Our Take
Professional investors aren't equipped to spot fraud in fields they don't understand, and that matters more when the loans are this large and the revenue models this fragile.
Why it matters
Banks and private credit funds have committed $1.5 trillion (per Morgan Stanley estimates) to AI data center buildouts. If the AI industry is overleveraged and revenue collapses when usage-based billing takes hold, the lending standards now in place are inadequate to absorb that shock.
Do this week
Finance teams: audit your AI consumption costs under token-based billing before month-end, because hidden subsidies are ending and the revenue cliff is real.
A private credit fund lost money on an obvious fraud
Lawyers for the bankrupt neobank Aspiration Partners are being sued by former investors from a UBS private credit fund. The investors claim the lawyers actively participated in the fraud. Aspiration promised to be a neobank that would plant millions of trees to offset carbon emissions, and counted Leonardo DiCaprio, Robert Downey Jr., and Los Angeles Clippers owner Steve Ballmer as investors.
On its face, the fraud was not sophisticated. A neobank is a risky investment. Carbon offset schemes are riskier. Together, they are substantially risky. Adding celebrity endorsements as a selling point should have triggered skepticism, not confidence. CEO Joseph Sanberg was convicted; the fund lost money anyway.
The U.S. Attorney's office defended the outcome by noting that "anyone can get duped by a con man." True. But also: a professional private credit fund should have the expertise to smell an obvious, multi-layered risk. If they didn't, what does that say about the standards being applied to other, more complex bets?
Private credit is financing AI, and AI's revenue base may be an illusion
The Aspiration case matters because it raises a specific question about the loans banks and private credit funds have extended to the AI industry. Morgan Stanley estimates loans will cover $1.5 trillion of AI data center spending over the next few years. That capital depends on AI companies generating enough revenue to service the debt.
But AI companies may have been subsidizing user costs for years without realizing it. Before moving to token-based billing, Anthropic and OpenAI charged per user, not per usage. One AI consultant reported to Axios that a client spent $500 million on Claude usage in a single month under the new token model. Under the old per-user model, that client's actual costs were invisible because power users faced no incentive to limit usage.
What happens when companies start paying by usage and pull back? The AI industry cannot afford to lose revenue when it has borrowed more than $1 trillion to build data centers and needs to spend billions more on chips to stay competitive. If AI is in a bubble, this revenue cliff is the mechanism that pops it. Private credit funds are heavily invested in AI. Banks are increasingly invested in private credit. That chain of exposure is what worries observers now.
The Aspiration case is instructive because it shows that even sophisticated investors struggle to apply standards to bets they don't fully understand. A neobank plus trees plus celebrity is obviously risky. AI data center loans backed by uncertain, newly-transparent usage costs are less obviously risky, but the structural problem is the same: the revenue model has not been tested at scale.
Audit your AI costs and demand visibility into the lender's assumptions
If your organization has been running AI workloads on per-user pricing, your true cost is unknown. Under token-based billing, that cost is now visible and likely far higher than budgets assumed. Before month-end, calculate your actual usage costs and model what happens if consumption drops 20, 30, or 50 percent as usage-based pricing sinks in across the industry.
If you work in finance or investment: ask your AI and infrastructure lenders what revenue assumptions underpin their loan books. If the answer is "customers won't pull back on usage because the AI is valuable," ask for the evidence. If the answer is "we're not sure," that is the right answer. Aspiration taught us that a fund missing obvious risk is a problem. But missing hidden risk is worse.