PIMCO finds the credit market pricing four tiers of AI risk as one trade

Where the four rungs separate is duration. "Leases can outlast GPUs, GPUs can outlast customer contracts, and debt can outlast both," PIMCO wrote of neocloud-tenanted sites ^[10]. A creditor to a hyperscaler-tenanted site underwrites the tenant's balance sheet and the lease, not utilization at that site. When the chips age, the tenant swaps them and keeps the same shell and power infrastructure ^[8]. One rung down, the creditor is underwriting a neocloud's ability to resell compute profitably. That depends on utilization, pricing, customer retention and access to capital ^[9].

The payment for taking that difference has the same shape at every rung: coupon, principal, and at most some spread compression, against leverage, execution, utilization, technological obsolescence and refinancing risk ^[4]. PIMCO calls the proposition asymmetric ^[4]. The published commentary describes the dispersion qualitatively and does not give spread levels for the four categories ^[3].

A buyer of AI credit at compressed spreads is therefore declining to be paid for a gap. It is the gap between a leased shell with a hyperscaler covenant behind it and a GPU fleet that has to be re-contracted as the hardware ages ^[9].

There is a case for the compression, and PIMCO makes it before qualifying it. Vendor financing, hyperscaler equity stakes in neoclouds and residual-value backstops link the categories, though "a hyperscaler guarantee can be softer than it appears" ^[12]. If the links hold, lending at the far end is partly lending at the near end, and one spread for the chain is defensible. If they do not, the same links carry a neocloud's utilization problem back toward the balance sheets that were supposed to be insulated from it ^[7].

PIMCO's preference is the top of the chain at both tails. Hyperscaler credits, it argues, offer the cleanest risk/return profile there. Spread upside is limited in the good case, because successful monetization prolongs the capex cycle and keeps issuance elevated ^[14]. In the bad case, balance sheet flexibility and rapid capital discipline provide meaningful protection ^[15].

In my view the dispersion arrives with the supply. PIMCO expects the funding gap to persist and debt supply to keep growing, bringing a wider variety of issuers, structures and risk exposures to market ^[5]. A market cannot separate issuers it does not yet own. The thesis fails if the next wave is mostly hyperscaler-tenanted leases with completion obligations and guarantees attached. Then utilization and technology risk stays inside the enterprise ^[7], the creditor is underwriting a tenant it can already price ^[8], and flat spreads are the right answer.