The AI infra trade cracks, the 5-year hold dies

The AI infra trade cracks, the 5-year hold dies

The Carry · October 11th. 6 min 0 sec.

Cold open

Private credit is falling over itself to fund AI data centers. But an 85 billion dollar real estate debt platform is walking away. Why? Zero alternative use value. If the single tenant defaults, the lender is left with an empty concrete shell. That's the signal. The hottest trade in infrastructure is cracking under the weight of physical reality.

Intro

This is The Carry for October 11th. We listened to dozens of hours of the top private equity podcasts this week so you don't have to. Here's what actually mattered across the market. Let's get into it.

The AI infra trade cracks

The AI data center trade is showing serious cracks. A massive wave of private credit is chasing hyperscale yields right now. But Invesco is pulling the plug. On Capital Allocators, Charlie Rose laid out why his 85 billion dollar real estate debt platform absolutely refuses to originate loans on data centers. His thesis is simple. The underlying collateral suffers from rapid technological obsolescence. If the single tenant defaults, the asset has zero alternative use value. Equity sponsors are hitting the exact same physical limits. Over on The Infrastructure Investor Podcast, Marion Calcine revealed that Ardian actually slashed the expansion pipeline for its Verne data center platform by 30 percent. Her team audited the local Icelandic electrical grid. What they found directly contradicted the seller's aggressive tenant demand projections. Transmission bottlenecks are real. Compute can't just scale infinitely. The broader municipal infrastructure required to support these facilities is also fracturing. Chris Beall brought this up on Dry Powder. He noted that the massive capital pouring into AI data center capex is blinding allocators to underlying physical decay. Computing capacity requires basic municipal resources. Beall points out that severe groundwater depletion and regional flight disruptions show acute physical bottlenecks. Digital capex can't solve physical problems. That dynamic completely breaks aggressive underwriting models. The market is pricing in limitless expansion. But the grid, the water supply, and the debt markets are saying something else entirely.

The five-year hold dies

The standard five-year exit math is completely dead. Investment committees aren't defaulting to those clean, five-year underwriting models anymore. On the Private Equity Funcast, Accordion CFO Jon Apter detailed how the historical private equity hold period has permanently fractured. Sponsors are now forcing one of two extremes. They execute rapid two-to-three-year flips. Or they stretch capital into seven-to-eight-year continuation vehicles. When those longer holds try to find liquidity, broad enterprise auctions are stalling out on price. Cooley partner Danielle Fortier explained this dynamic on M&A Science. Buyers are bypassing full buyouts right now. Instead, they target bilateral corporate carve-outs. Sellers are increasingly willing to accept these partial exits. It generates liquidity for core operations without taking portfolio-wide valuation discounts in a tight market. These extended hold periods are breaking the operational benches that sponsors rely on to actually sell the business. Over on Private Equity Spotlight, Ted Bililies argued that static management equity plans signed at acquisition lose their alignment as the hold stretches out. Second-tier commercial leaders just quit during buyer diligence. Bililies insists that sponsors are being forced to reset payout structures 12 to 18 months before a sale just to keep the team intact. The entire timeline has shifted. The playbook from 2019 just falls apart when the capital gets trapped.

Charlie Rose

Charlie Rose brought a rare flash of discipline to a private credit market that's totally desperate for AI exposure. Speaking on Capital Allocators, he laid out exactly where his firm draws the line on infrastructure risk. He told Ted Seides, I'll share with you a controversial position that we've taken, we're not lending on data centers. That's a massive signal from an 85 billion dollar platform. LPs are paying close attention to who actually prices risk and who just chases the trend.

Sam Lissner

Sam Lissner mapped out the mid-market infrastructure playbook for surviving an era of permanently elevated interest rates. Appearing on The Infrastructure Investor Podcast, he explained how smaller funds are delivering major outperformance by avoiding the highly competitive mega-fund auctions entirely. He summed up the capital structure strategy perfectly, saying, our philosophy is that we view use as an enhancer of returns, but never the driver of returns. That captures the exact shift happening across investment committees right now. True operational turnaround is the only reliable lever left to hit target hurdles.

Steven Rosenblatt

Steven Rosenblatt pushed back hard on the idea that the current AI capital expenditure cycle is going to end in an immediate 1999-style operational collapse. On How I Invest, he recalled the intense operational pressure of pre-selling 60 million dollars in mobile ads for an Apple product that didn't even exist yet. He explained the difference between then and now, saying, during the dot-com boom, the reality is that you didn't have balance sheets, everything was funded on hype and vaporware, but today, you have fortress balance sheets. Buyers are pricing this in. The biggest tech players have the cash reserves to absorb these massive infrastructure bets without snapping.

The bottom line

Mega-funds are piling blind capital into AI compute and data centers right now. But the sharpest operators are recognizing that physical grid limits and 90-day software lifecycles are quietly breaking the old underwriting models. The physical world is pushing back on infinite digital scale. The fundamental timeline to build, hold, and sell a business is permanently rewired. That's The Carry. See you next Sunday.