Goldman's 72% Equities Surge Masks a 2% NII Decline — The Bond Rout Gave the Trading Desk a Windfall While Quietly Pushing Funding Costs the Other Way

Goldman Sachs reported Q3 2026 net revenues of $18.68 billion and diluted EPS of $17.01 on Oct. 13, handily clearing the $13.02 consensus by 32% and marking the sixth consecutive beat. Revenue was the second-highest quarterly total in firm history. Yet the print is not the clean victory the headline suggests: net revenues fell 8% sequentially from Q2's $18.3 billion (net of certain reporting items), operating expenses ran 20% higher YoY, and the bond-rout macro backdrop simultaneously produced a trading-desk windfall and an interest-expense drag — the two halves of the trade that the brief framed, and that this quarter finally settled.

Goldman's firmwide net interest income of $3.77 billion was 2% lower year over year and 5% lower sequentially from Q2's $3.954 billion — a striking outcome in a quarter where the 10-year yield averaged north of 5.20% and touched 5.36% intraday on Oct. 7. The reason is straightforward: gross interest expense grew roughly 12% YoY (to about $19.0 billion), while gross interest income grew only ~9% (to $22.8 billion), compressing the spread. On a funding-economics basis, higher rates are a tax on Goldman's wholesale-funded balance sheet — the exact mirror of the trading-desk windfall that volatility-driven desks capture.

This is the half of the bond-rout thesis that doesn't fit the trading-desk-windfall narrative. Compare it with JPMorgan Chase, whose net interest income rose 13% YoY to $27.1 billion in Q3 — JPM raised full-year NII guidance to roughly $106.2 billion ($97.2B ex-markets), a function of its larger, deposit-funded retail balance sheet. Goldman runs a far more wholesale-funded, capital-markets-dependent book, and the higher-rate regime is now visibly reducing NII even as the trading desk captures the volatility premium. The combination explains why JPMorgan Chase reported an EPS of $5.57 on revenue that missed estimates, while Goldman Sachs printed $17.01 — the two banks are monetizing different pieces of the rate cycle.

On Sept. 16, CEO David Solomon warned that the third quarter's fixed income, currencies and commodities business had been \"a little bit softer on a relative basis,\" with shares falling nearly 4% on the day. Q3 FICC delivered exactly that picture: net revenues of $3.471 billion were essentially flat YoY (−1%) and down 24% sequentially. FICC intermediation — the part most exposed to client positioning — fell 7% YoY to $2.264 billion, with declines in rates, mortgages, and credit only partially offset by stronger currencies and commodities. The 10-year's 24-year-high move to 5.36% on Oct. 7 did not translate into a trading-revenue windfall the way it did for Morgan Stanley, whose fixed-income line dropped 7% but was overshadowed by a 41% surge in equities.

Equities, by contrast, ran the other direction. Equities net revenues of $6.441 billion were up 72% YoY, with intermediation up 52% and financing up 96% to $3.365 billion — the second-highest financing revenue print in firm history. CFO Denis Coleman framed mid-migration on the call: financing is now over 40% of FICC-plus-Equities revenue year-to-date, and \"demand for Equities financing continues to exceed available industry capacity,\" supporting attractive pricing. The takeaway is that this was not a vol-driven trading windfall so much as a balance-sheet-financing story, with hedge-fund and prime-brokerage balance-sheet demand paying Goldman to warehouse risk.

Investment banking fees of $3.22 billion were 21% higher YoY, despite a 41% YoY fall in Q3 M&A volumes (to $993 billion, per Reuters). The disconnect is the franchise: Goldman Sachs sits at the top of every M&A league table that exists, and on the Reuters dealogic league table YTD 2026 sits at $1.58 trillion of M&A value, ~43% above JPMorgan Chase at $1.10 trillion. Advisory revenue specifically hit a record $1.654 billion — proof that the pipeline of large-cap strategic transactions is converting, even when headline M&A volume doesn't show it.

More important for the thesis: equity underwriting of $666 million was up 43% YoY, and several of the quarter's marquee AI-infrastructure financings are visible in that line. On Aug. 10, Intel announced an upsized $23 billion common-equity follow-on (joint lead bookrunner) to fund AI capacity build-out; SK Hynix's $26.5 billion US listing was a Goldman global-coordinator mandate; and in the high-yield market, Applied Digital's $3.5 billion junk bond led by Goldman to fund AI data-center operations in Louisiana priced in the mid-7% to 8% range in mid-September. These are exactly the trades that put Goldman Sachs at the structural center of the AI capex funding machine — but the fee dollar is still small relative to where the issuance curve is headed.

On the call, Solomon put a number on the prize: U.S. data-center build-out could require \"north of $10 trillion of investment through 2032,\" averaging roughly 3.6% of U.S. GDP annually. Goldman Sachs Research's earlier forecast — published in the firm's June 2026 \"Private Markets Are Expected to Have a Growing Role in Data Center Financing\" piece — projected $5.3 trillion in combined hyperscaler capex from 2025 through 2030, plus another ~$2 trillion in adjacent grid and power infrastructure. AI-related debt issuance globally is already approaching $500 billion, per the Goldman Sachs Exchanges Aug. 5 piece on AI debt reshaping credit markets. This is not a theoretical pipeline: Applied Digital's $36 billion contracted lease book alone is bigger than the market cap of most of the AI data-center names.

Solomon positioned the firm as at the \"epicenter\" of advising, originating, structuring, and distributing across both public and private markets — and stressed disciplined underwriting because financing rates now depend on hyperscaler offtake. The risk he flagged is also structural: as financing needs get bigger, returns and project viability \"viable\" could be pressured by the same rate cycle that's currently pushing Goldman's NII down. That tension — bigger fee pipeline and bigger funding-cost headwind — is the single most important thread for the stock.

The Oct. 13 bank-print pack shows how unusual Goldman's profile is this cycle. JPMorgan Chase delivered $15.3 billion in net income on $50.9 billion of revenue (+8% YoY), with NII up 13% to $27.1 billion and IB fees of roughly $3 billion (+14%) — the deposit-rich, NII-positive profile that benefits from a steep curve. Wells Fargo reported Q3 net income of $6.0 billion and EPS of $1.87 on revenue of $21.87 billion (a ~2% miss), with NII of $12.52 billion (+4.7%) below consensus, ROTCE of 16.3%, and provision for credit losses up sharply to $1.029 billion from $681 million a year ago — the consumer-credit deterioration story. Citigroup reported $5.2 billion in net income on $23.8 billion in revenue (best Q3 revenue in a decade), up 40% YoY, with IB fees of $1.35 billion (+9%). Goldman Sachs, by contrast, is the only one in the group where NII declined — because its funding stack is structurally different.

The sector read-through is simple: a5%-yielding 10-year Treasury and a 3.7% CPI print are still constructive for net interest income at the deposit-heavy banks, but actively negative for Goldman's wholesale-funded book — even as the same volatility supercharges its equities and advisory franchises. That is why the print had a 32% EPS beat alongside a 2% NII decline, and why the stock will trade on the AI capex pipeline rather than on quarter-to-quarter NII.

Operating expenses of $11.308 billion were 20% higher YoY and pushed the efficiency ratio to 60.5%, vs. 57.4% in Q3 2025. Compensation and benefits rose to $5.12 billion (a function of the revenue mix tilting back toward trading and intermediation) and transaction-based expenses climbed to $3.02 billion. Provision for credit losses was modest at $368 million, with a 0.3% annualized net charge-off rate on a $276 billion loan book — credit quality is not the story here. Return on common equity was 18.7% annualized for the quarter and 20.7% year-to-date. Book value per share reached $372.64, up 1.4% for the quarter and 4.2% year-to-date.

Capital returned during the quarter was $4.49 billion: $3.00 billion of buybacks at an average ~$1,035/share on 2.9 million shares, plus $1.49 billion in common dividends. The standardized CET1 ratio fell to 12.4% from 12.9% in Q2 (and 14.3% at year-end 2025) — Denis Coleman described the firm as \"running with a 50 to 100 basis point buffer\" for capital and now at the 100bp level. The message: capital is being deployed into the trading and financing books that drove Q3, and management is comfortable that the buffer holds.

Three things are now established with this print. First, the bond rout has been a clean trading-desk tailwind for Goldman Sachs: Equities +72% YoY, record advisory, and an equities-financing book running at the second-highest level in firm history. Second, the same rate cycle that produced the trading windfall has reduced Goldman's net interest income by 2% YoY because the funding stack is wholesale, not deposit-funded — making Goldman the NII-decline outlier among the big banks. Third, the AI capex role is now visible in the fee line — Intel's $23 billion equity offering, Applied Digital's $3.5 billion high-yield deal, the SK Hynix US listing — but those are still a fraction of the $5.3+ trillion issuance curve Goldman Sachs Research has mapped through 2030.

The investor takeaway: in the short term (days–quarters), the stock should trade on the strength of the equities-financing book and on whether Q4 FICC intermediation can recover from the −24% sequential drop. Over1–3 years, the bigger question is whether Goldman's underwriting, structuring, and balance-sheet financing of the $5–10+ trillion AI infrastructure build-out can lift the fee mix structurally enough to offset the NII drag from a higher-for-longer rate regime. The risks to the thesis are two-fold: a re-acceleration in funding costs that compresses NII further, and any underwriting misstep in a complex, multi-trillion-dollar issuance cycle where project offtake is concentrated in a handful of hyperscalers.