Autoresearch: GS SLR, book quality, and Q3 2026 print after the Q2 FY2026 release
Q3 2026 is unprinted (IR date Oct 13). Last official SLR is still 4.3% (June 30 10-Q). Solomon's 6–18 month recalibration has not fired. Parent wholesale nonaccrual 1.5% / $3.486B is not a private-credit book rate. Do not upgrade step 3.
Autoresearch: GS SLR, book quality, and Q3 2026 print after the Q2 FY2026 release
Generated by
/autoresearchon 2026-09-19. Synthesized across 3 rounds from 16 successful web fetches (1 failed), anchored by GrokipediaGoldman_Sachs. See Provenance. Treat as raw material — review before promoting into a project or thread. Context: vault/projects/stock-marketPriors capture skipped — user instructed skip.
Tight gap for ai-financing-supercycle-to-gs-capital-markets-rerate step 3 (
partial). Record Q2 revenue is realized; the re-rate is forward; SLR caps; the July 14 call offered no GS-book quality number. GS Private Credit Corp Q2 repurchase ~3.24% below the 5% cap (vs BCRED/Apollo/Blue Owl gated). Do not treat gating as defaults. Do not re-rate GS. No buy / sell / size.
Summary
Q3 FY2026 has not printed. Official IR still dates the next earnings release Tuesday, October 13, 2026 (GS conference-call calendar). The IR hub and the /financials/10q page still feature the July 14, 2026 Q2 release as the latest print (IR; 10-Q hub). EDGAR’s latest parent 10-Q is the June 30 filing accepted 2026-08-03 (EDGAR 10-Q index).
SLR is still the last official constraint print, still 4.3% as of June 30, 2026 — confirmed in the 8-K exhibit and restated in the 10-Q, not updated since (2Q26 earnings exhibit; June 30 10-Q). After the January 1, 2026 early adoption of the modified eSLR, the holding-company requirement is 3.75% (3% minimum + 0.75% buffer). 4.3% sits 55 bp above that floor. Coleman on the July 14 call still treated SLR as one of several “oscillating, binding constraints” with “a limit to our appetite to expand” the balance sheet — he did not name a target floor (Fool Q2 transcript).
Solomon’s mid-July “recalibration in 6–18 months” has not fired. On September 16 at Barclays — the last scheduled public appearance before October 13 — he still used future tense: bumps “when there’s a recalibration,” and “at the moment, there’s certainly a lot of opportunity” (earningscalls.dev Barclays transcript). Two months into a 6–18 month window is not a conversion.
Book-quality: do not upgrade step 3. The July 14 call still had no private-credit portfolio non-accrual rate. What arrived after the call is (a) parent wholesale-loan tables in the August 3 10-Q — nonaccrual $3.486B / 1.5% of amortized-cost loans, not a named $31B private-credit book rate — and (b) GS Credit (the evergreen BDC vehicle, CIK 1920145) 0.1% / $27.21M, one name, as of June 30. That vehicle is not the firm’s $31B deployment. Silence on the load-bearing “GS originates better credit, not just more” test is still silence.
Findings
Theme 1 — Q3 FY2026 is unprinted; October 13 remains the date
The August 18, 2025 IR calendar still lists Third quarter 2026 – Tuesday, October 13, 2026, 7:30 a.m. ET release / 9:30 a.m. ET call (GS press release). The December 18, 2025 dial-in update kept the same date (updated call details — search-confirmed; not re-fetched).
As of 2026-09-19:
- IR homepage “Latest Featured Financials” still opens with Jul 14, 2026 — “Goldman Sachs Reports 2026 Second Quarter Earnings Per Common Share of $20.98 and Annualized Return on Common Equity of 23.5%.” The next-newest earnings item is Apr 13 (Q1). The page also flags a Sep 16 Barclays 24th Annual Global Financial Services Conference presentation — a conference, not a print.
- IR 10-Q hub still headlines the same Q2 release.
- Parent EDGAR 10-Q list: newest row is 2026-08-03, accession
0000886982-26-000297, period June 30, 2026. Next-older is 2026-05-01 (Q1). No September 2026 10-Q (EDGAR).
The July 14 print itself is unchanged from the known Q2: net revenues $20.34B, net earnings $6.63B, diluted EPS $20.98, annualized ROE 23.5% (GS Q2 release; 8-K exhibit 99.1). ROTE 25.5% is on the call, not in the short HTML release (Fool transcript).
Post-July parent 8-Ks that were filed (Series AA preferred, July 20 / July 23) are capital-structure items, not an earnings or SLR update (search-only; not treated as fetched bodies).
Theme 2 — SLR is still 4.3%; it is still a named binding constraint, now against a 3.75% eSLR floor
| Date | SLR | Source | |
|---|---|---|---|
| Dec 31, 2025 | YE 10-K / 10-Q compare | 5.2% | June 30 10-Q leverage table |
| Mar 31, 2026 | Q1 (8-K prelim. / call) | 4.7% (call); 8-K Q1 prelim. had 4.6% | Fool Q2 transcript |
| Jun 30, 2026 | Q2 8-K + 10-Q | 4.3% | exhibit 99.1; 10-Q |
| Sep 19, 2026 | Q3 | unprinted | IR + EDGAR |
The June 30 10-Q restates the 8-K preliminary 4.3% as the quarter’s SLR: Tier 1 capital $114.382B, total leverage exposure $2,646.205B, Tier 1 leverage 5.4%, SLR 4.3% (10-Q). Standardized CET1 12.9%, 150 bp above the 11.4% requirement — same as the call.
eSLR context (why 4.3% is no longer “below a 5% floor”): “On January 1, 2026, the firm early adopted the modified Enhanced Supplementary Leverage Ratio (eSLR) standards, which replaced the 2% buffer applicable to G-SIBs, with a buffer equal to 50% of the firm’s G-SIB surcharge (Method 1).” Holding-company SLR requirement = 3% + 0.75% = 3.75% as of June 2026 (vs 5.0% requirement at December 2025). GS Bank USA SLR 5.6% vs a 3.75% requirement (10-Q).
So: SLR is still the lowest-among-peers number management cited on July 14, still the last official print, and still inside the band Coleman called binding. It is not below the new regulatory floor. Whether it remains the governing constraint on financing growth is still Coleman’s qualitative claim — no post-June 30 ratio, no named floor, no Q3 update.
July 14 Q&A (exact): Christian Bolu: “Your SLR ratio is now the lowest among peers. How much does SLR govern constraints on growing the financing business?” Denis Coleman: “we have a number of different, often oscillating, binding constraints as a firm. We manage to all of them. We’re managing the CET1, we’re managing the SLR. … Ultimately, there will be a limit to our appetite to expand that” (Fool transcript).
The September 16 Barclays transcript as fetched does not mention SLR (earningscalls.dev; Seeking Alpha reprint timed out).
Theme 3 — Recalibration has not fired (future tense on Sept 16)
The dated falsifier is still the July 14 call. Solomon, answering Ebrahim Poonawala: “I’m not smart enough to tell you whether or not there can be recalibrations, in the short term, sometime in the next six months, the next 18 months. … It feels like that will continue, I know that it won’t be a straight line, and there’ll be bumps and there’ll be recalibrations” (Fool transcript). Same call: financing demand is “outstripping what we think is the appropriate quantum.”
On September 16, 2026 at Barclays, Solomon did not say the recalibration had started. On AI financing: “if everyone is right and the build-out of the compute capability in the next 5 years is going to take $8 trillion, there’s going to be a lot of financing to do that. Now I’m not sure it’s going to be a straight line.” On credit structure: “we know what a real investment-grade offtake agreement looks like. And then we also know when people are doing structured things that are getting investment-grade ratings where fundamentally the risk is not the same as true investment-grade risk. … It’s not at a scale at the moment that I’m overly concerned.” Close: “I’m sure we’ll have bumps like everybody else when there’s a recalibration. But at the moment, there’s certainly a lot of opportunity.” (earningscalls.dev).
That is still a forward warning, not a conversion. Mid-July + 6 months is ~January 2027; +18 months is ~January 2028. This pass is day ~67 of that window.
He also said AI financing is “not driving all the financing activity” and financing-revenue growth “is not going to be a straight line because it’s just correlated to market activity and market cap” (same transcript). Softening-if-drawdown is a correlation caveat, not a declaration that the 6–18 month recalibration has arrived.
The fetched Barclays file is truncated by the host (“first half of this call”). SLR, FICC sequential color, and any expense guide that secondary write-ups attributed to the conference are not treated as fetched.
Theme 4 — Book quality: two numbers exist; neither is the $31B private-credit book rate. Step 3 stays partial.
What the July 14 call still did not give: a GS private-credit / alternatives portfolio default, non-accrual, or underwriting-standard rate. Coleman’s credit comment was the $102 million provision, “primarily reflected impairments related to wholesale loans,” and loans $261B, +3% sequential (Fool transcript; exhibit 99.1). That is a P&L provision, not a book-quality rate.
What the August 3 parent 10-Q did give (wholesale loan book, amortized cost):
| Metric (June 30, 2026 vs Dec 31, 2025) | June 2026 | Dec 2025 | Source |
|---|---|---|---|
| Nonaccrual loans | $3,486M / 1.5% of gross amortized-cost loans | $3,387M / 1.6% | 10-Q Note 9 |
| of which Corporate | $1,784M | $2,065M | same |
| of which Commercial real estate | $1,186M | $1,079M | same |
| of which Other collateralized | $424M | $121M | same |
| Past due (30+ days) | $851M / 0.4% | $770M / 0.4% | same |
| Allowance for loan losses | $2,188M / 0.9% | (Q2 start $2,345M) | same |
| Allowance / nonaccrual | 62.8% | 63.4% | same |
| Q2 provision (loans) | $124M in the allowance rollforward; $102M P&L provision net of lending-commitment release | — | 10-Q + exhibit 99.1 |
| Q2 charge-offs | $185M | — | 10-Q |
| Modified loans (Q2) | $229M; defaults after modification “not material” | — | 10-Q |
This is the consolidated wholesale loan book (corporate, CRE, residential, securities-based, other collateralized). It is not a disclosed non-accrual rate on the $31B private-credit deployment or on alternatives AUM. The other-collateralized nonaccrual jump ($121M → $424M) is the only intra-book move that sits near FICC-financing / structured lending — and the 10-Q does not name those credits as AI-infra or private-credit. Do not invent that link.
What GS Credit (evergreen BDC) gave, August 7/10 10-Q + shareholder letter:
“As of June 30, 2026, the Company had an investment held in one portfolio company on non-accrual status, which represented 0.1% and 0.1% of the total investments (excluding investments in money market fund, if any) at amortized cost and at fair value.” As of December 31, 2025 the same one name was 0.2% cost / 0.1% FV. Table: performing $18,361.98M; non-accrual $27.21M; total $18,389.19M (GS Credit June 30 10-Q).
The August shareholder letter repeats 0.1% “slightly lower than last quarter,” PIK 3.6% of investment income, “Bad PIK” 0.2%, and claims repurchase requests “have so far remained below the 5% quarterly cap and have, to date, been fulfilled in full” (EX-99.1). Platform color in that letter: GSAM Private Credit >$180B AUM as of June 30, 2026; “over 275 investment professionals.” That is vehicle + platform marketing, not a parent-book default tape.
Q2 repurchase (already known; now final-amended): expired June 23; 12,076,761.403 shares accepted (~3.24% of the March 31 share count vs a 5.0% cap); paid on or about July 30 at $24.57 NAV (SC TO-I/A, July 31). Gating ≠ defaults.
Q3 repurchase is open, not reported. New SC TO-I filed 2026-08-25: offer to purchase up to 18,707,703 shares (~5.0% of 374,154,077 shares outstanding as of June 30) at September 30 NAV; expires 11:59 p.m. ET September 22, 2026 (SC TO-I; Offer to Purchase). EDGAR’s TO list has no amendment after August 25 (EDGAR TO index). This pass cannot say whether Q3 requests came in under or over the cap. Do not treat an open tender as a gate, and do not treat a gate as a default.
Boundary that keeps step 3 partial: a parent wholesale-loan nonaccrual table plus a GS Credit vehicle 0.1% is not a disclosed quality number on the $31B one-quarter private-credit deployment that step 2 cites. The “discipline, not reach” claim is still management assertion + vehicle-level (not firm-book) data. Do not upgrade step 3 from this pass.
Theme 5 — Grokipedia is a historical primer, not a 2026 print
Goldman Sachs (fetched via _lib/grokipedia.py, 25,017 chars capped, 340 citations) covers founding through the 2008 BHC conversion, 1MDB, and ~2024 revenues ($53.5B). It does not carry Q2/Q3 2026 EPS, SLR, private-credit deployment, or Solomon’s recalibration line. Vocabulary only.
Contradictions and open questions
- Which book is “the GS book”? Parent wholesale nonaccrual 1.5% vs GS Credit 0.1% vs the undisclosed $31B deployment. Using either disclosed number as the step-3 quality test would be a category error.
- Other-collateralized nonaccrual $121M → $424M in the parent 10-Q is the only intra-book deterioration near financing. The filing does not name the credits. Watch, do not infer AI-infra defaults.
- Is SLR still governing after eSLR? 4.3% vs a 3.75% requirement is a wider cushion than “4.3% vs the old 5%.” Coleman still called it binding among oscillating constraints. No Q3 ratio.
- Q3 GS Credit tender results (expire Sept 22) are not on EDGAR as of this pass. A print above 5% would be a liquidity event, not a default.
- Barclays transcript is truncated on the fetched host. Secondary write-ups that put sequential FICC / expense color in Solomon’s mouth are not used.
- Risk.net (Aug 19: “Morgan Stanley joins Goldman below old 5% SLR threshold”) is paywalled / not fetched. Peer-low remains the July 14 call claim, not a 2026-09 table.
Provenance
Rounds run: 3 of 3 (full)
Sub-questions by round:
Round 1 (broad survey):
- Has Goldman Sachs posted Q3 2026 earnings, or only scheduled a date, after the July 14 Q2 FY2026 print?
- Has GS disclosed SLR after Q2, and is it still described as the governing constraint?
- Has GS disclosed portfolio non-accruals, underwriting standards, or book credit quality (firm or Private Credit Corp) since Q2?
- Has David Solomon said the mid-July “recalibration in 6–18 months” has started?
- What did the Q2 10-Q / 8-K actually disclose on credit quality and SLR as the last official print?
Round 2 (drill-down):
- Exact parent 10-Q SLR table, eSLR requirement, and wholesale nonaccrual / past-due / allowance figures — targeting “no GS-book number” vs standard 10-Q loan tables.
- GS Credit June 30 10-Q footnote for the 0.1% claim, and whether a post-June repurchase TO exists — targeting vehicle vs parent.
- Full Barclays Sept 16 remarks on SLR / recalibration / financing — targeting whether the falsifier has fired.
- Any parent 8-K after July 14 that updates SLR or credit — targeting a hidden Q3 pre-announce.
Round 3 (resolve remaining uncertainty):
- IR 10-Q hub + parent EDGAR 10-Q index — targeting “Q3 still unprinted.”
- GS Credit EDGAR SC TO-I index — targeting a Q3 tender.
- Q3 SC TO-I + Offer to Purchase (filed 2026-08-25) — targeting cap, expiration, and whether results exist.
Anchor source (Grokipedia, fetched before round 1):
- Goldman Sachs — 25,017 chars extracted (capped) / 340 citations — historical primer through ~2024; no 2026 print, no SLR, no recalibration line.
URLs fetched (16 successful, 1 failed):
Round 1:
- GS conference-call dates (4Q25 and 2026) — official IR — Q3 date Oct 13, 2026.
- GS Q2 2026 earnings release — official IR — $20.34B / $20.98 / 23.5% ROE; no Q3.
- GS Investor Relations — official IR — latest featured financials still Jul 14; Sep 16 Barclays listed as a presentation.
- Fool: GS Q2 2026 earnings-call transcript — transcript reprint (Reliable; periodic timeout on other Fool URLs) — SLR 4.3%, Coleman “oscillating, binding constraints,” Solomon 6–18 month recalibration, $102M provision.
- GS Credit EX-99.1 Q2 2026 fund update — SEC primary — vehicle 0.1% non-accrual, PIK 3.6%, platform >$180B.
- earningscalls.dev: GS at Barclays, 2026-09-16 — transcript reprint (truncated) — recalibration still future tense; $8T / IG-offtake color.
Round 2:
- GS June 30, 2026 Form 10-Q — SEC primary (filed 2026-08-03) — SLR 4.3%, eSLR 3.75% requirement, wholesale nonaccrual 1.5% / $3.486B, past due 0.4%, allowance 0.9%.
- GS Credit June 30, 2026 Form 10-Q — SEC primary (accepted 2026-08-10) — one name, 0.1% cost / 0.1% FV, $27.21M.
- GS Credit SC TO-I/A (final), July 31, 2026 — SEC primary — Q2 tender 12,076,761.403 shares accepted in full; ~3.24% vs 5% cap.
- 2Q26 GS Earnings Results (8-K Ex. 99.1) — SEC primary — $102M provision; SLR 4.3% vs 4.7%.
[Failed: https://seekingalpha.com/article/4947212-the-goldman-sachs-group-inc-gs-presents-at-barclays-24th-annual-global-financial-services]— timeout (not in Reliable; treat as Periodic-failure this pass).
Round 3:
- GS IR 10-Q hub — official IR — still the Q2 headline; no Q3 10-Q.
- EDGAR: GS parent 10-Q list — SEC index — newest 10-Q 2026-08-03.
- EDGAR: GS Credit SC TO-I list — SEC index — newest TO is 2026-08-25; no later results amendment.
- EDGAR index 0001193125-26-365299 — SEC index — Q3 TO documents.
- GS Credit SC TO-I, Aug 25, 2026 — SEC primary — 5% of June 30 shares; expires Sept 22; results not filed.
- GS Credit Offer to Purchase, Aug 25, 2026 — SEC primary — same cap / Sept 30 NAV.
SOURCE_RELIABILITY routing: Preferred goldmansachs.com IR, sec.gov, fool.com (Reliable, with known periodic Fool timeouts). Skipped Reuters (Hard-blocked), CNBC (Persistent-failure 403), Risk.net (paywall), federalregister.gov HTML (Hard-blocked; eSLR numbers taken from the GS 10-Q instead). goldmansachs.com fetched cleanly this pass (not previously listed). Seeking Alpha timed out.
Tools used: WebSearch, WebFetch, grokipedia.py fetch (Goldman_Sachs). No --include-x.
Generated: 2026-09-19 21:00 UTC