Product1 distinct publisher3 min readPublished
Adjusted EPS came in at $5.90 against a $3.27 consensus. The revenue beat was 0.26% and the full-year raise 0.43%, which is where the question about agents and seats actually sits.
The Product Desk · Product desk

Compiled by The Product DeskSomething wrong?How this is made
Start with the profit line, because that is what carries the argument. Net income rose $1.64 billion year on year, from $1.89 billion to $3.53 billion [3][18]. The gain Salesforce booked on its stake in Anthropic was worth more than $2.6 billion [4], following a May round that put the startup at a $965 billion valuation [5]. The investment mark is bigger than the entire increase in profit [18]. SiliconANGLE's report does not say whether that gain sits inside the $5.90 adjusted figure or outside it [2], and the distinction is the whole read: one version is a quarter that earned its money from customers, the other is a quarter that got a large piece of it from a private valuation nobody trades.
The demand evidence is thinner than the headline suggests. Revenue cleared consensus by $30 million, or 0.26% [1][19]. The full-year midpoint moved up by $200 million, 0.43%, to $46.25 billion [11][21]. And Salesforce's own third-quarter EPS guide of $3.42 to $3.44 is about 58% of the quarter it just posted [7][20], which is the company telling you the shape of this profit does not repeat.
On growth, the number that matters barely moved. The quarter grew 11%, and the raised full-year range implies roughly 11% at the midpoint [1][11][25]. That cuts against the cannibalisation story Marc Benioff called nonsense [12], since seats are plainly not falling off. It also cuts against the opposite claim, that agents are additive revenue now. Nothing in the top line is accelerating.
Benioff's supporting statistic deserves reading closely: nine of the top 10 AI companies use Salesforce and Slack, and their combined spend is up 435% from a year earlier [13]. That is ten logos and no disclosed base. A 435% increase on a small starting figure is compatible with almost any contribution to a $46 billion company. The structural case he makes is the more durable one, that models need the customer data, context and workflows held inside the platform [14], and the new Claudeforce plugin, which lets Claude draft emails for salespeople from Salesforce and Slack data, is that case in product form [15].
Valoir's Rebecca Wettemann, quoted in the same report, adds the complication: Slack is becoming the place agents live, but Salesforce has an Agentforce versus Slack identity problem, and customer confusion follows if it goes unresolved [16].
One more piece of arithmetic. The stock was down 22% for the year before the print [9]; a 13% jump in late trading [10] still leaves it around 12% below where it started [22]. Investors have repriced the seat model on the fear that firms will build with AI what they currently license [17]. This quarter did not disprove that fear so much as postpone the test, and it is worth noting that both supplied reports of the call are the same SiliconANGLE write-up [23], so everything above rests on one account of what management said.
Ranked by verification strength, evidence, and original report placement.
Salesforce reported second-quarter revenue of $11.35 billion, up 11% from a year earlier, ahead of the $11.32 billion analyst forecast.
Salesforce reported adjusted earnings of $5.90 per share against a Street target of $3.27 per share.
Net income jumped 87% to $3.53 billion, up from $1.89 billion a year earlier.
Salesforce gained more than $2.6 billion thanks to its strategic investment in AI startup Anthropic PBC.
In May, Anthropic announced equity funding that lifted its valuation to $965 billion; Salesforce was one of the main investors in that round.
Free cash flow rose 81% to $1.10 billion, against a Wall Street projection of $643.2 million.
Follow any of these and your For You feed starts watching them — no settings page required.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Specific figures, one vendor-sourced account
The financial figures are precise, internally consistent and checkable by arithmetic (revenue, EPS, net income, free cash flow, guidance ranges, and the Anthropic mark). But both cluster items are the same SiliconANGLE article, so there is a single account, and every operational statistic (435% AI-customer spend, 240% Agentforce growth, cRPO, contract values) is company-supplied on an earnings call with no independent verification in the supplied material.
Real revenue and contracts, unverified agent traction
Adoption evidence is substantive at the franchise level: $33.5 billion of current remaining performance obligations, a $1.6 billion Department of Veterans Affairs contract, $1.5 billion of annualized Agentforce revenue, and a pending $3.6 billion Fin acquisition. It is weaker on the specific AI-agent thesis the story rests on: the 435% AI-customer spend figure and 240% Agentforce growth are vendor-reported, Claudeforce was announced the same day with no usage data, and the CFO conceded license headwinds in integration and analytics.
Beat framing outruns the underlying numbers
The report and the CEO frame the quarter as a crushing beat that disproves the SaaSpocalypse, but the same numbers show a 0.26% revenue beat, a 0.43% full-year raise, roughly 11% guided growth matching the 11% reported, and an EPS beat whose scale depends on a more-than-$2.6 billion Anthropic mark that exceeds the entire year-on-year net income increase. Next quarter's guided EPS midpoint is about 58% of the reported figure, and the 13% pop leaves the stock roughly 12% below where the year began. The overstatement is in interpretation, not in fabricated facts, so the gap is moderate rather than extreme.
Vendor narrative under de-rating pressure
Every load-bearing statement originates with parties holding a direct interest: the CEO is rebutting a thesis that has taken his stock down 22% year to date, the growth statistics are selected and supplied by the company, and the earnings uplift partly reflects a mark on a stake in Anthropic, the same partner whose plugin Salesforce announced the same day. The one external voice, Valoir's Rebecca Wettemann, is an industry analyst quoted by the publisher and raises a positioning concern rather than testing the numbers.
Facts firm, interpretation single-sourced
Confidence is moderate: the numbers are specific and the derived arithmetic follows directly from them, so the quantitative picture is reliable. It is capped by the cluster containing a single publisher's account duplicated twice, no primary filing or transcript, and no independent check on the vendor-supplied adoption statistics.
invest
Claudeforce: Salesforce hands Anthropic the seller's screen and keeps the write path2 distinct publishers
invest
Salesforce's 14% cRPO is the best agent-demand read yet. Organic growth is 6.5%2 distinct publishers
invest
Salesforce's $3.9bn agent line grew 210%, inside a company growing 11%1 distinct publisher
invest
Salesforce's AI answer to the $2 trillion SaaS selloff is $50,000 a deal1 distinct publisher
Distinct publishers with included, body-backed reporting in this cluster.
2 articles · August 26, 2026