cRPO Clears the Guide and Claudeforce Lands, But the EPS Blowout Is a Portfolio Markup: Maintaining Outperform After a 25% Re-Rate

Results vs. Consensus

Q2 FY27 Scorecard

Year-over-Year Comparison

Sequential Comparison vs. Q1 FY27

Quality of Beat/Miss

• Revenue: The $25M beat against a $11.32B consensus is inside the rounding of a $11.3B quarter, and the more useful comparison is against the company's own $11.27B to $11.35B range, where the print landed at the top. Composition is the issue: $456M of the revenue is Informatica, which was not in the year-ago base, so organic revenue was roughly $10.889B against $10.236B, or +6.4%. Professional services revenue actually declined 3.8%. Growth is narrowing onto subscription, which is the right place for it to narrow, but the organic subscription line is not yet accelerating in reported dollars.
• Margins: This is the most under-discussed line in the print. Non-GAAP operating margin of 34.1% is 20 bps below the year-ago 34.3%, the first year-over-year contraction in that metric across our coverage window. The cause is not operating expense: non-GAAP operating expense grew 9.3% against revenue growth of 10.8%, so the leverage story is intact at roughly 60 bps. The cause is gross margin. Non-GAAP gross margin fell about 80 bps to 80.0% because cost of subscription revenue rose 22.9% against subscription revenue growth of 11.7%, and only $105M of that $376M increase is explained by purchase-accounting amortization and stock compensation. Management did not decompose the remainder.
• EPS: GAAP diluted EPS of $4.29 is up 119%, and non-GAAP of $5.90 is up 103%, but GAAP income from operations was $2,331M against $2,332M a year ago. Operating profit was flat on 11% revenue growth. Everything below the operating line did the work: a $2,613M strategic-investment gain, offset partly by interest expense that went from $67M to $473M, and a 14.7% smaller share count. Strip the portfolio gain and non-GAAP EPS is roughly $3.37 against a $3.27 consensus, which is a genuine but ordinary beat.
• Cash: Operating cash flow of $1,269M (+71%) and free cash flow of $1,098M (+81%) look spectacular on a Y/Y basis, but Q2 is Salesforce's smallest cash quarter and the comparison base was low. The more informative figure is the first half: operating cash flow of $7,970M, up 10.4%, against a full-year guide of approximately 4% to 5% growth. That guide implies second-half operating cash flow of roughly $7.7B against $7.78B a year ago, in other words approximately flat to slightly down. The interest bill is the reason, and it is now visible in the guide rather than only in the commentary.

Segment Performance

Salesforce reports subscription and support revenue in the two-category framework it adopted in Q1 FY27: Agentforce Apps (Agentforce Sales, Service, Marketing, Commerce, Apps Flex Credits, and Slack) and Data 360, Headless Platform, and Other (Data 360, Data 360 and Platform Flex Credits, Headless Platform, Informatica, Agentforce MuleSoft, Agentforce Tableau, and Other). Prior periods are recast. There is no operating-margin disclosure at the category level.

Revenue by Product Architecture

Agentforce Apps: $7.19B, +8% CC, 66% of subscription revenue

The dominant block accelerated 100 bps in constant currency, from +7% in Q1 to +8% in Q2. That is small in absolute terms and large in narrative terms, because this category is the one the bear case is about: the core Sales, Service, Marketing, Commerce and Slack estate, wrapped in Agentforce branding and sold with a flex-credit consumption layer on top. If artificial intelligence were cannibalizing seats, this is where it would show up first, and it did not show up this quarter.

The composition within the category shifted meaningfully. Slack delivered its fastest quarterly net new annual order value growth since the acquisition, and Slackbot users grew more than 150% quarter over quarter. Sales and service were characterized as resilient with durable seat growth. Marketing, which has been an explicit drag for five consecutive quarters, was described for the first time as recovering. Commerce, which was called out as a drag alongside marketing in Q1, was not addressed in the results commentary.

"Agentforce apps delivered solid growth, anchored in momentum in Slack and resilience in sales and service, with early signs of recovery in marketing. While it is too soon to call this a sustainable trend, we are very encouraged by the success enterprise customers are having on our next-generation marketing product."
— Robin Washington, President and Chief Financial and Operating Officer

Assessment: +8% constant currency is the second consecutive quarter in the 7% to 9% band for this category, after a 6% print in Q4 FY26, which is neither the collapse the bear case predicted nor the reacceleration the bull case requires. The hedged framing on marketing is appropriate and we would not extrapolate it. The genuinely new information is that Slack has become the growth engine of the applications block rather than an integration project, which changes what a "core CRM saturation" argument has to cover. For the FY28 framework to hold, this category needs to reach double digits in constant currency, and the mechanism management has named for getting there is the premium-edition upgrade cycle rather than seat expansion.

Data 360, Headless Platform, and Other: $3.62B, +20% CC, 33% of subscription revenue

The high-growth category decelerated 300 bps in constant currency, from +23% to +20%. Two forces are pulling in opposite directions. Informatica continues to contribute and continues to grow faster inside Salesforce than it did outside, and Data 360 volume metrics remain extreme: 104 trillion records ingested in the quarter, up 355% year over year, of which 82 trillion arrived through Zero Copy, up 731%. Against that, the legacy integration and analytics assets are still eroding, and the on-premises license component of that revenue is lumpy by construction.

"Headless platform, Data 360, and other growth was fueled by continued momentum in Informatica and Data 360, partially offset by license revenue headwinds and volatility in integration and analytics."
— Robin Washington, President and Chief Financial and Operating Officer

"Integration and analytics" is the disclosure wrapper for MuleSoft and Tableau. Tableau was flagged in Q1 for "increased softness in bookings and renewals," and bookings weakness typically shows up in revenue two to three quarters later, which places the effect in exactly this window. The word "volatility" is doing work that "decline" would do more honestly, and neither asset has been sized since it was folded into the combined category.

Assessment: The deceleration here is the arithmetic mirror of the Agentforce Apps acceleration, and net of the two the total subscription line slipped 100 bps in constant currency. We treat the Data 360 deceleration as lower quality than the Agentforce Apps acceleration is high quality, because the drag is structural (legacy analytics losing share to artificial-intelligence-native query tools) while the offset is cyclical (on-premises renewal timing). The Informatica proof point remains the strongest single piece of evidence that Salesforce has broken its own post-acquisition deceleration pattern, and it is now roughly two-thirds of the way through its first full year inside the company.

Revenue by Geography

Assessment: Europe is now the fastest-growing region in constant currency for the first time in this coverage window, accelerating 100 bps to +13% while the Americas decelerated 100 bps to +10%. That inversion matters more than it looks: the Americas is 65% of revenue, so a 100 bps deceleration there is worth more absolute dollars than the European acceleration recovers. Asia Pacific held at +12% constant currency but printed only +9.1% nominal, the widest currency gap in the disclosure, consistent with management's note that a stronger dollar reduced the expected currency tailwind. Salesforce's chief operating officer credited European teams with originating the internal Claudeforce deployment pattern, which is a small data point but consistent with where the growth is.

Key KPIs and Consumption Metrics

Assessment: The consumption metrics are impressive and largely uncheckable, which is the standing problem with this disclosure set. Two entries are genuinely load-bearing because they connect usage to money. Bookings from the premium Agentforce editions more than doubled sequentially, and half of Agentforce bookings came from customers refilling flex credits they had already consumed, which is the closest thing to evidence that consumption converts to renewal rather than to a one-time land. The 2,000 net additions of paying customers into production, up 70% quarter over quarter, is the funnel metric that should lead ARR by two to three quarters. Everything above those two lines is engagement, and engagement has never been the bear case.

Key Topics & Management Commentary

Overall Management Tone: This was the most combative call of the coverage window and the most confident. Management opened by naming the bear thesis, listing six of its specific predictions, and asserting that each had failed, before any financial disclosure. That posture is a clear shift from the measured, evidence-first framing of the prior two quarters, and it was earned by a booking metric rather than by a revenue metric. Where the call was least convincing was on economics: three separate questions on how Claudeforce and headless access will be priced produced descriptions of flexibility and optionality rather than a model, and the operating-margin decline was not addressed.

Claudeforce and What Was Actually Announced

Salesforce and Anthropic announced Claudeforce live on television minutes before the call opened, and the call led with a video clip of the announcement rather than with results. The product is a Salesforce plugin inside Claude, exposing Salesforce data, workflows, business rules, and governance to the model through an enterprise harness the company calls AIforce, built on Model Context Protocol servers, application programming interfaces, and command-line tools. General availability was set for September at Dreamforce.

"This is really the best of both worlds. This is the number one AI in the world, Anthropic, and the number one CRM, Salesforce, coming together for the first time in an incredibly powerful way to build a new product called Claudeforce."
— Marc Benioff, Chair, CEO and Co-founder

The strategic logic is the one Salesforce has been building toward for three quarters: if the frontier models become the interface layer, the company that owns the deterministic data, permissions, and workflow layer underneath them captures a defensible position rather than being disintermediated. The chief executive made that argument explicitly, contrasting probabilistic model behavior with the deterministic guarantees of the underlying platform.

"models are probabilistic systems. They are non-deterministic systems. But our system, those four layers that we talked about with data and apps and semantics and agents, those are deterministic systems."
— Marc Benioff, Chair, CEO and Co-founder

Assessment: Claudeforce is the first announcement in this coverage window that answers the commoditization bear case with a product rather than with a metric. It is also the reason the stock moved 22.6% in one session, and it contributes nothing to FY27 revenue guidance. Investors are being asked to underwrite a strategic repositioning on the strength of a launch that has not yet shipped to general availability. We think the underwriting is reasonable, because the monetization gate (premium editions) already exists and is already priced, but the burden of proof shifts to the Q3 and Q4 prints.

cRPO: The Guide Beat That Settles Last Quarter's Question

Current remaining performance obligation reached $33.5B, growing 14% year over year and 14% in constant currency against a guide of approximately 13% in constant currency. That is a one-point beat, and it is the first time in three quarters that this metric has printed above its own guide rather than exactly on it. The Street was at roughly $33.22B. Management attributed the beat to bookings and to attrition rather than to currency, acquisition, or duration effects.

"Exceptional Q2 bookings, along with near-record-low attrition, drove our Q2 CRPO beat. H1 net new AOV growth significantly outpaced AOV growth, keeping us on track for second half organic revenue re-acceleration."
— Robin Washington, President and Chief Financial and Operating Officer

"CRPO accelerated quarter-over-quarter with $33.5 billion, up 14% in constant currency. That is one point ahead of our guide, driven by strength in Slack, Agentforce, and Data 360."
— Robin Washington, President and Chief Financial and Operating Officer

One qualifier belongs on this: cRPO includes Informatica in the current period and not in the year-ago period, so the reported 14% is not a purely organic figure any more than reported revenue growth is. The company has never disclosed an organic cRPO figure, and did not this quarter.

Assessment: We wrote last quarter that cRPO landing on guide for two consecutive quarters was the empirical problem with the reacceleration story, and that a Q2 print at 14% to 15% in constant currency would validate the bull case. It printed 14%, at the low end of that bar, with an acquisition contribution that is not sized. That is a pass, not a triumph. It is enough to keep the rating and not enough to raise conviction.

The Organic Growth Gap and the Second-Half Math

Reported revenue grew 11% with a disclosed $456M Informatica contribution, so organic growth was roughly 6.4%, against roughly 8.7% in Q1. Read alone, that is a deceleration of more than 200 bps in a quarter where management declared the reacceleration on track. Read on a two-year stack, it is not: organic revenue grew about 16.8% over the two years to Q2 FY27 against about 17.0% over the two years to Q1 FY27, which is flat. The prior-year comparison base was simply harder this quarter.

The forward math is the more useful lens. The FY27 midpoint of $46.25B against first-half actuals of $22.478B implies a second half of $23.77B, or +10.8% against a second half of FY26 that was $21.46B. That is below the first half's reported +12.0%, but reported growth decelerates in the second half by construction: Informatica entered the prior-year base in November 2025, so its incremental contribution shrinks. Working through the Q3 guide midpoint of $11.46B (+11.7% reported, with slightly above four points of Informatica), organic growth is guided to roughly 7.5% in Q3 and to roughly 8% to 9% in Q4. That is the acceleration, and it is in the guide rather than in the print.

"Definitely, we are committed to the H2 revenue re-acceleration. That's already math. Q3 is math. Q4 is nearly math. But we are committed in sustaining that acceleration, and we are going to hear more at Dreamforce about FY 2028."
— Miguel Milano, President and Chief Operating Officer

Assessment: "Q3 is math" is the most falsifiable sentence management has spoken in four quarters, and it is a welcome change from the framework language of the past year. The mechanism is real: net new annual order value converts to revenue with a lag, and the first-half bookings strength is now visible in cRPO. But this is the fifth consecutive quarter in which second-half acceleration has been asserted rather than observed, and the definition of "second half" has moved forward each time. Q3 is now the last quarter in which the claim can be deferred rather than tested.

Agentforce ARR Crosses $1.5B, and the Definition Moves With It

Agentforce annual recurring revenue exceeded $1.5B, up more than 240% year over year, from $1.2B at the end of Q1. The disclosure carries a footnote that deserves more attention than it will get: effective this quarter, Agentforce ARR includes Slackbot and Headless 360, which were not in the metric when it read $1.2B. The chief financial officer confirmed the composition effect in the prepared remarks.

"Agentforce ARR reached $1.5 billion, as you heard, driven by Slackbot and headless launches and continued AI momentum. Upgrades to our premium Slack editions have tripled since we launched Slackbot."
— Robin Washington, President and Chief Financial and Operating Officer

No like-for-like figure was provided. The 25% sequential step and the 240% year-over-year figure are therefore both measured on a basket that changed mid-series. The combined Agentforce and Data 360 ARR of approximately $3.9B, up more than 210%, is subject to the same effect through its Agentforce component.

Assessment: This is not an accounting problem and it is not a governance problem, because the change is disclosed in the first line of the metric's own footnote. It is a modeling problem: an artificial-intelligence ARR series that has been the single most-watched number in the story since it was first disclosed just lost its comparability, at the exact moment the number crossed a round threshold. Analysts building an FY28 Agentforce ARR bridge should treat the $1.5B as a new series with one data point. We would like a restated prior-period figure at the Investor Day.

Premium Editions: the Monetization Mechanism, Finally Named

For two quarters the open question on headless access has been how Salesforce gets paid when a customer reaches its data through a coding agent or a third-party model rather than through a Salesforce screen. This quarter the answer arrived, and it is neither a per-call price nor a revenue share. It is an edition gate: Claudeforce, headless, and the agentic capabilities require an upgrade to the premium editions.

"Every customer will be able to buy it. They will have to upgrade to our premium editions, and it is going to be a new wave of demand. I am very excited about it."
— Miguel Milano, President and Chief Operating Officer

The company then sized the pool. Roughly 5% of the knowledge workers using sales and service have upgraded to the higher-end editions, and those editions carry a 60% to 80% price premium. Separately, management said customers that begin an agentic engagement are already at double the annual order value, with a path to three or four times. Neither figure is audited, and neither is in guidance.

"Only 5% of the knowledge workers that use sales and service have upgraded to the higher-end editions. We get a 60% to 80% premium. Imagine."
— Miguel Milano, President and Chief Operating Officer

Assessment: This is the most important disclosure of the quarter after cRPO, and it reframes the growth debate. If the applications block grows at 8% in constant currency on a 5%-penetrated premium tier priced 60% to 80% above standard, the arithmetic ceiling on that block is far above where it is running, and the constraint is upgrade velocity rather than seats or market share. It also explains why management is comfortable letting Claudeforce contribute nothing to FY27 guidance: the product is the reason to upgrade, and the upgrade is the revenue. The obvious risk is that edition upgrades are a negotiated, renewal-cycle event, so the conversion is measured in years rather than quarters.

Slack Becomes the Growth Engine of the Applications Block

Slack delivered its fastest quarterly net new annual order value growth since the acquisition. Slackbot passed one million active users five months after launch, growing more than 150% sequentially, and premium Slack edition upgrades have tripled since Slackbot shipped. Internally, Slackbot is credited with 8.1 million hours of annualized productivity, more than double the prior quarter. Management introduced Slack Code, a shared workspace for teams of humans and agents.

The strategic point is that Slack is the surface where the premium-edition upgrade actually happens. It is the product with the shortest path from free usage to a paid tier, and it is where the third-party agent ecosystem is being built. The chief executive's aside that he receives regular inbound interest in buying Slack was made in the context of arguing the stock was mispriced, not as a disposal signal.

Assessment: Slack has moved from an integration liability to the fastest-converting monetization surface in the portfolio inside four quarters, and it is now doing more work in the growth story than Agentforce Sales or Service. That is a genuine change in the shape of the business and it is under-modeled. The caution is that Slack's absolute revenue is not disclosed, so its contribution has to be inferred from order-value commentary rather than measured.

The $2.6B Strategic-Investment Gain

Gains on strategic investments were $2,613M in the quarter against $6M a year ago, and $3,171M in the first half. The balance-sheet strategic investment line grew from $7,591M at January 31 to $11,324M at July 31. The company does not attribute the gain to any individual holding in the filing. On the call, the chief executive referenced the value of the company's Anthropic position while arguing that the market was mispricing Salesforce, and connected that argument directly to the buyback decision.

"Take the value of Slack. Obviously, I get a call every day from somebody who wants to buy Slack. Hey, take the value of Slack. Then take the value of our Anthropic stock. That has been like half our value. Forget our cash flow or our customer base. That is why I said, 'Robin, go try to buy as much back, as much Salesforce stock as you can.' I think we got a lot of stock at a good price."
— Marc Benioff, Chair, CEO and Co-founder

Assessment: The gain is real, it is non-cash, it is non-recurring by the company's own guidance convention (forward periods assume no portfolio change), and it should be excluded from every earnings-power calculation. It is not, however, economically meaningless: a $3.7B increase in the carrying value of the strategic portfolio in six months is a real asset, and the position that appears to be driving it is in the same company Salesforce just announced its flagship partnership with. That concentration cuts both ways, and it is now large enough that a mark in the other direction would be equally visible in GAAP earnings. The chief executive's characterization of the position's scale was made loosely on a live call and is not a disclosed figure.

Gross Margin: the Cost of Serving Artificial Intelligence Shows Up

Cost of subscription and support revenue grew 22.9% against subscription revenue growth of 11.7%. GAAP gross margin fell 145 bps to 76.7%. Adding back the amortization of purchased intangibles in cost of revenue ($234M against $150M) and stock compensation in cost of revenue ($147M against $126M), non-GAAP gross margin still fell about 80 bps to 80.0%. That $105M of purchase-accounting and equity-compensation growth explains only about a quarter of the $376M increase in subscription cost of revenue.

Operating expense discipline was not the problem. Non-GAAP operating expense grew 9.3% against revenue growth of 10.8%, contributing about 60 bps of leverage. Netting the two produces the reported non-GAAP operating margin of 34.1% against 34.3%, the first year-over-year contraction in this metric across our coverage window. The full-year non-GAAP operating margin guide was nonetheless maintained at 34.3%, which implies roughly 34.1% in the second half.

Assessment: Two explanations fit the residual cost growth: Informatica's revenue mix carries lower delivery margin than the Salesforce base, and inference plus data-platform costs are scaling with agentic usage. Both are plausible, and management decomposed neither. The distinction matters a great deal: acquisition mix is a one-time reset that anniversaries, while inference cost is a variable that scales with exactly the consumption metrics the company is promoting hardest. Given that Agentic Work Units nearly doubled sequentially and Data 360 record ingestion more than doubled year over year, we would want this decomposed before extending the Rule of 50 margin path. This is our first question at the Investor Day.

The $25B Buyback at a $176 Average

The accelerated share repurchase is running at a $176 average price to date, and management expects to retire at least 14% of shares outstanding through the program, with final settlement expected in October 2026. The diluted share count is already down 14.7% year over year to 821 million. At the $25B program size and the disclosed average price, roughly 142 million shares are implied in total against the 103 million delivered initially, leaving roughly 39 million more at settlement.

"Execution of our $25 billion accelerated share repurchase continues. We expect to repurchase at least 14% of shares outstanding through this program, thus far at an average price of $176 per share."
— Robin Washington, President and Chief Financial and Operating Officer

The funding is visible on the balance sheet. Noncurrent debt went from $10,439M at January 31 to $39,288M at July 31, an increase of $28.8B against $24,842M of debt issued net of costs in the first half; current debt went from $4,000M to zero over the same period, and no corresponding repayment appears in financing cash flows. Treasury stock rose from $32,228M to $55,022M, and total stockholders' equity fell from $59,142M to $38,378M. Net of $8,310M cash and $3,093M marketable securities, net debt is approximately $27.9B. Interest expense of $473M in the quarter annualizes to roughly $1.9B.

Assessment: Last quarter we called the timing of this program into question, arguing it would have been better executed in October at potentially lower prices. That call was wrong on the facts available today: the average execution price is $176 against a current price of $257.56, and the program retired roughly a seventh of the company near the bottom of a two-year de-rating. The cash-flow cost is nonetheless permanent and is now embedded in a full-year guide that implies flat-to-lower second-half operating cash flow. The correct framing is that management traded four to six points of cash-flow growth for a 15% reduction in share count at roughly a 32% discount to today's price, and on the evidence so far that was a good trade.

Contentful and Fin: $200M of Guidance on Deals That Have Not Closed

Two of the three components of the full-year revenue raise come from acquisitions announced in June and still pending at the print. The company expects both to close independently during the third quarter and has incorporated them into every guided metric except cRPO, with the guidance explicitly conditional on the closings. The exclusion from cRPO is a deliberate conservatism: the company is waiting on accurate balances rather than estimating them.

Assessment: The asymmetry is worth naming. The revenue guide includes $200M that depends on regulatory and closing mechanics outside management's control, while the cRPO guide of approximately 14% excludes any contribution from the same deals. That means the Q3 cRPO number will be a clean organic-plus-Informatica read, which is analytically helpful, and it means the revenue guide has a conditional component that could produce a mechanical miss on a timing slip. Half the full-year raise is inorganic and unowned.

Bookings Quality and the Sales Capacity Build

Agentforce bookings doubled year over year, and half of them came from customers refilling flex credits they had already consumed rather than from new lands. Two thousand paying customers moved into production, up 70% sequentially. Bookings from the two premium Agentforce editions more than doubled sequentially. Against that demand, the company is adding capacity: roughly 15,000 account executives today, with 1,200 more to be hired before year-end.

Assessment: The refill share is the highest-quality number in the entire consumption disclosure, because it is the only one that demonstrates a second purchase. Half of bookings from customers who exhausted a prior commitment and came back is a genuine consumption flywheel signal, not an adoption signal. The capacity build is the confirming evidence: a company hiring 1,200 additional quota carriers into a market it believed was decelerating would be making an expensive mistake, and it also tells us where sales and marketing expense growth of 12.1% is going. We would watch for the sales-and-marketing line growing faster than revenue for a second consecutive quarter.

Leadership Changes and the Deferred FY28 Framework

Three organizational items landed with the print. The chief revenue officer was elevated to chief operating officer during the quarter. Investor relations changed hands, with the outgoing head moving to a broader finance role and the incoming head arriving from the sell side. Most consequentially for modeling, the FY28 framework was explicitly deferred to the Investor Day at Dreamforce on September 16.

Assessment: Deferring the multi-year framework to a scheduled event two weeks after the print is normal practice and not a red flag, but it does mean the current price embeds an FY28 expectation that management has not yet published. The prior framework, a $63B revenue target for FY30 raised from $60B at the Q4 FY26 call, is the only multi-year anchor currently on the record. Investor Day is now the single largest near-term catalyst, and the bar has risen materially with the share price.

Guidance & Outlook

Q3 FY27 Guide

FY27 Full-Year Guide

The company decomposed the raise on the call rather than leaving it to inference, which is a welcome disclosure practice and also removes any ambiguity about how little of it is organic.

"We are raising our FY 2027 revenue guide for organic performance by $100 million. This is driven by the continued momentum in Agentforce, Data 360, and Slack. These areas of strength offset the continued volatility in overall licensed revenue."
— Robin Washington, President and Chief Financial and Operating Officer

Implied second-half ramp: The FY27 midpoint of $46.25B against first-half actuals of $22.478B implies $23.77B in the second half, or +10.8% against $21.46B a year ago. Netting the Q3 guide midpoint of $11.46B leaves roughly $12.31B implied for Q4, or +9.9% against Q4 FY26's $11.201B. Reported growth steps down across the second half because Informatica enters the prior-year base from November 2025, while organic growth steps up from roughly 6.4% in Q2 to roughly 7.5% guided in Q3 and roughly 8% to 9% implied in Q4.

Street at: Consensus entering the print sat at $11.32B in revenue and $3.27 in non-GAAP EPS for the quarter, with cRPO at roughly $33.22B. The Q3 revenue guide midpoint of $11.46B and the non-GAAP EPS guide of $3.42 to $3.44 both sit above where the Street was carrying the third quarter before the print, and the full-year non-GAAP EPS guide is now materially above any pre-print model because of the portfolio gain rather than because of operations.

Earnings-power translation: Backing the disclosed first-half strategic-investment contribution of $2.98 per share out of the $16.69 full-year midpoint leaves approximately $13.71 of clean operating non-GAAP EPS. The equivalent figure implied by the prior guide was approximately $13.64. The operating earnings guide moved about seven cents this quarter.

Cash-flow guide: Maintaining approximately 4% to 5% growth after a first half that grew 10.4% requires the second half to be approximately flat to slightly down year over year. Applying the guide to FY26 operating cash flow of $14,996M gives roughly $15.67B for FY27, less the $7,970M first half, or roughly $7.7B in the second half against $7.78B a year ago. The same arithmetic on free cash flow gives roughly $7.4B against $7.5B. Interest expense running at $473M a quarter is the arithmetic explanation.

Guidance style: Salesforce has guided conservatively and beaten narrowly on revenue for the entire coverage window, and this quarter fits: the print landed at the top of the range rather than through it. Two lines cleared by a point rather than landing on the number: constant-currency revenue against a 10% guide, and cRPO, which had been guided precisely enough to land exactly on the number twice. That is the first evidence in a year that the guide is being set below what bookings support.

Analyst Q&A Highlights

The call took only five questions, with a strict one-question-per-analyst limit and an explicit request for brevity, which is unusual for a company of this size and had the effect of leaving the margin decline, the ARR definition change, and the second-half cash-flow implication entirely unexamined.

Claudeforce Go-to-Market and Availability

The first question of the call was about distribution rather than economics: how the new product reaches customers, and whether the two companies' sales organizations would work jointly. Management answered the operational half fully, committing to general availability in September at Dreamforce, confirming every seller would demonstrate the product, and describing an internal deployment in which operating-unit leaders had been using it to run their businesses during the quarter. The pricing half of the answer arrived unprompted and is the more important disclosure: access requires an edition upgrade.

Q: "A lot of great things going on in this quarter, but I have to ask about Claudeforce, and Marc, I think specifically, I would love to understand how you and Miguel are going to be taking this product to the field. What are you doing with Anthropic to make sure that this gets out and gets into the hands of your customers as soon as it is available?"
— Kirk Materne, Evercore ISI

A: "Every customer will be able to buy it. They will have to upgrade to our premium editions, and it is going to be a new wave of demand. I am very excited about it."
— Miguel Milano, President and Chief Operating Officer

Assessment: The most valuable sentence on the call came in answer to a question about field enablement, not about monetization. The edition-gate model resolves two quarters of ambiguity about how headless access gets paid for, and it does so in a way that is checkable against the premium-edition penetration disclosure. That the answer was volunteered rather than extracted suggests management now wants this understood.

Open-Weight Versus Frontier Models, and When Headless Shows Up in Revenue

The second question had two halves: a strategic one about how the industry debate between open-weight and frontier models resolves, and a financial one about when the new surfaces contribute to reported revenue growth. Management answered the strategic half at length and with conviction, arguing that enterprises will consume artificial intelligence through packaged software rather than operating models themselves, and that the model layer will abstract away the same way the hardware layer did. The revenue-timing half was not addressed at all.

Q: "Maybe talk about how you see this debate settling and about the monetization opportunity for both Claudeforce and Headless, and when we should start thinking and seeing it in revenue growth."
— Alex Zukin, Wolfe Research

A: "I believe what they are going to do is consume AI through packaged software. Not a huge statement, just they are consuming their AI through packaged software."
— Marc Benioff, Chair, CEO and Co-founder

Assessment: A clean, unremarked dodge on the half of the question that had a number attached to it. The strategic argument is coherent and we largely agree with it, but "when does this show up in revenue growth" is the question that separates a repositioning from a re-rating, and it went unanswered for the second consecutive quarter. Nothing in FY27 guidance depends on the answer, which is both the reassurance and the problem.

Production Mix and Pilot-to-Paid Conversion Inside Agentforce ARR

The sharpest question of the call went underneath the headline ARR figure to ask what share is paying production deployments rather than pilots, and whether conversion timelines were lengthening as buyers evaluated more alternatives. The answer supplied the only funnel disclosure of the quarter, and it was substantive: bookings doubled year over year, half of them came from customers refilling consumed credits, and 2,000 paying customers moved into production, up 70% sequentially. Management did not, however, give the requested production-versus-pilot percentage of ARR.

Q: "I wanted to circle back on the strong Agentforce ARR growth, and in particular, what proportion of current Agentforce is coming from pay production customers rather than pilots? And how has the timeline from pilot to paid deployment and refilling credits changed over the last couple of months?"
— Elizabeth Porter, Morgan Stanley

A: "50% of the bookings came from customers refilling the tank. So they consume, they use the Flex Credits, they want more, they raise their hand, we go there, which by the way, is good for, A, productivity, because these sales cycles are very short. The time to production. Production is very important for us because consumption is very important to us. We bring customers in the funnel. We want to bring them down. The first thing is to get them to production. We added 2,000 paying customers into production. That is 70% more quarter on quarter."
— Miguel Milano, President and Chief Operating Officer

Assessment: A partial answer that gave more than it withheld. The refill share is the single best evidence yet that Agentforce consumption is recurring rather than experimental, and short sales cycles on refills is a real operating advantage. The unanswered part still matters: without a production-versus-pilot split of the ARR base, the quality of the $1.5B remains an inference, and the definitional change this quarter makes that inference harder rather than easier.

Whether Headless Customers Are Worth the Same as Agentforce Customers

The most economically precise question of the call asked whether Salesforce captures equivalent annual contract value from a customer that reaches the platform headlessly versus one that uses the full Agentforce stack. This is the value-abstraction risk stated in its cleanest form. The response acknowledged that headless monetization is still being worked through and redirected to the premium-edition upgrade path and to pricing flexibility, without addressing the equivalence question. The follow-on commentary offered the 5% penetration and 60% to 80% premium figures, which is the closest thing to a quantitative answer the call produced.

Q: "My question is a little bit on the economic value to Salesforce from these different types of cohorts. I know you're still figuring out monetization on headless. Do you think in steady state, you'll be able to capture the same type of ACV from customers agnostic of whether they do Agentforce or whether they do some sort of headless configuration?"
— Gabriela Borges, Goldman Sachs

A: "We feel really comfortable with, like you said, we're working through headless monetization, but just to further double down on Marc's point, what we ultimately think is predictability and flexibility is what's important to customers."
— Robin Washington, President and Chief Financial and Operating Officer

Assessment: The question asked whether a headless dollar equals an Agentforce dollar and the answer described customer preference. "We're working through headless monetization" is the same status reported a quarter ago. The redirect to premium editions is not evasive, because the edition gate genuinely is the answer for Claudeforce, but it does not cover the case the question was aimed at: a customer that stays on a standard edition and extracts value through interfaces Salesforce does not own. That case remains unpriced.

Durability of Net New Order Value and Attrition

The final question pressed on the foundation underneath the bookings-to-revenue conversion: how durable the four-year-high net new order value growth and near-record-low attrition actually are, and how to think about them going forward. The answer was the most explicit forward commitment management has made in this coverage window, tying the second-half reacceleration to first-half order-value performance the company has already booked and calling it arithmetic. It also deferred the FY28 framework to Dreamforce.

Q: "There are two things that Robin said that I think are really important to that. One, net new AOV growth is the strongest it's been in four years. Two, you had near record low attrition. Those two things help that math work. We know that. Help us better understand what's happening underneath those statements. How stable or sticky is that foundation of your recurring revenue?"
— John DiFucci, Guggenheim Securities

A: "Definitely, we are committed to the H2 revenue re-acceleration. That's already math. Q3 is math. Q4 is nearly math. But we are committed in sustaining that acceleration, and we are going to hear more at Dreamforce about FY 2028."
— Miguel Milano, President and Chief Operating Officer

Assessment: This is the exchange that most changes our forward view, because it converts a narrative into a testable claim with a date attached. "Q3 is math" is checkable in November against the roughly 7.5% organic growth the Q3 guide implies. The hedge sits in the word "nearly" applied to Q4, which is where the Contentful and Fin contribution and the harder organic bar both land. Management chose to be pinned rather than to hedge, which is a meaningful posture change and raises the cost of a Q3 miss considerably.

What They're NOT Saying

1. Organic revenue growth, for the sixth consecutive quarter: Total and constant-currency growth are disclosed, the Informatica dollar contribution is disclosed, but the organic growth rate is never stated. Every quarter investors must derive roughly 6.4% from $456M against a $10.236B base while management speaks in terms of 11%. When the whole thesis turns on organic reacceleration, the absence of the organic number in the company's own reporting is a choice, not an oversight.
2. The Agentforce ARR restatement: The metric now includes Slackbot and Headless 360, and no like-for-like prior-period figure was published. The $1.5B is presented against a $1.2B that was measured on a different basket, and the 240% year-over-year figure inherits the same problem. A single restated line would have cost nothing.
3. What caused the gross margin decline: Cost of subscription revenue grew twice as fast as subscription revenue, and only a quarter of the increase is explained by the amortization and stock-compensation lines the company does disclose. Neither the prepared remarks nor the Q&A touched the residual. Whether it is Informatica mix (which anniversaries) or inference cost (which scales) is the difference between a non-event and a structural change to the margin path.
4. Organic cRPO: The 14% constant-currency figure that carried the quarter includes Informatica in the current period and not in the year-ago base, exactly like reported revenue. The company applies the organic-versus-reported distinction to its revenue commentary and not to its cRPO commentary.
5. Why total RPO is growing at 11% while cRPO grows at 14%: Noncurrent remaining performance obligation grew only 7.5% year over year and fell $1.5B sequentially. Management stated that contract length terms are increasing across all segments in new business and renewals. Both can be true if the mix of what is being signed has shifted, but no reconciliation was offered, and longer contracts would normally build the noncurrent balance faster, not slower.
6. Attribution of the $2.6B investment gain: The filing names no holding. The chief executive referenced the Anthropic position on the call in loose terms while making a valuation argument. For a gain that produced 96% of the quarter's earnings beat and grew the strategic portfolio by $3.7B in six months, concentration disclosure would be appropriate.
7. Commerce Cloud: Named as a drag alongside marketing in the prior quarter, and absent from this quarter's results commentary while marketing was upgraded to "early signs of recovery." Silence on a previously disclosed drag is not the same as its resolution.
8. Any FY28 anchor: The multi-year framework was deferred to the Investor Day, leaving the FY30 revenue target of $63B as the only forward anchor on the record. The share price now embeds an FY28 expectation the company has not published, two weeks before it publishes one.

Market Reaction

• Pre-print setup: CRM closed at $205.62 on August 26, 2026, the print day, having traded $198.95 to $206.44 intraday. The stock entered the print down 22.4% year to date against an S&P 500 that was up 12.1%, and down 15.6% over the trailing twelve months. It had, however, already run 18.4% over the prior thirty days from $173.60 on July 27, so the setup was a stock recovering off a deep base rather than one at its lows. The 52-week closing range entering the print was $150.12 to $266.23.
• Reaction session (August 27): The stock gapped open at $230.05, up 11.9%, traded to a high of $254.48, and closed at $252.05, up 22.6% or $46.43. Volume was 55.5 million shares against a 30-day average of 12.4 million, a 4.5 times multiple. The S&P 500 gained 0.7% on the session.
• Follow-through: The move held and extended. The stock closed at $256.00 on August 28, $257.54 on August 31, and $257.56 on September 1, leaving it 25.3% above the pre-print close four sessions later. There was no give-back, which distinguishes this from a short-covering spike.
• Peer reaction: The enterprise software complex rallied with it. ServiceNow closed up 10.0%, Adobe up 5.7%, Oracle up 2.1%, Workday up 1.5%, and Intuit up 0.6% on the reaction session.

The size of the move is out of proportion to the operating content of the quarter, and that is the point. Revenue beat consensus by 0.2%, the operating beat against consensus net of the portfolio gain was about $0.10, and the organic revenue growth rate decelerated. A 22.6% single-session repricing on that set of facts is not a reaction to the numbers. It is a reaction to the removal of a discount.

What was priced into $205.62 was the possibility that large-scale artificial intelligence would structurally disintermediate application software, and that Salesforce specifically would find its data extracted through interfaces it did not own while its seat base shrank. That is what a 22% year-to-date decline against a rising market was paying for. This quarter attacked that discount from three directions at once: cRPO beat its guide on bookings and attrition rather than on optics, the applications block accelerated rather than decelerated, and the company announced a partnership that positions it inside the frontier model rather than underneath it. Each is modest on its own. Together they made the disintermediation scenario harder to hold with confidence, and a scenario discount unwinds faster than it accumulates.

The peer response supports that reading. ServiceNow rising 10% on a Salesforce print tells you the market re-rated a shared risk factor rather than a company-specific result, since nothing Salesforce disclosed improves ServiceNow's own bookings. The sector had been carrying an artificial-intelligence disintermediation discount, and one credible counter-example was enough to compress it across the group.

The consequence for the risk/reward is direct. At $205.62 the stock was priced near the bear case in our prior framework and the asymmetry was wide. At $257.56 it sits above the middle of our prior base case, and the compensation for the remaining uncertainty (whether organic acceleration actually prints in Q3, whether the premium-edition upgrade cycle converts, whether inference cost pressures the margin path) is materially thinner. The thesis improved this quarter. The entry price improved considerably more.

Street Perspective

Debate: Did This Quarter Actually Refute the Disintermediation Thesis, or Just Postpone It?

Bull view: The bear case made six specific, falsifiable predictions about this quarter, and management enumerated them on the call: seats would decline, customers would leave, pricing power would erode, contract lengths would shrink, agents accessing the platform would kill application usage, and frontier models would remove the need for the customer relationship management layer. Sales and service seats grew, attrition sat near record lows, premium bookings more than doubled sequentially, contract terms lengthened, agentic usage of the platform surged, and the leading frontier laboratory chose to build a joint product rather than compete. Six for six against the bear case in one quarter.

Bear view: None of those data points is the disintermediation mechanism. The mechanism is that once a customer's data is reachable through a general-purpose agent, the application layer becomes a database with permissions and the price of that is lower than the price of a seat. That takes three to five years and would not appear in a quarter where agentic adoption is still being sold as an upgrade. Meanwhile the observable operating fact is that organic revenue growth decelerated to roughly 6.4% and gross margin declined, which is exactly the profile of a business absorbing rising cost-to-serve on flat pricing.

Our take: The bull case wins on the near term and the bear case remains unfalsified on the term that matters. What genuinely changed is that Salesforce now has a product-level answer rather than a metric-level rebuttal: Claudeforce puts the company inside the agent surface with its permission and governance model attached, and the premium-edition gate means the more agentic a customer becomes, the more they pay. That is a coherent structural answer, and it is the first one offered since the commoditization debate began. It is also entirely unvalidated commercially, since Claudeforce reaches general availability in September and contributes nothing to FY27 guidance. We would describe the disintermediation risk as materially reduced rather than resolved, and we would not pay a full re-rating for it before the Q3 print shows the upgrade cycle converting.

Debate: Is a 25% Re-Rate on a $0.10 Operating Beat Justified?

Bull view: The quarter was never about the operating beat. It was about the removal of a scenario discount that had taken the multiple to roughly 12 times enterprise value to free cash flow, well below both the peer group and the company's own history, on fears that did not materialize. Even after a 25% move the stock trades around 16 times FY27 estimated free cash flow and under 19 times clean operating earnings for a business growing revenue at 11% to 12%, expanding earnings per share at roughly 19% on the retired share count, and about to publish a new multi-year framework. That is not an expensive stock; it is a stock that stopped being distressed.

Bear view: A 22.6% single session is precisely the kind of move that borrows return from the next twelve months. The stock is now discounting an FY28 framework that has not been published, a Claudeforce revenue contribution that has not been guided, and a second-half organic acceleration that has been promised in each of the last five quarters and delivered in none. Meanwhile the same print showed the first year-over-year non-GAAP operating margin decline of the cycle and a full-year cash-flow guide that implies a flat-to-down second half. Buying after the discount closes and before the delivery is the worst point on the curve.

Our take: The bull case has the better valuation arithmetic and the bear case has the better timing argument. At roughly 19 times our clean FY27 operating earnings of approximately $13.71, against an FY28 estimate near $16.30, the stock is not demanding for a business with this cash-conversion profile once the debt-service step-up anniversaries. But the compensation for being early is gone: the same three uncertainties existed at $205.62 and were being paid for, and at $257.56 they are not. We would not chase, and we would not sell. The correct posture is to hold what you own and let the September Investor Day and the November print decide whether to add.

Debate: Was the Debt-Funded Buyback Vindicated or Just Lucky?

Bull view: Management issued $24.8B of debt in the first half and retired roughly a seventh of the company at a $176 average price, against a share price of $257.56 four sessions after this print. That is roughly a 46% gain on the deployed capital in under two quarters, and it permanently improves every per-share metric. Executing that at the bottom of a two-year de-rating, while the market was pricing an existential risk that did not materialize, is the highest-conviction capital-allocation decision the company has made in a decade.

Bear view: The mark-to-market gain is real and the reasoning was not. The stock rose because of a partnership announced three months after the debt was raised, not because of anything management could have known when it levered the balance sheet. What management could control was the cost, and that is permanent: interest expense of $473M a quarter, a full-year cash-flow growth guide cut to 4% to 5%, an implied second half of flat-to-down cash generation, and $27.9B of net debt on a company that carried almost none. Judging a leveraged bet by its outcome after one favorable quarter is exactly the error the decision framework is supposed to prevent.

Our take: We wrote last quarter that the program was net positive but mistimed, and that it would have been better executed in October at potentially lower prices. On the evidence available today that was wrong, and we would rather say so than relitigate it. The bear case is right that the process should not be graded on a single favorable outcome, but the process was not arbitrary: management bought when its own securities traded at a discount to a scenario it believed was mispriced, and it turned out to be mispriced. The permanent cash-flow cost stands, and it is the reason free cash flow growth is guided at 4% to 5% while free cash flow per share grows about 19%, from roughly $15.06 on FY26's 956 million weighted diluted shares to roughly $17.96 on the guided 838 million. That gap closes in FY28 as the interest step-up anniversaries. The residual risk is the remaining authorization: deploying more leverage at $257.56 would be a materially worse trade than the one just executed, and we would prefer to see it left undrawn.

Model Update Needed

Valuation impact. At the September 1 close of $257.56 and the guided FY27 diluted share count of 838 million, market capitalization is approximately $215.8B. Adding net debt of approximately $27.9B ($39,288M noncurrent debt less $8,310M cash and $3,093M marketable securities) gives an enterprise value of roughly $243.7B. That is approximately 16.2 times our FY27 estimated free cash flow of $15.05B, and approximately 18.8 times clean FY27 operating non-GAAP earnings of $13.71. The prior quarter's equivalent multiple at $177.86 was roughly 12 times enterprise value to free cash flow, so essentially the entire move has been multiple rather than estimate.

Our revised twelve-month framework, anchored on FY28 estimated clean operating earnings of approximately $16.30:

The distribution is now roughly symmetric, where a quarter ago it was skewed to the upside. That is the honest consequence of a 25% move in four sessions: the base case offers 11% to 20% against a bear case of 11% to 20% downside, and the case for owning the stock rests on the bull path rather than on the floor.

Thesis Scorecard Post-Earnings

Scored against the standing thesis carried since Q1 FY27, with the same seven bull pillars and six bear points rather than a fresh set.

Overall: The standing thesis strengthened. Six of seven bull pillars are confirmed or strengthened, with the headline reacceleration pillar still mixed and one pillar (margin expansion) downgraded on the gross-margin decline. On the bear side, the single most important structural point of the past year (that cRPO growth was empirically capped and would keep landing on guide) was refuted outright, and the commoditization point moved from unanswered to structurally answered. What did not improve is the number the thesis actually turns on: organic revenue growth decelerated in the period, and the acceleration remains a guided event rather than a delivered one. The thesis is in better shape than it was in May, and the price is 45% higher than the level at which we last recommended sizing in.

Action: Maintaining Outperform with conviction unchanged at moderate. The Q2 print resolved the binary catalyst we set last quarter in the bull's favor, but on the leading indicator only, and the market repriced the entire scenario discount inside one session. Existing holders should maintain: the business is demonstrably not being disintermediated, the buyback has retired a seventh of the company at a 32% discount to the current price, and the multiple at roughly 19 times clean FY27 operating earnings is not demanding for a mid-teens earnings compounder. New positions should not chase this level. The three events that determine whether the base case or the bull case is right are all dated: the Investor Day at Dreamforce on September 16, where the FY28 framework and, we hope, a gross-margin decomposition and a restated Agentforce ARR series arrive; the final settlement of the accelerated share repurchase in October; and the Q3 print in November, which is the first quarter in five where the second-half acceleration can be tested rather than deferred. A Q3 organic print at or above the guided 7.5% with the margin guide intact would justify adding at higher prices than these. A Q3 organic print near 6% would send the scenario discount back where it came from.