Private Credit, Public Risk: AI Is Eating The Debt Market

After a post-pandemic recovery that heavily leaned on consumer spending, the last two quarters have seen a flip. Corporate spending is now leading the charge, owing to relentless investments in artificial intelligence.

According to Goldman Sachs, global power demand from data centers is expected to increase by 50% by 2027, driving significant spending and potentially formalizing AI-dedicated data centers as a new class of infrastructure.

Datacenter Arms Race

Microsoft (NASDAQ:MSFT), Alphabet (NASDAQ:GOOG), Amazon (NASDAQ:AMZN), and Meta (NASDAQ:META) alone are expected to shell out over $200 billion in combined capital expenditures this year. Much of it is going into AI-related investments like data centers, custom silicon, and software infrastructure. This isn’t incremental; it’s a macro force. As JPMorgan noted, "The Fifth Industrial Revolution will be driven by the continued evolution of AI."

Meanwhile, traditional consumer spending has begun to plateau. Durable goods consumption is down, and excess savings from pandemic stimulus have been largely depleted. That change means the U.S. economy's recent resilience isn't coming from your average Joe buying TVs—it's from mega-corporations buying teraflops.

Employment in data center construction is surging, and suppliers, ranging from chipmakers to power utilities, are scrambling to keep up. Even GDP numbers are feeling the impact. Gross domestic product jumped 3% for the second quarter, far above the 2.3% estimate.

A Trillion Dollar Corporate Club

Still, this gold rush is exclusive to a very select club. Building the next generation of AI requires enormous upfront investment—think custom chips, water-cooled hyperscale data centers, and software stacks that can handle billions of tokens. For most companies, that kind of money simply doesn't exist.

The top five U.S. tech giants account for over 80% of all AI-related capex.

Meta is spending tens of billions to retool its entire infrastructure around AI, ditching GPU leasing in favor of in-house clusters. Amazon is rolling out its own Trainium chips to cut Nvidia out of the loop. Microsoft just pledged another $10 billion to expand Azure's AI footprint, while Apple (NASDAQ:AAPL), traditionally slow to spend, is quietly buying up bandwidth and training data to launch its generative ecosystem.

The barrier to entry is clear. If you're not already big, rich, and entrenched, good luck playing catch-up. But for the few who can afford it, the rewards are massive. Cloud AI platforms are expected to be the next great tollbooth of the internet, charging rent on every algorithmic transaction.

Private Credit Joins the Party, Raises System Risk

The shift in spending also brought forth a change in crediting. Private credit (PC), an opaque part of Wall Street mostly known for distressed loans and leveraged buyouts, is now an essential source of capital for the AI boom. And it's not just startups—it's growth-stage companies and even industry leaders tapping this pool.

At the end of the last quarter, Meta was looking to raise as much as $29 billion for its AI expansion. According to the Financial Times, that would be one of the largest private fundraising efforts of this kind.

The data from the Federal Reserve Bank of Boston shows PC financing grew from a mere $46 billion in 2000 to roughly $1 trillion in 2023, inflation-adjusted. The Bank identified this growth as largely funded from bank loans to credit lenders, outlining increased risks.

The problem arises when growth occurs through expansion (providing loans that banks wouldn’t make) rather than substitution (competing with banks for their loans). However, when credit lenders borrow from banks themselves, their spreads are wider, making it highly unlikely to compete with the banks.

Boston FED also notes distinctive differences between banks and PC funds. The latter are typically less leveraged and less vulnerable to capital runs due to the locked-up structure of limited partnerships (often for multiple years). Meanwhile, they note that 75% of bank funding consists of demand deposits, almost half of which are uninsured.

Thus, when PC lenders make the loans that the banks themselves wouldn’t make, aggregate credit risks in the financial system rise.

This influx of non-bank lending is reshaping capital flows. Instead of waiting for IPOs or venture equity rounds, firms are turning to direct lenders for faster cash. The leverage ends up on the balance sheet of a riskier borrower, and all is well until enough voices say, "I want my money back."

Read Next:

• ‘This Is Our Time to Do More’: SoundHound CEO On AI’s Enterprise Breakout

Photo: Who is Danny/Shutterstock