Salesforce Q2 EPS $5.90 vs $3.27 (26 August 2026): $2.53 of It Was a Portfolio Mark, and Operating Income Went Nowhere

Salesforce reported second-quarter fiscal 2027 results after the US close on Wednesday 26 August 2026, and the headline looks like one of the largest earnings beats of the season: non-GAAP diluted EPS of $5.90 against an LSEG consensus of $3.27, up 103% year over year. Then read the line above it. Income from operations was $2,331m, against $2,332m in the same quarter last year — a decline of one million dollars on $1.1bn more revenue. The gap between those two facts is a single item: $2,613m of gains on strategic investments, worth $2.53 of the $5.90, on a portfolio Salesforce did not sell.

This print was set up as a test of one line. Revenue and EPS had effectively been pre-announced in May, because a subscription business guides tightly on contracted revenue; the only figures capable of carrying new information were current remaining performance obligation and the cash flow statement. That held. cRPO accelerated. Revenue beat by $25m. And the EPS line, which nobody was supposed to learn anything from, moved by eighty times the revenue beat — because of an item that has nothing to do with selling software.

What actually happened

Every figure below is from the Exhibit 99.1 financial statements filed with the SEC on 26 August, with consensus figures as reported by CNBC.

Two rows in that table did the work, and they point in opposite directions. cRPO cleared its consensus and accelerated. Income from operations did not grow at all.

The $2,613m line, and why it is not a sale

Salesforce carries a large book of privately held equity stakes, reported on the balance sheet as strategic investments — $11,324m at 31 July 2026 against $7,591m at 31 January. Under US accounting rules, when a private holding reprices in a subsequent funding round, the holder marks its stake to that new value and the change runs straight through the income statement, below the operating line but above net income.

That is what the $2,613m is. It is not revenue, not a disposal, and not cash. The clearest proof sits in the cash flow statement, where the entire $2,613m is subtracted again as a non-cash adjustment when reconciling net income to operating cash flow — which is why net income of $3,526m converted into only about $1.3bn of operating cash.

CNBC reported that the company attributed the gain to its stake in the AI developer Anthropic, noting that Anthropic said in May it had raised equity funding valuing it at $965bn, and that Alphabet and Microsoft have flagged gains on the same holding in recent weeks. That last detail is the part worth holding onto: a single private valuation event is now large enough to be visible in the quarterly earnings of several of the largest listed companies in the world, none of which sold anything.

Where the extra $1.1bn of revenue went

Revenue rose $1,109m year over year and operating income fell $1m. That is not a rounding artefact; it is identifiable cost lines absorbing the entire increment.

The composition matters more than the total. Cost of subscription and support revenue rose 22.9%, to $2,021m from $1,645m — roughly twice the rate of the revenue it supports, which pulled subscription gross margin down about 170 basis points to 81.3%. Amortisation of purchased intangibles rose to $522m from $380m, the arithmetic consequence of the Informatica acquisition. Stock-based compensation was $904m against $793m. Restructuring and acquisition-related costs were $115m against $4m. Research and development grew 13.9% and sales and marketing 12.1%, both faster than revenue.

Only some of those are excluded from the non-GAAP figures, which is why non-GAAP operating income still grew 10.3%, to $3,872m, roughly in line with revenue — and why the non-GAAP operating margin still slipped, to 34.1% from 34.3%. On either measure, this was a quarter with no operating leverage. On the GAAP measure it was a quarter of 230 basis points of margin contraction.

The forward book accelerated while recognised revenue slowed

Here is the genuinely encouraging part of the release, and it is the part that carried new information.

Current remaining performance obligation reached $33.5bn, growing 14% year over year and 14% in constant currency. The constant-currency series now reads 10% in the year-ago July quarter, 13% in April, 14% in July — two consecutive quarters of acceleration in contracted revenue expected within twelve months. Management guided Q3 cRPO growth to approximately 14% again, and stated explicitly that the guide contains no contribution from the pending Contentful and Fin acquisitions. President and CFO Robin Washington said net new annual order value growth was "the strongest it's been in four years, keeping us on track for second-half organic revenue reacceleration."

At the same time, constant-currency subscription revenue growth decelerated from 12% in the April quarter to 11% in July. Both things are true because a subscription contract is recognised ratably: this quarter's revenue was set by deals signed several quarters ago, while cRPO reflects what was signed recently. A forward book accelerating into a decelerating revenue line is precisely the pattern that has to appear first if a second-half reacceleration is going to be real, which is why it is evidence rather than proof.

The segment split is the honest frame for the AI question. Agentforce Apps — Agentforce Sales, Service, Marketing, Commerce, Flex Credits and Slack, $7,193m of revenue in the quarter — grew 8%. The faster line, at 20%, is Data 360, Headless Platform and Other, which is where Informatica sits, and it decelerated from 23%. On the call, Washington cited "headwinds and volatility" in selling licences for integration and analytics software, per CNBC's report.

Agentforce crossed $1.5bn — with a redefinition inside the growth rate

Salesforce disclosed Agentforce annual recurring revenue above $1.5bn, up over 240% year over year, and combined Agentforce and Data 360 ARR of nearly $3.9bn, up over 210%. Marc Benioff, Chair and CEO, said the company is "seeing incredible demand for our AI and data products, with ARR about to cross $4 billion."

One sentence in the release governs how that 240% should be read: "Effective Q2 FY27, Agentforce ARR includes our AI offerings, Slackbot and Headless 360." The measured quantity changed in the same quarter the growth rate went from just over 200% to over 240%. That does not make the number wrong or the growth unreal — it makes the comparison with the $1.2bn reported in May not like-for-like, and there is no disclosure that lets an outside reader unwind it. Note it, and move to the units that were defined consistently: 3.2 billion Agentic Work Units delivered in the quarter, up 97% sequentially, out of 7.0 billion to date, and 104 trillion records ingested by Data 360, up 355%.

The scale arithmetic has not changed. Roughly $1.5bn of ARR against a full-year revenue guide of $46.1bn–$46.4bn is about 3% of the company. A line compounding at triple digits off that base can be transformative in three years and still be arithmetically incapable of moving this year's growth rate. Benioff addressed the opposite fear directly on the call, in remarks reported by CNBC: "This is not the SaaSpocalypse... these dire predictions about the end of software and how the models eat everything, but none of them have come true for us."

The share count, the interest bill, and a guidance raise that is mostly M&A

Three capital-structure effects are now running through the per-share figures at once, and none of them is operating performance.

Diluted shares fell to 821m from 962m, a 14.7% reduction, following the $25bn accelerated share repurchase that delivered 103 million shares up front; final settlement is expected in October 2026. Treasury stock rose to $55.0bn from $32.2bn since January and total equity fell to $38.4bn from $59.1bn. The debt that funded it is now visible: noncurrent debt of $39,288m against $10,439m at 31 January, and interest expense of $473m for the quarter against $67m a year earlier — an annualised run-rate cost of roughly $1.9bn. A smaller denominator lifts EPS; the interest is a charge that reduces the earnings being divided. Full-year operating and free cash flow growth guidance stayed at approximately 4% to 5%.

The guidance raise deserves the same decomposition, and Salesforce supplied it: full-year revenue moves to $46.1bn–$46.4bn, an increase of $200m made up of "$100 million of organic growth, $200 million from the pending Contentful and Fin acquisitions, and $100M FX headwind." Full-year GAAP operating margin guidance was cut to 20.1% from 20.6%; non-GAAP was held at 34.3%. Meanwhile the full-year non-GAAP EPS guide jumped to $16.67–$16.71 from $14.06–$14.12 — about $2.60 at the midpoint, on a $200m revenue increase that at a 34.3% margin is worth roughly seven cents. The rest is the mark already booked in the first half, carried forward on the stated assumption that the portfolio does not move again.

The dollar footnote

One thread runs from this release to the currency board, and it reversed direction this quarter. Salesforce's May guidance embedded a currency tailwind; the August release states that "with the U.S. dollar strengthening in Q2, Salesforce now expects a reduced currency tailwind," and the raise explicitly contains a $100m FX headwind for the full year. Q3 revenue growth is now guided at 11–12% both reported and in constant currency — a $0m year-over-year FX effect, where the prior quarter's reported growth sat above its constant-currency growth.

That is the transmission channel working in the ordinary direction. Europe was $2,764m of the quarter's revenue and Asia Pacific $1,168m, so roughly a third of billings are translated from non-dollar currencies. A firmer US dollar compresses reported growth for every S&P 500 constituent with that mix, a quarter or two after the move, without a single underlying contract changing. The dollar's own direction is a rate, growth and risk question rather than a software one — the five-factor read exists to keep those inputs separate.

For the wider complex: the AI capex side of the same week is in the Nvidia Q2 result, and the same discipline of reading a headline growth rate against its second derivative is in the CrowdStrike net-new-ARR note. What this print settles is narrow, and worth stating plainly: the forward book is accelerating, the operating margin is not expanding, and the earnings line that moved the stock was mostly a valuation somebody else set.

Educational macro context only — not investment advice.