The AI Infrastructure Frenzy – Letting Wall Street "not rest even if it's fake" by financing technology that is unknown five years from now for a term of 20-30 years

An unprecedented AI infrastructure financing frenzy is sweeping Wall Street, with hundreds of billions of dollars flowing into data center construction, leaving bankers no time to take a break even during the August holiday. Meanwhile, industry executives and analysts are beginning to question whether this investment boom is creating a new bubble, especially as investors provide financing for a technology whose shape remains uncertain five years down the line.

On August 23, it was reported that informed sources revealed JP Morgan and Mitsubishi UFJ Financial Group are leading a loan of over $22 billion this week to support Vantage Data Centers in building large data center campuses. Meta has secured $29 billion in funding from Pacific Investment Management Company and Blue Owl Capital to construct large data centers in rural Louisiana. This deal highlights the market's fervent enthusiasm for AI infrastructure financing.

However, behind the frenzy of capital inflow, doubts are starting to emerge. Key industry participants acknowledge that AI investors may face pain. OpenAI CEO Sam Altman stated that he sees similarities between the current AI investment frenzy and the internet bubble of the late 1990s. A study from the Massachusetts Institute of Technology shows that 95% of enterprise generative AI projects fail to generate any profit.

Analysts point out that this contrast is making credit watchers nervous, especially considering that many financing arrangements are based on forecasts of future cash flows from data centers, while the long-term profitability of these technologies remains uncertain. Citigroup's head of U.S. investment-grade credit strategy, Daniel Sorid, stated:

"Credit investors naturally recall the early 2000s when telecom companies may have overbuilt and over-leveraged, only to see significant write-downs on those assets."

AI Data Center Financing Reaches Historic Highs

For Wall Street bankers engaged in data center financing deals, this summer has been devoid of holidays.

Reports indicate that in July, there was a $10 billion debt and equity deal for xAI, as well as a $2.6 billion debt deal for CoreWeave.

In August, at the beginning of the month, Meta secured a $26 billion loan and a $3 billion equity deal for its data center construction. This week, JP Morgan and Mitsubishi UFJ Financial Group agreed to underwrite $22 billion in debt for Vantage Data Centers.

According to industry tracking agency Project Finance News, the market is expected to grow to $60 billion this year, doubling that of 2024. Given the total transaction volume in August, the market may need to revise its estimates for data center financing upwards.

UBS's head of credit strategy, Matthew Mish, stated:

"Private credit funding for AI has been at the low end of about $50 billion per quarter over the past three quarters. Even without considering the large deals from Meta and Vantage, the funding they provide is already two to three times that of the public market." Currently, most of the debt funding comes from the private credit market. Many new computing centers are obtaining financing through commercial mortgage-backed securities (CMBS), which are not linked to corporate entities but rather to the revenues generated by these complexes.

According to JP Morgan's estimates, the amount of CMBS supported by AI infrastructure has increased by 30% from the total for the entire year of 2024, reaching $15.6 billion.

Transition from Self-Funding to External Financing

It is noteworthy that the early construction of the infrastructure required to train and support advanced AI models was primarily funded by the AI companies themselves, including tech giants like Google and Meta. However, recently, funding has increasingly come from bond investors and private credit institutions.

A recent report from Bloomberg Intelligence pointed out that although the risks associated with AI-related investments vary widely, "AI super giants" like Microsoft and Amazon primarily finance the construction of new infrastructure by issuing high-quality bonds known as " gold-plated corporate bonds." These bonds are considered relatively safe because these companies have ample cash flow and strong repayment capabilities.

Borrowers are typically tech giants with some of the best balance sheets in the world. They choose to borrow because the amounts are substantial, and the debt is usually tied to the data centers they are building rather than the companies themselves.

The boom in private debt funds means more capital is seeking higher returns. Data center transactions meet this demand, offering higher yields than typical corporate loans. As a result, investors are seizing the opportunity to earn extra cash.

Analysts also noted that as Federal Reserve Chairman Jerome Powell becomes more open to interest rate cuts, the pursuit of yields will become more urgent.

Industry Key Players Acknowledge AI Investors May Face Pain

However, as the scale of financing for AI data centers is experiencing explosive growth, key industry players acknowledge that AI investors may face pain.

According to a Jianwen article, Sam Altman, CEO of OpenAI, the "flag bearer" of the global AI wave, believes that the current enthusiasm in the AI field has many similarities to the dot-com bubble of the past. He also pointed out that OpenAI will spend trillions of dollars on data center construction in the near future.

In this trillion-dollar gamble, there are bound to be losers. Altman ominously stated that "someone" will lose "an astonishing amount of money." But he quickly added, "We don't know who."

This week, a report from the Massachusetts Institute of Technology (MIT) revealed the harsh reality of corporate AI investments, stating that while companies are full of expectations for generative AI, the vast majority of projects have failed to produce actual financial impacts, with as much as "95% of companies receiving zero returns from their generative AI investments." On one side is the frenzied financing of AI infrastructure, and on the other side are heavy warnings from key industry participants, creating tension among credit observers.

Ruth Yang, Global Head of Private Market Analysis at S&P Global Ratings, candidly pointed out the risks of this financing model:

" Data center transactions are financed over 20 to 30 years for a technology that we don't even know what it will look like in five years. We take a conservative view on future cash flows because we don't know what they will look like, and there is no historical basis."

According to UBS Group, a type of loan called "PIK (Payment-in-Kind) loans" is becoming increasingly common in the tech private credit space, where borrowers, facing cash constraints, use increased debt to pay interest instead of cash. In the second quarter, this "paper interest income" accounted for 6% of total revenue among BDCs investing in small and medium-sized enterprises, the highest level since 2020, indicating rising financial pressure on lenders.

BDC (Business Development Company) is an investment company that specializes in investing in small and medium-sized enterprises or startups, providing investors with access to the private market. The rising proportion of PIK in BDC revenues means that an increasing amount of interest they collect is "paper interest" (non-cash), reflecting the growing financial pressure on borrowing companies.

Notably, market data shows signs of a bubble. According to CB Insights, there are currently 498 AI unicorns with a total valuation of $2.7 trillion. The valuation multiples for AI startups have exceeded 100 times.

Even more concerning is the unit economics of AI startups: Users pay $1, while application layer companies pay $5 to foundational model providers, who then pay $7 to hyperscale computing service providers, ultimately paying $13 to GPU manufacturers.

The well-known financial blog ZeroHedge cautioned in an article to be wary of the "discontinuity period" during the AI technology hype cycle, emphasizing that investors must clearly understand which stage each AI technology they invest in or deploy is currently at in the emerging technology cycle proposed by Gartner. Gartner's technology hype cycle model divides the development of emerging technologies into five stages:

Innovation Trigger, Peak of Inflated Expectations, Trough of Disillusionment, Slope of Enlightenment, Plateau of Productivity

Potential Ending Factors: Power Costs, Price Pressure

Many AI debt transactions are based on the idea that their constructed data centers will generate enough revenue to repay the loans. The cost competition among top AI providers will be closely monitored. If AI prices drop again, it may raise concerns among some investors about repayment.

The state of power demand may also become the end of the lending frenzy. Data centers consume a large amount of electricity, and electricity prices have risen nearly 7% this year. Regulators attribute the soaring demand to data centers. Rising electricity prices at least mean higher costs for operating data centers.

Texas has been particularly dissatisfied with the price increases, as the state has passed a law granting grid operators the ability to reduce power supply to data centers during crises.

The stock market is also beginning to show skepticism. CoreWeave, a typical representative of AI risks and returns, had a striking IPO earlier this year, but the company's stock price has fallen nearly 50% from its peak.