06-reference/research

apollo valor xai disclosure latency measured

2026-10-23·research-brief·source: deep-research·by Ray Data Co (deep-research synthesis)
disclosure-surfacesprivate-creditai-capexbdcinterval-funds

39 days: the Apollo/Valor/xAI exposure surfaced first in an interval fund's repurchase-offer notification, not in any surface the parent brief's latency table ranked

The question

"How long after signing did Apollo's $3.5B capital solution in the Valor/xAI transaction appear in ANY public filing, and in which one — converting the private-credit disclosure-latency table from reasoning into a measured number on a real closed deal?"

Context: this is the empirical test of the latency table in [[2026-08-13-ai-capex-lender-exposure-insurance-pension-vehicles]], which was built from regulatory structure rather than from a measured deal. One cell gets replaced with a number.

Measurement criterion, stated up front. "Appear" is scored at three distinct thresholds, because they are three different disclosures and collapsing them is how this kind of number gets faked:

  1. Borrower-identified holding — the borrower entity is named in a filing as something the filer holds, at any granularity (percentage of assets is enough). No deal terms required.
  2. Named-deal narrative — the filing describes the transaction: lender, borrower, lessee, dollar amounts.
  3. Loan-level schedule row — a mandated Schedule of Investments line with instrument, rate, maturity, par, cost and fair value.

All three are reported below, with their own deltas. The headline number is threshold 1, because the question asks for any public filing.

What we already know (from the vault)

What the web says

Sourcing note: all filing evidence below comes from SEC EDGAR full-text search and the filings themselves, retrieved with curl and a declared contact User-Agent (no 403s; the access pattern the sibling brief established holds). Press coverage was used only for the announcement date.

Convergences and contradictions

Synthesis for RDCO

The measured number is 39 days (±3), and it supports the parent brief's conclusion while invalidating its mechanism. Signing-to-first-filing is 2025-11-20 → 2025-12-29. Widen the close to the 11/17 acquisition date and it is 42 days; the uncertainty is the close date, not the filing date, because no filing states the close precisely — "November 2025" plus a 11/20/2030 maturity plus an 11/17/2025 equity acquisition date is the best available triangulation, and it should be labelled an inference every time it is reused. Threshold 2 is 64 days, threshold 3 is 90 days, Apollo's own BDC is 111 days, and Apollo Global Management as a filer is never. So the parent's clock argument survives intact: even the fastest surface is a month and a half behind the event, and the mandated surfaces are a full quarter behind. Applied to the Nvidia compute-financing platforms, first closings in Q4 2026 would surface named exposure around January to February 2027 at the fast end and Q2 2027 at the mandated end — still at or after the Q3 2026 to H1 2027 turn window. The parent brief's verdict does not need to be walked back. Its constant was wrong by about seven weeks in our favour; its conclusion was right.

What does need rewriting is the table's row order, and that has a buildable consequence. The fastest name-level private-credit disclosure surface in the US is not the BDC 10-Q. It is the Form N-23c-3 repurchase-offer notification filed by private-credit interval funds, because these funds attach retail marketing collateral — monthly fund updates with top-10 holdings and borrower descriptions — to a procedural filing nobody thinks of as a disclosure document. The exposure surfaced here not because a regulator required it but because Apollo's retail distribution machine wanted to advertise the trade. That is a durable incentive, not an accident: a $3.5B GPU-lease financing is a selling point to an interval-fund buyer, so it gets named in the collateral, which then gets filed. The monitoring hook is concrete and cheap — track N-23C3A, N-CSR and N-CSRS filings from the dozen-odd large private-credit interval funds (Apollo Diversified Credit, Blue Owl Alternative Credit, Diameter Dynamic Credit, Oaktree Asset-Backed Income, BlackRock HPS Credit Strategies, Cliffwater, Variant, FS Credit Income) via EDGAR full-text search on borrower-name patterns. All of them showed up in this deal's search results at some point in 2026.

The second structural finding is about where exposure hides, and it cuts against reading the listed parent. The entity that signed is not the entity that discloses. Apollo Global Management, the NYSE-listed filer a 13F or a 10-K reader would naturally interrogate, has never named this deal. The disclosure lives in affiliated vehicles — one interval fund, one non-traded BDC, two Apollo Origination capital trusts — and, notably, in competitors' filings: Blue Owl's four BDCs, Diameter Credit, Great Elm, Lord Abbett, Western Asset, T. Rowe Price, Invesco, Fidelity and Artisan all name VCI entities somewhere in 2026. A syndicated private-credit deal is therefore most findable by searching the whole filer universe for the borrower's name, not by reading the lead arranger's filings. That is the opposite of how RDCO's existing smart-money machinery is pointed [[2026-10-22-13f-eligibility-bdc-feeder-interval-funds]], and it is the generalisable method: borrower-name full-text search across all filers, not filer-name holdings pulls.

Calibration, stated plainly. Confidence that 2025-12-29 is the first filing appearance: high — it is the single EDGAR full-text hit in the window, across all forms. Confidence in the 39-day figure: medium-high, limited entirely by the inferred close date. Confidence that the monthly-update document was first publicly available on 2025-12-29: low — Apollo almost certainly posted that November update to its own website in mid-December, so 39 days is the filing-system latency, which is what was asked, and not the information-availability latency, which is shorter and unmeasurable from EDGAR. Do not let those two numbers get swapped in any downstream artifact. And n=1: one deal, one sponsor with an unusually large retail-distribution footprint. A deal syndicated purely into institutional separate accounts would have no interval-fund marketing surface at all and would print first at threshold 3, at 90 days or worse.

Why this is in the vault

It closes the sharpest of the seven follow-ups on [[2026-08-13-ai-capex-lender-exposure-insurance-pension-vehicles]] with a measured number, and it decides a specific build question that brief left open: the lender-exposure monitoring hook for the financing-fragility tail risk on the memory capital-cycle thesis should be borrower-name full-text search across all EDGAR filers, with private-credit interval-fund N-23c-3 filings as the fast leg — not BDC Schedules of Investments, and not the lead arranger's own filings.

Open follow-ups

Related

Sources

Vault:

Primary filings (SEC EDGAR):

Press (announcement date only):