimated that by 2030, the issuance of AI related debt would reach $4.1 trillion. Data center companies and other enterprises related to the artificial intelligence boom are competing to expand their capacity to meet what many industry experts consider to be an almost endless demand for AI services.

As borrowers return to the market, they now face a yield on 10-year U.S. Treasury bonds of nearly 5.17%, which is about 1 percentage point higher than at the beginning of the year. This means that companies issuing debt will have to offer more attractive returns in order to attract investors.

At least for now, the market has not fallen into panic. The highly indebted neocloud company CoreWeave has seen a stable stock price this week, rising by nearly 8%; whereas in AI, the oracle bone text that relies on the debt market for expansion fell by 7% this week and has already dropped by about 30% since the beginning of this year.

Meanwhile, SoftBank of Japan, as the main capital provider for the AI project, completed the issuance of $11.1 billion in junk bonds this week, with the yield on the 7-year bonds reaching as high as 9.75%.

Siebert Financial Chief Investment Officer Mark Malek said in an interview: "They are basically insensitive to the price of this financing, which means they are price takers. In my opinion, many companies of this kind need to be insensitive to price. They need to raise as much capital as possible to compete."

The core of this craze is the leading model developers OpenAI and Anthropic. Both companies are valued at nearly $1 trillion in the private market. To provide the necessary infrastructure for their advanced models, as well as those of numerous other companies and services, tech giants such as Amazon, Google, Meta, and Microsoft have committed to investing tens of billions of dollars in capital expenditures this year, with further increases expected in 2027.

Although a considerable portion of their funding comes from debt financing, these tech giants possess investment-grade credit ratings, which allows them to obtain capital at lower costs. However, some market participants believe that greater challenges lie ahead for other companies.

Risk signals?

A senior private credit investor, who wished to remain anonymous, told CNBC that in the future, it will be more difficult to finance neocloud transactions, as these companies have less buffer space and are unable to absorb the pressure brought about by rising costs.

Mitsubishi HC Vice President of Capital America, Riley Thompson, said in an interview that lending institutions have become more selective about the projects they are willing to fund, even if borrowers are willing to pay higher interest rates.

Thompson said, "Rather than saying that the market is really interested in 50 neocloud companies, it's more accurate to say that there are probably only about 20."

CoreWeave, which went public last year, warned about the risk of rising interest rates in documents submitted to the U.S. Securities and Exchange Commission. In its latest quarterly report, the company stated that as of June, based on its outstanding floating-rate debt balance, for every 100 basis points increase in interest rates (i.e., 1 percentage point), its interest expenses could increase by $30 million.

This week, Oracle's stock price fell, which may indicate an early warning sign. Previously, Bloomberg reported that the company issued a "force majeure" notice regarding its data center project in New Mexico to protect itself from higher costs. The report stated that if the facility fails to go into operation as expected by 2028, Oracle is seeking to defer payments for this project, which is referred to as Project Jupiter. Oracle stated that the project "is still progressing according to our planned timeline."

Rising interest rates are not the only issue. Before this week's surge in yields, the CEOs of Anthropic and OpenAI had already begun calling for a slowdown in the development pace of AI, as industry researchers publicly expressed concerns that advanced models might get out of human control.

Meanwhile, the backlash against the AI data centers across the country has become an important issue ahead of the mid-term elections in November. The latest survey conducted jointly by NBC News Decision Desk Poll and SurveyMonkey shows that 69% of respondents oppose the construction of such facilities in their regions. On Monday, Greg Abbott, the Republican governor of Texas who is in a tough re-election campaign, ordered a temporary suspension of all environmental permits related to data centers, following his previous suspension of power grid approvals last month.

Nevertheless, the market demand for AI services continues to grow explosively. The latest example is Meta's Muse personal assistant app, which has become very popular since its launch at the beginning of September. Muse achieved over 2.5 million global downloads in the two weeks before its release, surpassing ChatGPT to top the Apple App Store charts. Evercore's Mark Mahaney indicated this week that the app could reach 100 million users within 6 to 12 months.

Credit rating agency KBRA, the global head of corporate, project, and infrastructure financing at Andrew Giudici, stated that although rising interest rates may affect future transactions, he does not believe that this will have a significant impact on borrowing demand.

Giudici said, "In normal circumstances, people might take a step back and pause for a moment. But I don't think that will happen here. I believe you will continue to see relatively large-scale releases."

Latham and Watkins, Vice Chairpersons of Emerging Companies and Growth Businesses, Haim Zaltzman stated that, without a doubt, as costs rise, "someone has to bear the burden."

But he said, "Under that kind of demand structure, it would be much easier to absorb these costs because the demand is simply too large." Zaltzman is in charge of the infrastructure financing business for AI.

For Bernie Margulies, the Chief Executive Officer of American Compute, who provided risk management consulting for the financing of GPU, the situation is even simpler. He stated that even in the face of higher costs, borrowers are eager to obtain financing, especially when they have cooperation agreements with OpenAI and Anthropic – these two companies have already signed contracts to lock in computing power capacity several years in advance.

Margulies said, "If you have a transaction with Anthropic, will 50 basis points really stop you?"