AI bubble 2025: Analysts warn AI hype is a ‘Red Flag’ — bubble now bigger than 2008 Subprime crisis
The fight between Wall Street and Silicon Valley continues on many fronts. Tech firms want to run financial services. Private equity and crypto compete for retirement accounts. Some political groups want tech giants to replace major banks in shaping business rules. These disputes are ongoing. But the two sides agree on one thing. They are building risky credit structures that look ready to trigger a major financial breakdown.
Last month, attention centered on “round-tripping.” Major players in artificial intelligence funded their customers so those same customers could buy their products. Nvidia invested in AI companies that then bought Nvidia processors. This was only one piece of a much larger shift. Former “shadow banks” have adopted a new label. They now call themselves private credit firms. The purpose appears plain. They want to keep lending billions without the regulation that applies to traditional banks.
Private credit has become the main engine of AI infrastructure finance. Assets under management reached $1.6 trillion in February and have grown since. Many deals look detached from reality. The life span of the items being funded does not match the length of the loans. Experts warn that the country is building another bubble. Some speak openly about a coming financial crisis. The only question is the size.
The math explains the danger. AI computing and data center growth need roughly $2 trillion in revenue each year by the end of the decade to justify current investment. That level of income is not visible today. No company has it. Many companies may never have it. So investors rely on creative debt structures to bridge the gap. Big Tech generates huge cash flow but not enough for the required spending. Venture capital is cautious about funding infrastructure. Data centers and power plants do not promise explosive returns on their own.
At the same time, investors crave access to the AI boom. They believe large tech companies will drive U.S. industrial growth for many years. Big Tech historically borrowed very little. Even with increased bond issuance this year, major firms want to protect high credit ratings. Smaller AI businesses cannot borrow the hundreds of billions needed.
This gap gives rise to special purpose vehicles, known as SPVs. For example, a new firm appears to build a data center. A major tech company signs on as an anchor tenant. The SPV finances the construction by selling debt. Investors buy the debt under the belief that rental payments from the tenant will cover repayments.
One example is the $30 billion Hyperion data center in Louisiana. A private credit fund called Blue Owl holds a majority stake in the SPV. It put in limited equity while the SPV issued debt. Meta holds the remaining 20 percent. Investors feel confident because Meta pays rent but does not have to record the debt on its balance sheet. Rating agencies then give these instruments strong grades. Many experts question these ratings, calling them suspicious.
Private credit funds like Blue Owl gather billions from investors. The Wall Street Journal reports more than $1 trillion flowing into the sector. Blue Owl alone manages $295 billion. Fund managers sell these deals as near-risk-free. They say big tenants are locked into long leases, so the repayment stream is secure.
But cracks have appeared. After merging two of its credit funds, Blue Owl blocked investor withdrawals. That action forces investors to take losses around 20 percent. If a traditional bank did the same, it would resemble a bank run. The move has angered investors who have limited rights within private credit structures. Signs of trouble are visible across the industry. Debt contracts are being rewritten to protect private credit funds if things collapse.
Data centers are unreliable as long-term collateral. The internal hardware does not last. GPUs used to train AI models may burn out within two years. High-end AI companies want the latest processors as soon as they become available. Some lenders, including Blue Owl, are even financing purchases of Nvidia GPUs for Elon Musk’s xAI.
AI firms stretch depreciation schedules to claim the equipment will last far longer than expected. That results in unreal revenue and asset values. Some firms borrow against their existing GPUs to buy new ones. This inflates the appearance of growth while masking growing debt. Meanwhile, many data centers and attached power plants may become stranded assets once the hardware inside becomes obsolete.
The securitization pattern deepens the risk. Many data center loans roll into asset-backed securities. Investors choose risk levels through debt tranches. A report from the Center for Public Enterprise shows that more than 60 percent of relevant securitizations in this market come from data centers. These securities often mature long after the hardware becomes outdated.
OpenAI lost more than $11.5 billion last quarter. It plans to spend much more in coming years. There is no clear revenue stream to plug that gap. If more efficient Chinese AI models beat U.S. models, finance for U.S. expansion could fall apart. The companies driving model development also own the cloud platforms selling compute time. Revenue from cloud services is subsidizing model development. Each company funds the expansion of the others. That is a classic bubble pattern.
A growing group of real estate firms, sometimes called “neoclouds,” now build centers and lease space. CoreWeave is one of the major names in this category and carries heavy debt. The scale of the cycle resembles previous periods of U.S. financial excess. The structure mixes private lending from the 1920s, technology build-out comparable to the railroad boom, and complex financial instruments reminiscent of the mid-2000s housing crisis.
Banks are now entangled. Moody’s reports more than $300 billion in bank loans tied to private credit. Blue Owl’s Hyperion bonds are sold by firms such as BlackRock and PIMCO. Wall Street wants exposure because AI growth is the only strong sector in a weak economy. Some banks, including Deutsche Bank, appear cautious but still allow traders to finance data center debt.
If trouble spreads, private credit funds might struggle to get government support. Banks, in contrast, have clearer pathways to rescue programs. As banks absorb more private credit exposure, the burden may shift to the public.
Special purpose vehicles may lock in the losses while major firms trade capital back and forth. Retail investors, including retirement savers, could face large losses. Some influential figures, such as Peter Thiel, appear to be stepping away quietly. Blue Owl stock is falling. The company dropped 6 percent on Monday alone. Even Google’s chief executive says no company will escape the fallout if the AI bubble bursts.
The warning lights are on across the market. The country is loaded with debt built on assets that fade faster than the loans funding them. The result may be a financial shock with wide impact across banks, investors, and workers who never signed up for the risks.
#AIBubble #PrivateCredit #BigTechFinance #DataCenters #MarketRisk