Salesforce and the Price-Per-Outcome Problem

For twenty-five years Salesforce billed by the seat: one login, one invoice line, one renewal date. Agentforce bills by the conversation, or by a flat fee per digital worker, and that pricing model has now run through two full quarters of real results. Shares sit around $237.92 as I write this, 11.2% below the 52-week high of $268 and 63% above the low, which is either a fair discount for a business mid-transition or a market that still doesn’t trust the new math.

Second-quarter fiscal 2027 revenue came in at $11.3 billion, up 11% from a year earlier and 2% sequentially. That’s a slow number for a company still calling itself an AI story, but it’s also large enough in absolute terms that 11% growth adds close to a billion dollars of new annual revenue in a single quarter.

The question this quarter was supposed to answer is whether outcome-based pricing adds revenue on top of the seat business or eats into it. The numbers argue it’s doing more adding than eating, for now, though the company hasn’t published the one figure that would settle it cleanly.

Two prices for the same product

Agentforce annual recurring revenue passed $1.5 billion in the quarter, up more than 240% from a year earlier, per Salesforce’s own results release. Add in Data 360, the data and analytics layer that increasingly feeds the agents, and combined ARR reached near $3.9 billion, up more than 210%. Those growth rates don’t exist anywhere else in this business.

Set that against the whole company and the scale gap is obvious: roughly $3.9 billion of combined ARR against $41.5 billion of full-year revenue is still under a tenth of Salesforce. Agentforce is not yet why the top line moves. Core subscription and support, the seat business Salesforce has sold since 1999, is what’s producing the 10% of headline growth, helped along by acquisitions closed over the past year rather than by the AI story alone.

I think about the split the way I’ve thought about Amazon’s retail business sitting next to AWS: two different growth stories wearing one ticker. The Salesforce version is more entangled than that, though, because a customer adopting Agentforce is often trying to cut the number of seats it needs from that same account. Capturing the replacement revenue through agent fees beats losing the account to an AI-native rival with no seat business to protect. But it isn’t free money layered on top of the old model; some of it is defense.

What actually moved the quarter

Net income was $7.5 billion, up from $6.2 billion a year earlier, a jump that outran revenue growth, which points to costs growing slower than sales rather than just a bigger top line. Operating income was $8.9 billion, and the operating margin came in at 21%, with an EBIT margin of 21.5%. Gross margin ticked up to 77.7% from 77.2% a year earlier, a small move but the right direction for a company adding compute-heavy AI workloads to its cost base.

Shares jumped +22.6% on the day of the report, 2026-08-26, nearly three times the stock’s own average earnings-day move of 7.8%. That’s the market telling you expectations had gotten bruised beforehand, not that the quarter itself was extraordinary by Salesforce’s historical standard. A beat against a low bar still moves a stock a lot; it just doesn’t mean the bar was hard to clear.

Sequential growth of 2% from the first quarter is the number I’d flag to anyone modeling this business quarter by quarter. A company growing 11% year over year should show something close to 2 to 3 percent between adjacent quarters if the pace is holding steady. Two straight quarters meaningfully below that would tell you the year-over-year figure is coasting on a soft prior-year comparison rather than current strength, and that’s worth checking again once the easy comparisons roll off in the back half of the fiscal year.

Margins before the mix shifts

Here’s the part I keep coming back to: net margin sits at 18%, and it’s been rising even as Salesforce pours money into agent infrastructure and Data Cloud capacity. If the AI build-out were quietly wrecking profitability, that line would show it first, before management said a word on a call. It hasn’t shown it yet.

That won’t last forever. Every enterprise software company that has tried to bundle a metered AI product into a subscription business has eventually had to explain a margin dip somewhere, usually in the compute line, and Salesforce has not had to yet because Agentforce is still small next to the base it’s built into. Watch this line, not the ARR headline, for the first sign the honeymoon is ending.

A forward multiple that argues with itself

Salesforce trades at 21.8 times trailing earnings of $10.92 per share. Consensus estimates put forward EPS at $7.56, well under the trailing figure, which pushes the forward P/E up to 31.5, above the trailing multiple. That’s an unusual setup: a forward multiple higher than the trailing one only happens when the Street expects earnings to fall, and that’s a strange thing to see from a company that just raised full-year revenue guidance.

I don’t have a clean explanation for that gap, and I’d rather say so than invent one. The likeliest reading is that consensus estimates haven’t fully caught up to a mix shift toward lower-margin, usage-based AI revenue, or that they’re still baked around a prior guidance cut. Either way, it’s the one number in this setup that doesn’t fit the rest of the story, and it’s the uncertainty I’d want resolved before sizing a position around the growth numbers alone.

That’s an honest limit in the data, not a convenient hedge to dodge a call.

The valuation picture splits further depending which multiple you use. On sales, Salesforce trades at 4.9 times trailing revenue against a five-year average of 6.7, meaningfully cheaper than its own history. On book value, it trades at 5.6 against a five-year average of 3.9, meaningfully more expensive than its own history. A stock can’t be cheap on one multiple and expensive on another for free; it usually means the market is repricing something structural about the business, in this case probably the buybacks and acquisitions that have shrunk the share count and thinned the balance sheet relative to revenue.

How this compares against similar names

Against Microsoft at 27.5 times earnings, Salesforce’s 21.8 looks inexpensive for a company still growing revenue in double digits. Against Cintas at 40 times earnings, a business growing far slower but priced for near-certainty, the gap looks even wider. Cheap relative to peers isn’t the same as cheap in absolute terms, though, and the 31.5 forward multiple is the number that would concern me if I were buying today rather than the trailing one everyone quotes.

The Street is not particularly split on direction: 34 analysts carry an average target of $276, implying 16% upside from here, with 76% rating the stock a buy. The high target of $400 and low of $160 bracket a wide range, which tells you the ARR-versus-cannibalization debate hasn’t been settled on Wall Street either. Short interest is a modest 3.5% of the float, so this isn’t a stock the bears have organized around; it’s one the bulls and the skeptics are both still watching. The dividend, a token 0.72% yield on $1.71 paid over the trailing year, isn’t why anyone owns this name.

None of these peer comparisons settle the Agentforce question by themselves. They only tell you where the market currently prices patience for a pricing-model transition, and patience is exactly what this transition is asking for. A cheap multiple against Microsoft or Cintas is a starting point, not a conclusion, if the seat-cannibalization risk I described earlier turns out to be real.

Metric	Value	Context
Price	$237.92	11.2% below 52-week high of $268
Trailing P/E	21.8	forward P/E 31.5, above trailing
Price/sales	4.9	five-year average 6.7
Price/book	5.6	five-year average 3.9
Quarterly revenue	$11.3 billion	up 11% year over year
Net margin	18%	net income $7.5 billion, prior year $6.2 billion
Analyst target	$276	16% upside, 76% buy-rated

My honest read is that Salesforce earns a hold-and-watch stance rather than a strong conviction call in either direction. The core business is compounding at a rate that supports the current multiple without needing Agentforce to work at all, and everything above that is optionality the market hasn’t fully priced because it can’t yet measure it. The concrete case against me: if Salesforce ever reports net revenue per customer falling among accounts that have adopted agents, even while combined ARR keeps climbing, that’s the tell that outcome pricing is cannibalizing faster than it’s adding, and I’d change my stance on the spot.

The number I’m watching into the next print is core subscription growth excluding the AI product lines. If that number holds at 8% or better while combined ARR keeps compounding above 150%, the pricing bet is working as designed. If core growth slips into mid-single digits while AI ARR keeps accelerating, that’s the cannibalization case showing up in the one place spreadsheets can’t hide it.

Analysis and opinion only, not investment advice. Figures come from Salesforce’s second-quarter fiscal 2027 results on its investor site and its filings on SEC EDGAR; valuation multiples are approximate and were checked on September 23, 2026.