In the quarter ended 2026-06-30, Goldman Sachs and SoFi each reported a segment with Platform in its name going sharply backwards, inside a quarter the rest of the business made look excellent.
Goldman's Platform Solutions posted net revenue of $221M, down 64%, with a pre-tax loss of $48M — the only one of Goldman's three segments losing money, in a record quarter. SoFi's Technology Platform posted net revenue of $84.5M, down 23%, with contribution profit down 65%.
Drawn on a chart, those are the same picture. They are opposite diagnoses. Goldman's segment is shrinking because Goldman decided it should. SoFi's shrank because a large client left.
The useful question a reader can carry away from this is not which company is in trouble. It is was this decline chosen — and these two quarters happen to answer it about as cleanly as it is ever answered.
First, the units, because they are a trap
Goldman reports segments in millions; SoFi reports them in thousands. Goldman's 221 is $221M. SoFi's 84,505 is $84.5M. Read straight off the two tables the segments look 380x apart. They are 2.6x apart — closer in size than almost anything else about them.
Goldman Sachs — segment net revenues, Q2 2026 ($M)
| Segment | Net rev | YoY | Pre-tax | Segment ROE |
|---|---|---|---|---|
| Global Banking & Markets | 15,520 | +53% | 7,499 | 27.5% |
| Asset & Wealth Management | 4,597 | +20% | 1,112 | 13.8% |
| Platform Solutions | 221 | −64% | (48) | (5.8)% |
SoFi — segments, Q2 2026 ($K)
| Segment | Net revenue | YoY | Contribution profit | YoY | Contribution margin |
|---|---|---|---|---|---|
| Lending | 724,798 | +63% | 399,033 | +63% | 55% |
| Financial Services | 466,278 | +29% | 212,676 | +13% | 46% (from 52%) |
| Technology Platform | 84,505 | −23% | 11,772 | −65% | 14% (from 30%) |
Goldman: a business being closed on schedule
Platform Solutions is Goldman's consumer-banking remnant, and the cause of the 64% is named in the release: Apple Card portfolio markdowns, with the portfolio held for sale. That is not a business losing customers. It is a business being marked down on the way out the door, by a bank that has been executing this retreat for years.
The size makes the point better than the percentage. Platform Solutions is $221M against Goldman's record $20,338M of net revenues — 1.087%, or about one dollar in ninety-two. The segment could go to zero and Goldman's quarter would still be a record. Meanwhile Global Banking & Markets was up 53% and diluted EPS was up 92%.
Our July analysis got this right on the day, calling the collapse "closer to 'mission accomplished' than 'problem emerging'". Nothing here revises that. It is the correct reading and it is worth repeating, because it is exactly the reading the SoFi line does not support.
SoFi: the margin is the number, not the revenue
SoFi's Technology Platform lost revenue for a reason with a name: accounts fell 16% year over year to 135 million after one large client fully transitioned off the platform before 2025-12-31. A customer left and took its volume.
The revenue line understates what that did. Contribution profit fell 65% on revenue down 23% — and the ratio between them is the sharpest figure in either company's release:
Contribution margin: 14%, from 30%. Check it: 11,772 / 84,505 = 13.9%.
That is operating leverage running backwards. When a segment's revenue falls 23% and its profit falls 65%, the cost base did not leave with the customer. The margin halving is what separates SoFi's line from a smaller version of Goldman's — Goldman is disposing of assets it has decided not to own, while SoFi is carrying a platform's fixed costs against a materially smaller book of accounts.
Derived, and worth showing because it self-checks: back-solving the disclosed changes puts prior-year Technology Platform revenue near $109.7M and prior-year contribution profit near $33.6M, a margin of 30.6% — which reproduces the "from 30%" the release states. The two disclosed percentages and the two disclosed margins agree with each other.
Our July piece printed the revenue decline and the profit decline. It never printed the margin. That is not an error — it flagged the segment under What to Watch with precisely the right open question, "does it stabilize, or is there more churn coming?" — but the two numbers sat in one sentence without the division between them being taken, and the division is where the story is.
Why neither of these made news
Both segments are small enough to vanish into their own quarter's headline, and both did.
Goldman's headline was record net revenues of $20.3B and diluted EPS of $20.98, up 92%, with investment-banking fees of $3,395M. A 1.1% segment losing $48M does not survive that.
SoFi's headline was its own company-stated Rule of 40 score of 70, its 19th consecutive quarter above the threshold — on total net revenue of $1,218.7M, up 43%, and net income of $156.6M, up 61%. Worth one caveat this site owes its readers: that 70 is SoFi's definition, adjusted revenue growth plus adjusted EBITDA margin. r40's own score uses revenue growth plus FCF margin, so the two will not match and are not meant to.
Neither company hid anything. Both segments are broken out in the release. Both companies named the cause. The gap was never disclosure.
Why this site never made the comparison
It would be easy, and wrong, to say either article missed something. Both reached the right conclusion about their own segment, independently, on the day of the print.
The reason the comparison went unwritten is visible in where each piece filed its segment. Goldman's went into a paragraph under The Rest of the Franchise, as a completed strategic story. SoFi's went into a one-line bullet under What to Watch, as an open risk. Both placements are correct. Filed correctly in two different articles two weeks apart, by writers each holding one company, the two lines had no way to meet.
The site has published this shape before, from one company at a time: Robinhood's crypto revenue fell 38% to $100M inside a record $1,308M quarter, and that piece is the closest precedent here — same pattern, different structure, since crypto is a product line inside a segment rather than a reported segment of its own.
What this piece does not claim
- No forecast for SoFi's Technology Platform. The capture records one client exiting and accounts up 2 million sequentially. One quarter of sequential growth against a 16% annual decline is not a trend in either direction. The next print is the test.
- No judgement that Goldman's segment is finished. Held for sale is a disclosed intention, not a completed transaction.
- Nothing about either share price. This repository stores no previous close for any company, so no piece here can say how either stock reacted, and this one does not.
- No claim that either July article was wrong. They were not, and the figures in both reconcile to the captures.
The transferable part is the question. Two segments fell hard in the same three months, at similar absolute size, inside two quarters that were otherwise excellent. One was a plan working and one was a customer leaving, and the fastest way to tell them apart was not the revenue line — it was asking what happened to the margin underneath it.
Every reported figure is from this site's captures of the two Q2 2026 releases, both for the quarter ended 2026-06-30. Goldman: this site's capture, from Exhibit 99.1 to an 8-K (Item 2.02), accession 0000886982-26-000294, reported 2026-07-14 — the segment table, the $20,338M total net revenues, $20.98 diluted EPS, $3,395M investment-banking fees, and the Apple Card markdown and held-for-sale cause. SoFi: this site's capture, from Exhibit 99.1 to an 8-K, accession 0001818874-26-000050, reported 2026-07-29 — the segment table including the 14%-from-30% contribution margin, total net revenue $1,218,676K, net income $156,592K, the company-stated Rule of 40 score of 70 and its 19th consecutive quarter, and the 135 million accounts, −16% YoY, after one large client fully transitioned off before 2025-12-31. Goldman's segment figures are in $ millions and SoFi's in $ thousands; every cross-company comparison here is made after converting both to millions. Derived figures, marked as such above: Platform Solutions at 1.087% of net revenues (221 / 20,338); Technology Platform's 13.9% current contribution margin (11,772 / 84,505); the 2.6x size ratio between the two segments ($221M / $84.5M); and the back-solved prior-year Technology Platform revenue of ~$109.7M, contribution profit of ~$33.6M and ~30.6% margin, obtained from the disclosed −23% and −65% and offered only because they reproduce the release's stated "from 30%". Two quotations come from this site's own July articles rather than from either release and are attributed where they appear. No share price, price move or market-capitalisation figure appears here.