Salesforce's 22% Rally: A $2.6 Billion Paper Gain Meets a Real Reacceleration

Salesforce shares are up 22.6% today to about $252 after the company reported a "record" fiscal second quarter and unveiled Claudeforce, an expanded partnership that makes Anthropic's Claude the default reasoning model inside its own products. Same facts, opposite reads. Bulls see the bookings acceleration that finally buries the "SaaSpocalypse" bear thesis. Bears see a $2.6 billion paper gain on the Anthropic stake wearing an earnings beat's clothes — an adjusted EPS number that more than doubled on a mark-to-market, not on operations. At $252, up from a June low near $146, the whole question is which reading the current price already assumes. Here is the shared record first, then the fight.

Shared facts

As of Aug. 27, 2026:

• Price and capital: ~$252 (+22.6% on the day; prior close $205.62), market cap ~$206B, enterprise value ~$234B, 52-week range ~$146–$269.
• Q2 FY27 (ended July 31): revenue $11.35B, up 11%, including ~$456M (about 4 points) from Informatica. cRPO $33.5B, up 14% — held at 14% for a second straight quarter after running ~11% a year ago. Free cash flow $1.1B, up 81%.
• The earnings question: adjusted EPS $5.90 reported; ~$3.37 excluding the Anthropic paper gain, against consensus of $3.27, so the real operating beat was about a dime.
• AI metrics: Agentforce ARR >$1.5B, up 240% (on a definition broadened this quarter to include Slackbot and Headless 360); combined Agentforce and Data 360 ARR ~$3.9B, up 210%.
• Guidance:FY27 revenue raised to $46.1–46.4B (net +$200M: $100M organic, $200M from pending Contentful and Fin acquisitions, −$100M FX); FY27 adjusted EPS raised to $16.67–16.71, up ~$2.60 from the $14.06–14.12 given in May.
• Valuation move: from ~12x forward earnings in early July (stock $166.11, street target ~$246) to ~15–18x on this year's guided and normalized earnings, and ~5.3x trailing EV/sales today.

Round 1: The earnings that doubled were mostly a mark-up

The bull's best punch is the headline: adjusted EPS of $5.90 against $3.27 expected — a 103% year-over-year jump on top of two raised forecasts, free cash flow up 81% and well above the ~$643M StreetAccount consensus, and a raised FY27 guidance. It reads like a demolition.

The bear's answer is that the gap is mostly a paper entry. The $5.90 includes $2.6B from marking up Salesforce's stake in Anthropic. That stake was worth roughly $5B after the lab's latest funding round valued Anthropic at about $965 billion. Subtract the gain and adjusted EPS was ~$3.37, about ten cents over consensus. The full-year guidance raise, from $14.06 to $16.71 a share, is almost exactly the size of the one-time gain; the company's own footnote says it cannot forecast future changes in its investment portfolio. Stripped of the mark-up, fiscal-2027 operating EPS guidance is essentially unchanged. And this gain can reverse: the stake is re-marked every quarter, and a mark-down would hit the same reported line.

The bull concedes the optics. But so does the bear have to concede the rest of the quarter: bookings, cRPO and cash flow — the numbers that do not run through an investment portfolio — all came in ahead. Hiring conviction against the headline is cheap; the honest caveat is that the "earnings crush" investors were celebrating was mostly the appreciation of a stake, not a step-change in operations.

Round 1: the bear wins the earnings headline; the bull wins the operating signals underneath it.

Round 2: Does AI eat seats or feed them?

The bear thesis at full strength: Agentforce is a self-disrupting machine. Every task an agent handles is a human seat that stops billing, and consumption revenue cannot fill the hole fast enough — so the seat-based model decays and bookings decelerate. This is the story that cut Salesforce down 30% in 2025, drew a "Has SaaS outlived its long run?" question from Goldman's Kash Rangan on a 2025 earnings call, drove the stock to a June 22 low that was 45% below its high, and by early July left the market paying only ~12x forward earnings and ~3x sales for a company growing double digits — a price built for a shrinking business.

The bull's case that it is the opposite, all of it from the last two quarters:

• cRPO — revenue already contracted for delivery in the next year — has held 14% growth for two straight quarters after running ~11% a year ago, and the company's own earnings headline says "cRPO growth accelerates."
• Management says net new annual order value grew at its fastest pace in four years, with near-record-low attrition and contract lengths improving.
• Agentforce ARR is past $1.5B, up 240%; combined with Data 360 it is ~$3.9B, up 210%, with Benioff saying it is "about to cross $4 billion."
• Usage is compounding: 3.2B agentic work units in the quarter, up 97% sequentially; Agentforce bookings more than doubled quarter over quarter; 2,000 paying production customers added.
• The most inconvenient fact for the bear: nine of the top ten AI companies use Salesforce or Slack and raised their spending 435% year over year. If AI were about to eliminate the CRM category, the labs building the agents would be the first to stop paying — they are instead the heaviest users.

The bear's answers are narrower but real. The AI ARR bases are tiny — under 4% of a ~$46B revenue stream — and the 240% growth runs on a definition that was widened this quarter, so it is not like-for-like. Monetization is unsettled, with pricing tried per seat, per conversation, and per outcome. And the reacceleration is still not visible in the income statement: Q2 grew 11%, down from 13% in Q1, roughly 7% once Informatica is set aside.

Round 2: the bull wins on direction. The leading indicators — cRPO, bookings, agent usage, and the frontier labs' own spending — point the opposite way from the demand-collapse thesis. The bear is right that the scale is still small and part of the growth is definitional, which is exactly why this round was close.

Round 3: Claudeforce — the hedge that ends the war

Claudeforce, announced with the earnings, makes Claude the default reasoning model inside Agentforce (the Atlas Reasoning Engine, Agentforce Vibes and Agentforce Coworker) and the default model across Slack, while a "Salesforce in Claude" plugin ships 37 prebuilt sales skills. Actions agents take route back through Salesforce's workflows and permissions, and regulated deployments run through Amazon Bedrock inside Salesforce's security boundary. Open beta is September.

The bull reading: this is the strategic exit from the nightmare scenario. The bear's deepest fear was that enterprises would bypass the CRM and talk to a frontier model directly. Instead, Salesforce becomes the data and governance layer agents cannot operate without, Claude stays inside the house, and access is monetized as a premium-edition upgrade. Management's own numbers say customers starting the agentic journey run at double the average order value, with the potential to go three or four times higher.

The bear reading: Salesforce just made its most dangerous potential disruptor the default brain of its own products, with no revenue share disclosed, and lets token spend flow to Anthropic and AWS. Monetizing as "premium edition upgrades" is the old seat license wearing an AI costume; if agents genuinely take over the interface, charging premiums for seats is the exact model the bear says will not scale.

The strategic point goes to the bull — a company the market feared would be replaced found a way to be indispensable to the replacer, and the frontier lab that could have ignored it instead built on top of it. The economics are a draw, because nothing about Claudeforce's pricing or revenue split is public and the beta does not even launch until next month.

Round 3: the bull wins strategy, the bear wins on "show me the economics." Two of three rounds to the bull on substance, with the valuation the real battlefield.

What the price requires now

Six weeks ago this stock was $166.11, the street's average target was ~$246, and Guggenheim's July Buy at $228 argued the AI-doom scenario was "misaligned with reality." Today it trades at $252 — above that average target from six weeks ago. The multiple is now whatever you believe about the gain: ~15x the reported FY27 EPS guidance, ~18x that same guidance stripped of the paper gain, and roughly 30x on the street's forward estimates once the gain is normalized out.

Reverse-engineer the price. For $252 to pay off, three things must be true by the time the reacceleration lands in the income statement: cRPO must hold 14% or better; organic revenue, currently ~7%, must actually reaccelerate as management has promised for the second half; and combined AI plus Data 360 ARR must keep compounding in triple digits while the 34.3% non-GAAP margin holds against token costs. That is a real operating path — but the market started paying the full ticket before the report card shows the result. The "priced for extinction" discount that made the $146–166 trade compelling is gone; what remains is the harder question of whether a ~15–18x multiple on a ~8–9% organic grower is justified by an AI story that is real but still small.

The ruling

The business case goes to the bull. On the evidence that matters — cRPO at 14% for two straight quarters, bookings at a four-year best, AI revenue compounding, and the frontier labs themselves paying 435% more — the SaaSpocalypse thesis is now the weaker side, and Claudeforce removes the "bypassed" tail risk that gave it teeth.

The stock call is more careful. The cheap-insurance trade of June and July has been consumed, the market has adopted the reacceleration before the income statement has shown it, and the earnings number that capped the move is mostly a paper gain on a stake that can mark back down. The burden of proof now sits with the bull's reacceleration, not with the bear's extinction.

The ruling flips if, at the Q3 FY27 report in late November or December, cRPO slips below ~13% constant currency, organic revenue has not reaccelerated in the reported numbers, AI and Data 360 ARR stops compounding in triple digits, or the Anthropic stake marks down and takes the EPS optics with it. The earliest bear confirmation would be Q3 guidance with cRPO below 14%. The 22% day proved the demand-collapse half of the bear thesis wrong. Proving the bull's reacceleration — and paying for it at this price — is the next quarter's job.