PGIM's Peters says the credit crowding has barely started

Gregory Peters, co-chief investment officer at PGIM Credit, used an episode of the podcast Credit Exchange with Lisa Lee to argue that credit crowding is in its early stages, and he built the case on demand pools that the direct lending market has not yet priced. Per Creditflux's summary of the episode, those pools are the franchising of the AI digital infrastructure buildout and government spending, both interest rate-insensitive and requiring enormous amounts of debt.

That is a different argument from the price debate the market has been having, in which two years of crowding talk has meant too many lenders in one sponsor process, thinner documentation, tighter spreads. A borrower with no sensitivity to the coupon is not shopping on spread at all; it is shopping on size and execution, which turns the contest into a capacity test, and capacity is where the field narrows fast.

Both pools are asset pools rather than companies, which is where this publication has argued private credit's next frontier sits, and reading a buildout or a government program for credit means underwriting contracted or budgeted cash flows instead of a sponsor's EBITDA at longer duration and far larger ticket sizes than the typical mid-market unitranche. The skill set that wins here overlaps with infrastructure debt and public finance more than with sponsor M&A, so the crowding Peters describes will not stay contained inside the direct lending market he competes in.

The demand is not hypothetical. In August, NVIDIA signed memorandums of understanding with six firms toward $500 billion of compute financing, with the hard terms left private, which is as close to a template as the infrastructure pool has produced. The allocator side is moving too, if EIG's $2.1 billion separate-account sleeve out-raising its $1.9 billion infrastructure fund is a guide, with limited partners choosing a bespoke mandate over the commingled vehicle that gives the platform its name.

The strain lands on teams. Origination desks built to win mid-market unitranche compete on proprietary deal flow and closing speed, while desks built to hold enormous, rate-insensitive, long-dated paper compete on warehouse capacity, syndication, and laying off risk in size, and those are different businesses with different talent. The marginal dollar of new demand is arriving in the second.

With no figure for either pool, no target allocation, and no timeline in Creditflux's paywalled summary, "early stages" stands as a manager's claim rather than a measured share of the market, and the falsifiable version arrives with the first priced facility. The NVIDIA memorandums of understanding signed in August still have no public terms. When the first of those prints, the market will find out what rate-insensitivity costs in spread.