Anchor story: On Monday, November 9, 2026, the Federal Reserve Banks of Dallas and New York announced a voluntary pilot survey of the U.S. private credit direct-lending market. It closes on November 18, 2026, and aggregate findings are due in the first quarter of 2027. The stated reason: visibility into the market is limited.
Why it matters now: The private credit market is becoming a key funding channel for AI infrastructure, while bond investors are already repricing risk. Reporting from CNBC and Deutsche Bank describes CoreWeave’s bonds due 2032 paying a 9.625% coupon but now yielding above 16%, with five-year credit default swap spreads above 1,289 basis points. The Fed survey and the bond market are asking the same question from two directions.
Practical lesson: Make a two-column map this week. Column one: the AI infrastructure exposure you already own, and whether it is backed by contracted customers and hard collateral, or by future-capacity promises. Column two: the names where the re-price is creating an entry. Answer it before the next stress print comes out.
Charon: Welcome to Ad Valorem Syndicate Room, episode fifty five. I'm Charon. This week the signal is not a headline about how much AI builders want to spend. It is the bond market quietly refusing to fund it. And two Federal Reserve banks just opened a survey to find out who is lending that capital in the first place.
Kore: Right. On November ninth, the Federal Reserve Banks of Dallas and New York announced they are opening a pilot survey of the U.S. private credit direct lending market. It is voluntary, it closes on November eighteenth, and aggregate findings are due in the first quarter of twenty twenty seven. The reason it matters is the Fed says visibility into this market is limited.
Charon: That is doing a lot of work in that sentence. Limited visibility means the regulators who would be first to notice stress are, by their own account, the last to see it. The survey will break the market into upper, middle, and lower middle market borrowers, and it will look at how fast credit is being extended, whether lending standards are loosening, and what loan documentation actually looks like. So when the data lands, it will be the first systematic read on the health of a multi trillion dollar shadow credit market.
Kore: And the private credit market right now is the fuel for the AI build out. Which brings us to the second signal, and it is harder than the survey. CoreWeave, the neocloud that rents out GPU clusters, has become the poster child for AI credit stress. Its bonds due twenty thirty two pay a nine and three quarter percent coupon, but they are now yielding above sixteen percent.
Charon: And that gap is the whole story. A new unsecured lender would not be paid nine and three quarter. They would be paid close to sixteen, if they could get in at all. Reporting from CNBC and Deutsche Bank puts CoreWeave five year credit default swaps above twelve hundred and eighty nine basis points, which is roughly twelve point nine percent of the debt insured per year. That is distressed credit pricing on a company that was still raising a year ago.
Kore: And it is not just CoreWeave. Reports from this week, including from Bloomberg and Tilt, describe CoreWeave, Nebius, and Nscale as effectively shut out of the corporate bond market. Lenders are demanding much higher coupons, tighter covenants, and customer guarantees. In the broader data center sector, at least eighteen live deals are now requiring annual returns above ten percent.
Charon: So why do the Fed survey and the CoreWeave bond belong in the same episode? Because they are the same question from two directions. The bond market is the price signal. It is telling us that leveraged, unsecured financing for AI infrastructure no longer trades at growth prices. The Fed survey is the search for the missing map. If the credit behind the build out is concentrated in opaque private vehicles, then a widening in one name like CoreWeave can become a widening in the whole private credit complex, and the first people to see it might be the people whose job is to monitor it.
Kore: Here is the context that keeps this grounded. This is not a prediction that AI capex collapses. The build out is still real, and the compute commitments behind it are enormous. What is changing is the price of risk. The market is starting to distinguish between AI companies with hard collateral and contracted customers, and ones that are asking unsecured bondholders to carry a bet on future capacity. The first group still raises. The second group is paying for that distinction.
Charon: Why does that matter now for a smaller investor or family office? Because the mispricing is on both sides, and both sides are reachable. On the sell side, if you own exposure to levered AI infrastructure through funds, secondaries, or public wrappers, you want to know whether that exposure is the collateralized, contracted kind, or the unsecured, future capacity kind. On the buy side, the same stress that is scary is also where secondaries and special situation opportunities start to appear, because holders who entered at growth valuations are now the ones feeling the re price.
Kore: The downside to hold against that is simple and honest. If the bond stress stays contained to a few names and the Fed survey comes back benign, the discount may just be noise, and the special situation angle under performs. And if the survey flags loosening lending standards in the lower middle market, the problem is not AI at all, it is the whole private credit complex, and you want to know that before you concentrate a position.
Charon: Our suggestion, and this is not advice, is one concrete step this week. Make a two column map. Column one, the AI infrastructure exposure you already own, and for each holding, whether it is backed by contracted customers and hard collateral, or by future capacity promises. Column two, the names where the re price is creating an entry, so you are not caught buying only on the way up. The data to build both columns is public. The question is whether you will have answered it by the time the next stress print comes out.
Kore: Position before you predict. And remember, the numbers here are dated and reported. The bond yields and CDS levels are from this week, the survey window is open through November eighteenth, and the Fed aggregate findings are not due until the first quarter of twenty twenty seven. None of this is a call on any single company.
Charon: For sources, the primary references are the Federal Reserve Bank of Dallas and New York announcement on the private credit pilot survey from November ninth, the CNBC and Deutsche Bank reporting on CoreWeave bond yields and credit default swap spreads, and the Bloomberg and Tilt reporting on the AI data center credit tightening. All linked in the episode notes at gp dot ad valorem dot io.
Kore: Thanks for listening. This content is AI generated, for informational purposes only. If you want to walk through your own AI infrastructure exposure, book a thirty minute call with the Frontier Alternatives Fund team. Until next time, from Ad Valorem.