Salesforce CRM

A note on the calendar, because everything below depends on it. Salesforce's fiscal year ends 31 January, so the fiscal-year label runs a year ahead of the calendar. Fiscal 2026 is the year ENDED 2026-01-31. Fiscal 2027 is the year ending 2027-01-31, and the most recent reported quarter — Q1 fiscal 2027 — is the three months ended 2026-04-30. Where this dive says "FY2027E" it means the year ending 2027-01-31, eleven months away, not calendar 2027.

Reference table

1. What the business is — one segment, six product lines that just became two

Salesforce operates as ONE reportable segment. Both the 10-K and the Q1 fiscal-2027 10-Q state it plainly: "We operate as one segment." Everything the vendor file calls seg_prod is therefore a revenue disaggregation, not a segment note — a distinction that matters for how much weight it can bear.

The profile describes a global leader in customer relationship management technology, 83,334 employees at 2026-01-31 (10-K, matching the vendor), chief executive Marc R. Benioff, listed on the NYSE since 2004-06-23.

Revenue by service offering, fiscal 2026 (year ended 2026-01-31) — vendor seg_prod, which sums EXACTLY to reported revenue:

The sum ties to the reported $41.525B to the dollar — no double-count, no missing sub-segment. That is worth saying because it is the vendor failure mode this batch was told to hunt for, and here it does not occur.

Three readings. First, the two largest and oldest lines — Service Cloud and Sales Cloud, 45% of revenue between them — grow at 8.4% and 8.5%. That is the core seat business, and it is growing at roughly nominal-GDP-plus. Second, Salesforce Platform and Other at +22.6% is the only fast line, and the 10-K discloses that it includes $388M of Informatica subscription revenue — strip that and the line grows about 17%. Third, Professional Services is shrinking, which the 10-K attributes to "less demand for larger, multi-year transformation engagements, which may continue in the near term" — a soft but real datapoint on enterprise appetite for big software projects.

The labels are already obsolete, twice over. The 10-K discloses that "In the third quarter of fiscal 2026, we renamed our service offerings to reference Agentforce. There were no changes in the allocation of revenue between these service offerings as a result of this change." And the Q1 fiscal-2027 10-Q goes further:

> "Effective as of the first quarter of fiscal year 2027, we have revised the presentation of our disaggregated revenue disclosures to reflect the evolution of our product architecture to deliver the Agentic Enterprise... subscription and support revenue is now reported across two primary categories: Agentforce Apps and Data 360, Headless Platform, and Other."

Agentforce Apps comprises Agentforce Sales, Service, Marketing, Commerce, Agentforce Apps Flex Credits and Slack. Data 360, Headless Platform, and Other comprises Data 360, Data 360 and Platform Flex Credits, Headless Platform, Informatica, Agentforce Mulesoft, Agentforce Tableau and Other. The vendor file carries neither presentation — it carries the pre-rename six-line split and no fiscal-2027 data at all. Flagged in Section 8.

Revenue by geography, fiscal 2026 (seg_geo, also summing exactly to $41.525B): Americas $27.193B (65.5%, +8.2%), Europe $10.017B (24.1%, +12.7%), Asia Pacific $4.315B (10.4%, +11.8%). Europe and Asia Pacific are growing faster than the Americas — but revenue is booked "based on the region of the Salesforce contracting entity, which may be different than the region of the customer" (10-Q), and total revenue was helped by about one percentage point of currency in fiscal 2026 and two points in Q1 fiscal 2027. The geographic mix shift is therefore partly an exchange-rate artefact.

Concentration and structure. The 10-K: "The number of paying subscriptions at each of our customers ranges from one to hundreds of thousands. None of our customers accounted for more than ten percent of our revenues in fiscal years 2026, 2025 or 2024." Assets located outside the Americas were 16% of the total. Subscription and support was 95% of revenue in fiscal 2026 and in Q1 fiscal 2027; term software licences, recognised at a point in time rather than ratably, are about six percent of subscription revenue. That six percent is the only lumpy piece.

Non-equity tripwire — checked and passed. Common stock on the NYSE, IPO 2004-06-23, price $190.99 (nowhere near a $25 or $1,000 par-like level), beta 1.178, a variable and recently initiated dividend ($1.712 trailing, 0.90%), volume 6.76M shares, and a 52-week band of $150.12 to $266.23 — a 77% range. This is ordinary common equity, not a preferred or a baby bond inheriting parent fundamentals.

2. The single most important thing to understand about these numbers — you are looking at two different earnings figures, and the gap is $5.22 a share

Every forward multiple published on Salesforce is a non-GAAP multiple, and the difference is not cosmetic.

The proof is in this file. Sum the four most recent quarterly actuals from earn_cal — the figures against which "beats" are measured — and compare them with the reported GAAP diluted EPS in inc_q for exactly the same quarters:

Trailing GAAP diluted EPS is $8.63. Trailing non-GAAP EPS is $13.85. GAAP is 62.3% of non-GAAP. The same relationship holds for the full fiscal year: reported GAAP diluted EPS of $7.80 for fiscal 2026 against $12.55 of summed non-GAAP quarterly actuals.

The bridge is mostly stock-based compensation, and stock-based compensation is a real cost. Fiscal 2026 SBC was $3.509B — 8.3% of revenue (stockBasedCompensationToRevenueTTM 0.0829, independently confirming). On 819.0M shares that is $4.28 per share, pre-tax, every year. Add the fiscal 2026 restructuring charge of $586M and amortisation of acquired intangibles (not separately extractable from this filing text, but consistent with $6.815B of intangibles and $3.631B of total depreciation and amortisation) and the pre-tax add-backs plausibly total $5.9B to $6.3B — which, taxed at the effective rate of roughly 22%, matches the observed $4.75 per-share gap for fiscal 2026 closely enough to confirm the composition.

What this means for every multiple in this dive, stated once:

• Forward multiples are on the NON-GAAP basis, because that is the basis the consensus estimates are built on. est.epsAvg of $11.776 for fiscal 2026 against a reported GAAP $7.80 and an actual non-GAAP $12.55 settles it beyond argument.
• At $190.99: 13.5x FY2027E ($14.155, 32 analysts) and 12.2x FY2028E ($15.627, 32 analysts).
• The GAAP equivalents are roughly 21.7x and 19.6x if the 62.3% conversion holds. Trailing GAAP is 22.1x.
• When this dive quotes a free-cash-flow multiple it charges SBC as a cost, because a company issuing $3.5B of stock a year and spending $12.6B buying it back is not generating $14.4B of shareholder cash. It is generating about $11.15B.

And the buyback proves the point. Weighted diluted shares went 984M (FY2024) → 974M (FY2025) → 956M (FY2026) — down 18M, or 1.8%, in a year in which the company repurchased approximately 50 million shares for approximately $12.7 billion (10-K). Roughly 32 million shares were reissued into equity plans. About two-thirds of a $12.6B buyback went to standing still. State it plainly: on a weighted-average basis, the majority of fiscal 2026's repurchase spend offset compensation dilution rather than shrinking the company.

That changed in March 2026, and Section 5 deals with how.

3. Growth and margins — the forward indicator is remaining performance obligation, and it is decelerating

Reported revenue is the least informative number Salesforce publishes. It is ratable recognition of contracts signed one to five years ago. The forward indicator is remaining performance obligation — all future revenue under contract not yet recognised — and specifically current RPO, the portion expected to be recognised within twelve months.

The full disclosed series from the filings:

Read it constant-currency and the picture flattens: roughly +13% at both of the last two balance dates. Read it ex-acquisition and it is lower still — Informatica closed in November 2025 and is inside both of those figures. Its contribution to cRPO is not disclosed; on the basis that it added ~$444M of revenue in the April quarter, a contribution of $1.5-2.0B to a $29.5B prior-year cRPO base would be roughly 5-7 points. That would put organic constant-currency current-RPO growth somewhere near 7-8%. That is our arithmetic on undisclosed inputs and is labelled as such — but it is the same answer the revenue line gives.

The revenue decomposition, which is disclosed:

A headline of 13.3% is an organic constant-currency rate of roughly 7%. The 10-K states Informatica contributed approximately $399M of revenue in fiscal 2026 (essentially all in Q4, since it closed in November) and the 10-Q states $444M in Q1 fiscal 2027 and "approximately two percent" of currency benefit. This is the most important single adjustment in the dive and it is built entirely from disclosed figures.

Where the growth comes from is disclosed too, and it is not price. The 10-Q: "The increase in subscription and support revenues for the three months ended April 30, 2026 was primarily caused by volume-driven increases from new business, which includes new customers, upgrades and additional subscriptions from existing customers. Pricing was not a significant driver of the increase in revenues for the period." The identical sentence appears in the 10-K for fiscal 2026. Salesforce is not raising prices. It is selling more seats. Which is exactly the variable the bear case says is about to stop.

And on attrition, the one sentence that most directly addresses the bear case:

> "We continue to maintain a variety of customer programs and initiatives, which, along with increasing enterprise adoption, have helped keep our attrition rate consistent as compared to the prior year. Consistent attrition rates play a role in our ability to maintain growth in our subscription and support revenues." — 10-Q, quarter ended 2026-04-30

No dollar-based net retention figure is disclosed anywhere in this file. "Consistent" is a word, not a number, and it is management's word — half-weighted accordingly. But it is the only attrition disclosure that exists, and it does not corroborate the bear case as of 2026-04-30.

Margins — real expansion, at half the vendor's rate.

The vendor excludes restructuring from fiscal 2026 operating income and includes it in fiscal 2025. The 10-K is unambiguous: "income from operations was $8.3 billion... Operating margin... increased to approximately 20 percent for fiscal 2026 compared to approximately 19 percent in the prior year period." The filing wins: margin expansion is about 100 basis points a year, not 250. Full treatment in Section 8, item 4.

The company's stated intent is more of the same: "We expect to continue to experience improvements in our operating expenses as a percentage of revenue", with research and development "consistent as a percentage of revenue", and sales and marketing and general and administrative "decrease as a percentage of revenues over time... which includes the use of AI and agents." Management guidance, half-weighted — but the direction is corroborated by three consecutive filing-basis datapoints. Restructuring charges have fallen $988M → $461M → $586M across fiscal 2024-2026, with $220-250M of future cash payments still expected, and the 10-Q states "We do not expect to incur significant additional charges in connection with our restructuring initiatives in the near term."

4. Balance sheet — the vendor's is one quarter stale, and that quarter changed the company

This is the largest data-integrity finding in the dive, and it is not a rounding error.

The vendor's most recent balance sheet is 2026-01-31:

• Cash $7.327B · short-term investments $2.238B · total $9.565B
• Short-term debt $4.000B · long-term debt $10.439B · finance leases $2.737B · total debt $17.176B
• Total assets $112.305B · total liabilities $53.163B · total equity $59.142B
• Goodwill $57.941B · intangibles $6.815B — 57.7% of total assets

Twelve weeks later, per the 10-Q for the quarter ended 2026-04-30, it looks nothing like that:

> "As of April 30, 2026, we had senior unsecured debt outstanding, with maturities starting in March 2028 and extending through March 2066, with a total carrying value of $33.3 billion... In March 2026, we also issued unsecured Senior Notes with an aggregate principal of $25.0 billion... We used the net proceeds from the March 2026 Notes to fund an accelerated share repurchase program... As of April 30, 2026, the full $6.0 billion was outstanding under the 2026 Term Loan Credit Agreement... our principal sources of liquidity were cash, cash equivalents and marketable securities totaling $11.8 billion and accounts receivable of $5.1 billion."

Borrowings at 2026-04-30: $33.3B of senior notes plus $6.0B of term loan = $39.3B, with nothing drawn on the $5.0B revolving facility. The 10-Q's balance-sheet metrics table carries the note "Amounts do not include operating or financing lease obligations", so finance leases sit on top; the last disclosed figure is $2.737B at 2026-01-31.

The corrected enterprise value, rebuilt from the filing:

The vendor's EV is $2.8B high — about 1.5% — and the direction of the error is the familiar one: marketable securities excluded from the net-debt calculation. The same defect is provable at fiscal year end, where netDebt of $9.849B is exactly total debt ($17.176B) less cash ($7.327B), silently discarding $2.238B of short-term investments. The correct fiscal-2026 figure is $7.611B. We use $30.2B of net debt and $186.6B of enterprise value throughout.

The corrected leverage picture, and it is not trivial. Net debt of ~$30.2B against trailing free cash flow of ~$14.66B is about 2.1x, or 2.7x on free cash flow after charging stock compensation ($11.15B). Trailing interest coverage of 20.8x is measured almost entirely before the March 2026 notes existed — fiscal 2026 interest expense was only $324M on the old $14.4B of borrowings, and the 10-Q states: "We expect this debt offering to cause interest expense to increase as compared to prior year throughout fiscal 2027." The coupons on the March 2026 notes are not in the extracted filing text; at a blended 5% the incremental cost would be roughly $1.25B a year pre-tax. That is scenario arithmetic on an undisclosed input and is labelled as such.

Two structural features that are correct and should not be mistaken for errors. First, tangible book value per share is NEGATIVE $36.53 and tangibleAssetValueTTM is −$31.706B — arithmetically right, because goodwill and intangibles are 61.8% of assets on the trailing basis while the accelerated repurchase cut equity from $59.142B to roughly $33B. Book-value valuation is unusable for this name and no price-to-book figure appears in this dive. Second, the current ratio of 0.786 and negative working capital of −$5.889B are normal for enterprise software: unearned revenue is a current liability that will be settled by delivering software, not cash. Neither is a liquidity signal.

5. Cash flow and capital returns — a self-funding machine that stopped self-funding in March

The operating business generates cash prodigiously and consumes almost no capital.

Operating cash flow has more than doubled in three years while capital expenditure has FALLEN in every one of them — capex is now 1.4% of revenue, and operating cash flow is 36.1% of it. On any conventional reading this is an outstanding cash business. Trailing free-cash-flow yield is 9.4% of market capitalisation.

Charge stock compensation as a cost and the number changes materially. Trailing free cash flow of ~$14.66B less SBC of $3.509B is $11.15B — a 7.1% yield on market capitalisation and 6.0% on the corrected $186.6B enterprise value, i.e. ~16.7x EV/FCF. That 16.7x is the honest multiple, and it is the number Section 6 uses.

Fiscal 2026 was the last self-funded year. Buybacks plus dividends of $14.183B were 98.5% of free cash flow — but adding $9.268B of acquisitions took total deployment to $23.45B against $14.4B of cash generation, funded by $6.0B of Informatica borrowings and a $4.5B drawdown of cash and investments (total cash and short-term investments fell from $14.032B to $9.565B).

Then the March 2026 quarter, which is not in the vendor's balance sheet at all:

> "The net cash used in financing activities during the three months ended April 30, 2026 was primarily related to proceeds from the issuance of debt, net of issuance costs of $24.8 billion... partially offset by repurchases of common stock of $27.2 billion, which includes our March 2026 accelerated share repurchase, and payments of dividends and equivalents of $365 million."

> "In February 2026, the Board authorized $50.0 billion in share repurchases... In March 2026, we entered into the ASR Agreements... to repurchase an aggregate of $25.0 billion of our common stock and received an initial delivery of approximately 103 million shares at an average price per share of $198.34, which represents approximately 80 percent of the total shares expected to be repurchased... As of April 30, 2026, we were authorized to purchase a remaining $22.9 billion."

Read the prices. Fiscal 2025: 30 million shares for approximately $7.8 billion — ~$260 a share. Fiscal 2026: approximately 50 million shares for approximately $12.7 billion — ~$254 a share. March 2026 accelerated repurchase: 103 million shares at $198.34. The stock closed today at $190.99. Roughly $45 billion of repurchases over twenty-seven months, every tranche of it above the current price. That is not a scandal — buybacks are struck against an unknowable future — but it is the plainest available evidence on management's own valuation judgement, and it should be weighed against the fact that the company is now borrowing to do it.

Is the levered buyback accretive? Only clearly on the adjusted numbers. The $25B accelerated repurchase will ultimately retire roughly 126 million shares (103M delivered, being about 80% of the expected total; the remainder settles in the second half of fiscal 2027). Those shares would have carried roughly $1.75B of trailing non-GAAP earnings ($13.85 × 126M) or roughly $1.09B of trailing GAAP earnings ($8.63 × 126M). The financing cost, at a blended 5% on $25B, is about $1.25B pre-tax and roughly $0.98B after tax. Strongly accretive on non-GAAP. Roughly break-even on GAAP. (Our arithmetic; the coupon is not disclosed in the extracted filing text.) A company whose largest capital-allocation act only clearly works on the adjusted earnings basis is a company to watch rather than to underwrite — and it is a direct consequence of the $5.22 per-share gap in Section 2.

The dividend is small, real and growing slowly. $1.587B in fiscal 2026 against $1.537B in fiscal 2025; $365M in the April 2026 quarter against $402M a year earlier — lower in dollars because there are fewer shares. dividendPerShareTTM is $1.712, a 0.90% yield, and a 19.3% payout of GAAP earnings.

6. Valuation — priced in or room?

At $190.99 (market capitalisation $156.42B, ~819.0M shares, corrected enterprise value ~$186.6B):

Analyst counts in brackets. GAAP forward figures are derived by applying the observed fiscal-2026 GAAP-to-non-GAAP conversion of 62.3% and are labelled as derived, not as a consensus figure.

Estimate coverage is excellent through FY2028 and thin beyond. 32 analysts on both FY2027E and FY2028E EPS, 38 and 39 on revenue. Then it falls off a cliff: 12 analysts on FY2029E EPS, 6 on FY2030E, 11 on FY2031E — and the FY2031E revenue of $70.505B implies +16.5% growth, faster than anything the company has produced in three years. No conclusion in this dive rests on any row beyond FY2028E. The niAvg row is also internally inconsistent past FY2028 (FY2029E net income of $14.827B is BELOW FY2028E's $15.011B while EPS rises from $15.63 to $17.95) — flagged in Section 8.

est.ebitdaAvg and est.ebitAvg are rejected outright. The FY2026E rows show EBITDA of $19.288B and EBIT of $17.192B against a reported GAAP operating income of $8.331B and an EBITDA of roughly $11.96B (operating income plus $3.631B of depreciation and amortisation). Even adding back the entire $3.509B of stock compensation reaches only $15.47B. No EBITDA multiple appears anywhere in this dive, and evToEBITDATTM (13.527x), enterpriseValueMultipleTTM (13.527x) and netDebtToEBITDATTM (2.354x) are all rejected with them.

Peer context, used carefully. The vendor peer set is half irrelevant (Micron, Lam Research, Uber). The genuine comparables are SAP ($195.57, $227.9B), IBM ($235.11, $221.5B), ServiceNow ($118.13, $122.1B), Intuit ($323.60, $88.5B) and, from this batch's own table, Adobe ($257.64, $102.4B). No peer multiples are supplied in the file, so no peer-multiple comparison is drawn. The one cross-reference that is genuinely informative is Adobe's price-target consensus of $261.16 against a $257.64 price — +1.4% — versus Salesforce's +39.1%. Same sector, same narrative, same trading day. That contrast is used in 6c and Section 8.

The vendor's own model rates Salesforce B+ with an overall score of 3, its worst sub-scores being debt-to-equity (1) and price-to-book (2) — both of which are consequences of the accelerated repurchase rather than of the operating business, and the price-to-book one is meaningless given negative tangible book.

6a. What today's price assumes (the inversion)

At $190.99 — 13.5x FY2027E and 12.2x FY2028E consensus non-GAAP EPS, and ~16.7x corrected enterprise value to free cash flow after charging stock compensation — the price embeds roughly the following falsifiable claims:

• Revenue compounds about 11% then 9% through the years ending 2027-01-31 and 2028-01-31. (Consensus-derived: $41.525B actual → $46.108B → $50.428B, on 38 and 39 analysts.) Our number is that organic constant-currency growth is running near 7%, so consensus needs the acquisition contribution to persist and currency not to reverse. It is not assuming a re-acceleration; it is assuming no further deceleration. Falsifiable at every print.
• Operating margin keeps expanding about 100 basis points a year, and the share count keeps falling. (Filing-derived: ~19% FY2025 → ~20% FY2026 → ~21% in Q1 FY2027. Share count: 984M → 974M → 956M weighted diluted, then 871M in Q1 FY2027, then ~819M today.) Consensus FY2028E EPS of $15.627 on revenue of $50.428B implies a non-GAAP net margin near 25%, which needs both legs.
• The market keeps paying ~13x forward non-GAAP. This is the assumption doing the most work and receiving the least scrutiny. 13.5x forward is roughly a market multiple. It already embeds a very large de-rating — the knowledge base's own most-cited valuation claim (all_in, 2026-02-07, conviction 80) put the reset at "~30x to ~15-17x FCF", and on our corrected enterprise value with stock compensation charged, Salesforce is at ~16.7x today. The price is not assuming terminal decline. It is assuming maturity.
• Current remaining performance obligation growth stays in double digits. (Filing-derived: +11% at 2025-10-31, +16% at 2026-01-31, +14% at 2026-04-30, with 3 and 1 points of currency in the last two and Informatica inside both.) This is the most fragile assumption in the price, because cRPO is the earliest place seat erosion would appear and because it is reported on 2026-09-02.
• Interest expense does not swamp the buyback, and no further debt-funded acquisition arrives. (Filing-derived: borrowings from $14.4B to $39.3B in one quarter; fiscal 2026 interest expense $324M; $22.9B of authorisation remaining.) The balance sheet has already been spent once.

Most fragile: current remaining performance obligation, reported 2026-09-02. Everything else in the list can drift for a year without breaking anything. That one cannot.

6b. The return bridge (why the multiple moves)

Expected return over twelve months decomposes as non-GAAP EPS growth (+10.4%, consensus FY2027E $14.155 → FY2028E $15.627) + multiple drift (we assume a HOLD at ~13.4x, neither expansion nor compression) + shareholder yield (0.90% dividend only).

Our base of $210 assumes NO re-rating whatsoever. It is 13.4x FY2028E non-GAAP EPS against 13.5x FY2027E today — the same multiple, rolled forward one year. All of the expected return is earnings growth. We do not count the buyback as yield, because it is now financed with debt rather than free cash flow, and a repurchase funded by borrowing transfers value between capital providers rather than creating it.

We assume a hold rather than expansion deliberately, and it is the most consequential judgement in this dive. The street's $265.75 requires 17.0x FY2028E — a 26% multiple expansion from here — and nothing in the published target set explains why the multiple should expand on a business whose organic growth is decelerating. Our bull case of $275 is essentially the street's case, held at low probability. Our bear of $130 is ~8.3x FY2028E — compression to the "Ford multiple" the knowledge base names repeatedly — on a modestly missed number.

If you need multiple expansion for this to work, you are betting on sentiment reversing on the single most crowded negative narrative in software. We are not making that bet, which is why the base case is only +10%, and why +10% is not enough.

6c. Variant perception (where we differ, what would surprise)

• We differ from the street on the target, not the direction — and we think the targets are stale. The consensus is $265.75 (+39.1%) with a low of $215, still 12.6% above spot. Not one covering analyst out of a rating set of 97 has a target below the current price, on a stock down 23.8% over twelve months and 28.3% below its high. The controlled comparison sits inside this batch: Adobe, the same AI-disruption narrative, closed at $257.64 against a $261.16 consensus — +1.4%. Adobe's street has marked to the new world; Salesforce's has not. Watchable event: whether targets are cut in the week after 2026-09-02 regardless of the result. A wave of cuts on an in-line print is confirmation that the gap was staleness, not opportunity.
• We differ from the knowledge base on how much is already in the price — their thesis is right and their trade is late. The most valuation-specific bear claim in the lane called for terminal multiples resetting from "~30x to ~15-17x FCF" (2026-02-07, conviction 80). On our corrected enterprise value and free cash flow after stock compensation, Salesforce trades at ~16.7x — inside that band. Another (2026-02-28, conviction 40) observed the stock had "rerated to a ~10 PE like Ford." The de-rating the bears demanded has happened. From here, the bear case must be delivered by earnings, not by the multiple — and no bear claim in the file forecasts an earnings decline with a date attached.
• We differ from the bears on what the filings currently show. Seat erosion is not visible: growth was "volume-driven increases from new business" with "pricing not a significant driver"; attrition is "consistent as compared to the prior year"; current RPO is +14%. None of this is proof of safety — these are multi-year ratable contracts and damage would surface late — but it is the only evidence in existence, and it does not corroborate the bear case as of 2026-04-30. Watchable number: cRPO growth on 2026-09-02.
• We agree with the bears, and disagree with the price, on organic growth. A headline of +13.3% that is ~7% underneath, achieved partly by paying $9.6B for Informatica, is not the company the FY2029-FY2031 estimate rows describe. Watchable number: the quarter ending 2027-01-31, when Informatica laps and the reported rate steps down mechanically.
• Positive surprise that would force a re-rate: Salesforce sizing its consumption revenue. The plumbing exists and is named in the accounts — Agentforce Apps Flex Credits and Data 360 and Platform Flex Credits — and the new two-category disclosure makes introducing a third line easy. Nowhere in this file is that revenue quantified. A disclosed, fast-growing consumption line would answer the terminal-value question directly and is the single most probable trigger for a genuine re-rating.
• Negative surprises that would break it: cRPO growth below ~10%; any disclosure of declining seat or subscription counts; a change from the word "consistent" on attrition; or another large debt-funded acquisition on a balance sheet that has already been levered once.

Synthos fair values

All three anchors are multiples of FY2028E consensus non-GAAP EPS of $15.627 (32 analysts, year ending 2028-01-31), cross-checked against corrected enterprise value on revenue and on free cash flow after charging stock compensation.

• Bear ~$130 — 8.3x FY2028E (or ~9.2x a number missed by 10%). Cross-checks: market capitalisation $106.5B, enterprise value ~$136.7B, 2.7x FY2028E revenue and ~12.3x trailing free cash flow after stock compensation — the zombie multiple the knowledge base repeatedly names. The scenario: cRPO growth breaks into single digits, organic growth decays toward 4-5%, and the multiple compresses as terminal value is written down. −31.9%.
• Base ~$210 — 13.4x FY2028E. Cross-checks: enterprise value ~$202.2B, 4.0x FY2028E revenue, and roughly 15x enterprise value to FY2028E free cash flow after stock compensation if that cash flow compounds at the consensus earnings rate. This requires no re-rating at all — it is today's 13.5x multiple, rolled forward one year, on consensus earnings. The scenario: the business does what the last three prints say it is doing. +10.0%, plus a 0.90% dividend.
• Bull ~$275 — 17.6x FY2028E. Cross-checks: enterprise value ~$255.4B, 5.1x FY2028E revenue; $9 above the street's $265.75 consensus and $50 below its $325 high. The scenario: consumption revenue is disclosed and growing fast, cRPO re-accelerates above 15% constant-currency ex-acquisition, and the multiple expands toward the low twenties as the terminal-value discount unwinds. This is essentially the street's case, and it is the fragile leg. +44.0%.

Base is 10.0% above spot; asymmetry roughly 1.4:1 to the upside (32% down, 44% up). A 1.4:1 payoff on a name whose deciding disclosure arrives in 29 days is a wait, not a buy.

7. Knowledge base — 120 raw hits, 89 used, 31 discarded, and the worst false-positive problem in the batch

The triage is itself a deliverable here, so it is reported before any claim is quoted.

The eighteen category collisions, named so the discipline is checkable. The worst is a 2025-11-24 claim about Solana lacking "a paid salesforce to onboard merchants" — the literal common noun, nothing to do with this company. Others: a 2020 claim about IQVIA that mentions a Salesforce Health Cloud partnership; a 2022 claim about a "healthcare CRM" as a product category; CrowdStrike described as "the Salesforce of security"; an NFT-loyalty claim about what "email/CRM" cannot do; a "4-person CRM build" used as an engineering-velocity example; "building your own CRM/AI is nonsensical TCO"; agents deployed "atop existing CRM/telephony stacks"; a claim about SAP's moat that names ERP/CRM as categories; a claim about Sierra as an "agent-native Salesforce"; a claim about Adobe that uses Salesforce as a comparator; and five Bitcoin or macro claims that invoke "the Google/Salesforce IPO era" as a historical analogy. Every one of these would have been counted as a Salesforce claim by a naive search, and every one would have been noise.

The eight attribution-quarantined discards matter more. These carry thinker: quarantine_misattributed, which is our internal flag that the claim's speaker could not be safely resolved from the channel. They are almost all bearish and, if admitted, would make an already negative lane look overwhelming. They are excluded on principle: an unattributable claim cannot be weighted, regardless of which way it points.

One subtlety in the stance field that changes the count. The stored stance describes the claim, not the claim's view of Salesforce. Four claims tagged bullish are long-Chevron-short-Salesforce pair trades — bullish on the trade, bearish on this security. We reclassify them. After that, the 89 usable claims split roughly 52 bearish / 27 bullish / 10 neutral. Net stance: negative-high.

And one honest caveat on breadth. Seventeen distinct sources is wide. But roughly 31 of the 89 usable claims come from a single speaker, Jordi Visser (skill 2.0, the highest-weighted thinker touching this name), repeated across many appearances between September 2025 and August 2026. The lane is loud, not independent. Weight accordingly.

The bear lane

Jim Bianco — the claim this dive was specifically directed to treat properly. (Entity match on CRM/NOW/MSFT, bearish, conviction 60, skill 0.8, guest attribution.)

> "Software companies are cratering because AI will replace corporate software spend on Office, Windows, CRM (Salesforce/ServiceNow) — you'll still need them a couple years, but that's where AI revenue comes from."

Stated honestly: the date field on this record is EMPTY. It is a named, entity-matched, testable claim from an independent voice, and it carries the useful qualifier "a couple years" — a bear who concedes the revenue does not vanish immediately. But an undated claim cannot be graded on a timeline, and that is a defect in our own store, not in the source. It is used as a directional input and excluded from any dated test.

Gavin Baker is the most rigorously formed bear claim in the file, because it names its own test:

> 2024-11-01 · bearish · conviction 75 · entities CRM, WDAY · speaker Gavin Baker, independent

> "Application SAS is done — per-seat software like Salesforce and Workday are bets against ASI; as AI replaces human labor, per-seat (per-human) revenues shrink."

> Stored test: "Per-seat application-software revenues shrink as AI reduces headcount."

That claim is now twenty-one months old, and on its own test it has not yet come true. Salesforce revenue over that period went from $9.444B in the quarter ended 2024-10-31 to $11.133B in the quarter ended 2026-04-30, and current RPO from an implied ~$26.5B to $33.6B. This is exactly the kind of dated, named, falsifiable claim the scorecard exists to grade, and as of today the numbers have not delivered it. That is not a refutation — the horizon is thesis, not 12 months — but it belongs on the record.

Jordi Visser, the dominant voice, at his two most specific:

> 2026-02-07 · bearish · conviction 85 · entities CRM, ADBE, MSFT, IGV · speaker jordi_visser (guest on the Pompliano channel — attributed to the SPEAKER, channel noted separately per house policy)

> "Enterprise SAS (Salesforce, Adobe, Microsoft) is directly at odds with AI-native agents; multiples must be re-rated down like Macy's — companies survive but stop growing with nominal GDP."

> Stored test: "Enterprise-software names (Salesforce, Adobe) should NOT bounce significantly; their earnings/revenues decouple from nominal GDP."

> 2026-07-19 · bearish · conviction 78 · entities ADBE, CRM

> "Adobe and Salesforce have downward-sloping 200-day moving averages — 'these stocks are toast'; may bounce to the 200-day but must pivot their business, disrupted by AI agents."

> Stored test: "200-day moving averages keep declining; no durable recovery without business pivot."

Both tests are live right now and one of them is remarkably precise. The stock has bounced 27.2% off its June low — which contradicts "should NOT bounce significantly" — and it now sits 6.6% below a 200-day moving average of $204.56 that is indeed still declining, which is exactly the level his July claim named as the natural stopping point. A bear who tells you where the bounce ends is a bear worth reading, and $204.56 is now a tracked level in this dive's kill criteria.

Chamath Palihapitiya supplies the only quantified damage estimate in the entire lane:

> 2026-04-24 · bearish · conviction 80 · entities CRM, NOW, SNOW, ADBE

> "Agents got so cheap enterprises spin up bespoke alternatives to vertical SaaS; unit cost is ~30% of value, so customers cut renewals 50-75% and cash flows become unpredictable."

A 50-75% renewal cut would be visible in current RPO within two quarters. It is not visible. cRPO grew 14% in the most recent print. This is the most testable bear claim in the file and the filings currently contradict it.

Two further named bears narrow the case. Bill Ackman (2026-06-03, bearish, conviction 55): "Worries more about Salesforce than others; niche software charging monopolistic prices (~$30k/yr) is really at risk if not fully AI-enabled, unlike low-per-seat platforms like Microsoft." Josh Brown supplies the most recent macro datapoint (2026-07-14, bearish, conviction 62, stored test "enterprise SaaS spending stays weak through summer 2026"): "IBM's pre-announcement was a warning shot: enterprise SaaS spending is not a corporate priority and probably won't be for the rest of the summer as capex shifts elsewhere."

And the most precisely reasoned bear claim of all, which is also the mildest:

> 2026-03-06 · neutral · conviction 48 · entity CRM · speaker "Alex", guest, a16z

> "Nobody eliminates Salesforce's back end, but because its front end is divorced from the back end, seat counts may shrink as people build custom AI frontends on the data."

That is the analytical core of this dive stated in one sentence by someone else. Not death; erosion of the unit of sale while the data position stays intact. It is the outcome that produces dead money rather than disaster, and it is the outcome our base case of +10% roughly describes.

Two further bear claims are paraphrased rather than quoted, because their stored text names specific AI vendors and this house does not print AI model or laboratory names. Alex Kantrowitz (2026-07-31, bearish, convictions 35 and 40, guest) argues that the real threat is not customers building their own systems — which he calls ridiculous — but a frontier model developer offering an agentic customer-relationship product over a simple database, in which case "it's unclear what Salesforce is worth." Note the low conviction on the most sophisticated bear claim in the file: 35 and 40. A 2026-01-31 Visser claim naming a specific assistant product is paraphrased on the same basis.

The bull lane

Steven Sinofsky is the strongest recent bull claim, and it is a claim about architecture rather than about the stock:

> 2026-07-07 · bullish · conviction 85 · entities SAP, CRM · speaker Steven Sinofsky, a16z

> "Legacy enterprise software that codifies a company's business rules (SAP, insurance systems, Stripe) is 'not going anywhere ever'; the SaaS-death-by-AI-agents narrative is wrong."

Jensen Huang makes the direct expansion argument — and is discounted as a promoter-class voice for AI adoption generally (2026-02-26, bullish, conviction 80): "Market wrongly fears a SaaS apocalypse; agents are tool users, so tool/system-of-record usage rises as enterprises deploy agents."

The switching-cost argument, from a source otherwise neutral on the name (invest_like_the_best, 2026-01-29 at conviction 80 and 2026-02-24): "To displace incumbents (Salesforce), AI-native companies must spend 1-2 years building migration tooling; customers won't move data without seamless migration", and "systems of record (ERP/CRM) with strong distribution stay hard to displace, like Walmart surviving e-commerce."

And the bear's own bull case, which is the most interesting claim in the file precisely because of who made it:

> 2026-02-28 · neutral · conviction 40 · entity CRM · speaker jordi_visser

> "Salesforce rerated to a ~10 PE like Ford (assumed no growth); if you believe it grows earnings and embraces AI agents over two years, there's a 3-5x setup — speculative if/then."

> Stored test: "grow earnings significantly and embrace AI agents over the next two years."

The most persistent bear in the lane concedes a 3-5x setup conditional on execution. He repeats it on 2026-04-04 at conviction 60. That is a bear defining his own falsifier, and it is the cleanest statement of what the bull case actually requires.

Dan Ives appears three times, bullish at convictions 82, 78 and 55 (2026-02-13 and 2026-05-20), calling the selloff "a generational buying opportunity" and predicting AI use cases within 6-18 months. He is sell-side and promoter-class, and is given LOW weight — noted for completeness rather than relied upon.

Two claims the filings have already settled

First, and most usefully, a bull claim that is now factually wrong.

> 2025-10-18 · bullish · conviction 58 · entity Salesforce · we_study_billionaires

> "Salesforce is attractive — 10yr CAGRs of 27% revenue, 28% EPS, 31% FCF, no debt, founder-led — but a watchlist name pending research."

"No debt" was already wrong when written and is now emphatically wrong: $39.3 billion of borrowings at 2026-04-30. This is exactly why filings outrank commentary, and it is a good reminder that a low-conviction watchlist mention is not research.

Second, a claim that contradicts a filing's own framing.

> 2026-07-07 · neutral · conviction 60 · entity CRM · speaker Sema Amble, a16z

> "Salesforce's headless 360 launch was a classic marketing rebrand of APIs already exposed — 'nothing actually changed' in the Salesforce context."

The 10-Q for the very next quarter states that "Salesforce is the platform that powers how humans and agents work together, whether using Customer 360 apps, Slack, Headless 360, or other user interfaces." One of these is a company describing a product architecture and the other is a practitioner calling it a rebrand. The dive cannot adjudicate — but it can note that the company's chief evidence of an agentic pivot was publicly described as repackaging four weeks before the quarter closed.

Management's own voice appears once, and is given LOW weight per house policy: a 2025-04-24 claim (bullish, conviction 80, entities Mark Benioff and Salesforce) that "We're at the precipice of the greatest moment in the history of enterprise software and cloud computing." That is the chief executive on a podcast. It is not evidence.

Historical context worth one line. The oldest usable claim is Arthur Hayes on 2022-12-10 (bearish, conviction 85): "Salesforce is the emperor with no clothes; pulling guidance hides that S&M is 44% of revenue vs ~1% net profit — can fall much further." He was right on the facts: fiscal 2023 net income was $208M on $31.352B of revenue, 0.7%. The trailing net margin today is 18.7%. The margin half of that bear case has been comprehensively falsified by the company's own execution — which is the strongest single argument for taking the current cost discipline seriously.

8. Data integrity — what we rejected from the vendor file and why

Sixteen items. This section is the point of the exercise.

1. bal_a is one quarter stale, and that quarter contained the largest balance-sheet event in the company's history — the most consequential defect in this file. The vendor's most recent balance sheet is 2026-01-31: total debt $17.176B, total equity $59.142B. The 10-Q for the quarter ended 2026-04-30 discloses $33.3B of senior unsecured notes plus $6.0B of term loan = $39.3B of borrowings, following a $25.0B note issue in March 2026 used to fund a $25.0B accelerated share repurchase, with equity implied at roughly $33B by the vendor's own bookValuePerShareTTM of $39.44. Anyone reading bal_a sees a company with modest leverage. The filing shows a company with roughly $30B of net debt. We use the filing throughout.

2. netDebt omits short-term investments — provable at fiscal year end. The vendor reports netDebt of $9.849B at 2026-01-31, which is exactly total debt ($17.176B) less cash and equivalents ($7.327B), silently discarding $2.238B of short-term investments. The corrected figure is $7.611B. The same omission propagates into enterprise value.

3. enterpriseValueTTM of $189.366B is ~$2.8B (1.5%) too high; enterprise value REBUILT. The vendor's figure implies net debt of $32.949B. From the 10-Q: borrowings $39.3B, plus finance leases of approximately $2.7B (last disclosed at 2026-01-31; not separately given at 2026-04-30 — the 10-Q's metrics table explicitly notes "Amounts do not include operating or financing lease obligations"), less $11.8B of cash, equivalents and marketable securities = ~$30.2B. Corrected enterprise value ~$186.6B, and every EV multiple in this dive is computed on it.

4. Operating income is computed INCONSISTENTLY across years, overstating margin expansion by 150 basis points. Vendor fiscal 2026 opinc is $8.917B (21.5% margin); the 10-K states "income from operations was $8.3 billion... Operating margin... increased to approximately 20 percent." The gap is the $586M restructuring charge, which the vendor excludes. But vendor fiscal 2025 opinc of $7.205B matches the 10-K's $7.2B and ~19% despite a $461M restructuring charge in that year — so restructuring is INCLUDED in fiscal 2025 and EXCLUDED in fiscal 2026. The same defect appears in the quarter: vendor $2.427B (21.8%) against the 10-Q's "$2.3 billion" and "approximately 21 percent", with $80M of restructuring in the period. The vendor implies +250bp of margin expansion. The filings show +100bp. We use the filings.

5. est.ebitdaAvg and est.ebitAvg are unusable — REJECTED, along with every EBITDA multiple. The FY2026E rows show EBITDA of $19.288B and EBIT of $17.192B against a reported GAAP operating income of $8.331B and an EBITDA of roughly $11.96B. Even adding back the entire $3.509B of stock compensation reaches $15.47B. evToEBITDATTM (13.527x), enterpriseValueMultipleTTM (13.527x) and netDebtToEBITDATTM (2.354x) are all rejected. No EBITDA figure appears anywhere in this dive.

6. est.niAvg is internally inconsistent beyond FY2028E. FY2029E net income of $14.827B is BELOW FY2028E's $15.011B while epsAvg rises from $15.627 to $17.949 — which would require a ~19% share-count reduction in one year. The FY2029E-FY2031E rows are used for context only and are separately thin (12, 6 and 11 analysts on EPS against 32 for FY2027E and FY2028E).

7. Estimates are NON-GAAP; reported EPS is GAAP — a $5.22 per-share gap. est.epsAvg of $11.776 for fiscal 2026 against a reported GAAP diluted EPS of $7.80 and an actual non-GAAP result of $12.55. Trailing: non-GAAP $13.85, GAAP $8.63. All forward multiples in this dive are stated on the non-GAAP basis and labelled; GAAP equivalents are given alongside; and stock-based compensation of $3.509B, 8.3% of revenue, is treated as the real cost it is wherever a cash-flow multiple is quoted.

8. seg_prod labels are obsolete, twice over — though the arithmetic is clean. The six-line fiscal 2026 split sums exactly to the reported $41.525B (verified; no double-count, which is the failure mode this batch was told to hunt), and seg_geo sums exactly too. But the 10-K discloses that the service offerings were renamed to reference Agentforce in Q3 fiscal 2026, and the Q1 fiscal 2027 10-Q re-cut the disclosure into just two categories — Agentforce Apps, and Data 360/Headless Platform/Other. The vendor carries neither presentation and has no fiscal 2027 data. Note also that Salesforce operates as one reportable segment; seg_prod is a revenue disaggregation, not a segment note.

9. tech 52-week extremes disagree with quote. tech reports hi52 $266.23 and lo52 $150.12; quote reports yearHigh $269.11 and yearLow $146.32. We use the tech values, because they are computed from the same closing series as the moving averages this dive quotes, and because $150.12 is independently corroborated by an insider tax-withholding transaction priced at exactly $150.12 on 2026-06-22. The quote figures are presumably intraday extremes.

10. profile.lastDividend of 2.128 contradicts the rest of the file. dividendPerShareTTM is $1.712 and dividendYieldTTM is 0.896%, and the 10-Q's $365M of dividends and dividend equivalents on roughly 870M weighted shares is about $0.42 a quarter, or $1.68 annualised. We use $1.712 and a 0.90% yield; lastDividend is rejected.

11. Per-share trailing metrics use ~868M shares while marketCap implies exactly 819.000M — both are defensible, and the roundness is flagged. $156,416,715,000 ÷ $190.985 = 819,000,000 to the share, which is suspiciously round. Our rollforward supports it: approximately 930M outstanding at 2026-01-31, less the 103M initial accelerated-repurchase delivery and roughly 13M of open-market repurchases (the $27.2B of quarterly repurchases less the $25.0B prepayment, at prevailing prices), plus equity-plan issuance. We use 819.0M for market capitalisation and enterprise value, and weighted diluted counts for per-share earnings, and we say which is which at every use.

12. tangibleBookValuePerShareTTM of −$36.53 and tangibleAssetValueTTM of −$31.706B are CORRECT, not errors. Goodwill of $57.941B and intangibles of $6.815B at 2026-01-31 (57.7% of assets; 61.8% on the trailing reading) against equity cut to roughly $33B by the repurchase produce genuinely negative tangible book. Reported so it is not mistaken for a fault — and because it means no price-to-book figure can be used for this name. priceToBookRatioTTM (4.842x) and priceToFairValueTTM (4.842x) are rejected as uninformative.

13. currentRatioTTM and quickRatioTTM are identical at 0.786 — also correct, not a defect. Salesforce carries no inventory (inventoryTurnoverTTM, daysOfInventoryOutstandingTTM and averageInventoryTTM are all zero, correctly), so the two ratios coincide. A sub-1.0 current ratio and negative working capital of −$5.889B are structural for subscription software, where unearned revenue is a current liability settled by delivering service rather than cash. Neither is a liquidity signal. payablesTurnoverTTM and daysOfPayablesOutstandingTTM of zero are, however, unpopulated rather than accurate, and are rejected.

14. The 2026-05-27 8-K is a cover page only — MANAGEMENT GUIDANCE IS ABSENT. Exhibit 99.1, the FQ1 fiscal-2027 earnings release containing the outlook, is not in the extracted text. No guidance figure appears anywhere in this dive. Given the half-weight rule that is a modest loss, but it is stated rather than papered over. Earnings-call transcripts are separately unavailable on the current data plan.

15. earn_cal.epsActual is the NON-GAAP series, used deliberately. Q1 fiscal 2027 shows $3.88 against a reported GAAP diluted EPS of $2.42. The series is used as the non-GAAP actuals in Section 2 and labelled at every use. The forward row (2026-09-02, epsEstimated $3.28, revenueEstimated $11.331B) is likewise non-GAAP.

16. Headcount cannot be trended, and it matters. The 10-K discloses 83,334 employees at 2026-01-31, matching the vendor profile exactly. No prior-year figure is in the extracted text, so revenue per employee ($498,000 for fiscal 2026) cannot be compared with anything. This is a real gap, because Salesforce's own headcount is the single cleanest read on whether agents replace users — a software company that automates its own seats away is the most credible possible witness for the bear case.

Minor and correctly reported: the 10-K's "approximately 50 million shares for approximately $12.7 billion" against the vendor's cash-flow buyback of $12.596B — a $104M difference consistent with excise tax and settlement timing, not a defect. capexToRevenueTTM of 1.3% and capexToOperatingCashFlowTTM of 3.7% are correct and remarkable. researchAndDevelopementToRevenueTTM of 14.4% ties to the fiscal 2026 R&D of $5.993B on $41.525B.

9. Technicals — a mid-recovery chart below a declining long-term average

• Price $190.99, −28.3% from the 52-week high of $266.23, and +27.2% ABOVE the 52-week low of $150.12. Maximum drawdown from peak: −48.1%. This stock halved and has since retraced about a quarter of it.
• Above the 50-day, below the 200-day. 50-DMA $171.44 — the price is 11.4% above it. 200-DMA $204.56 — the price is 6.6% below it, and that average is declining. This is the exact configuration one knowledge-base claim named on 2026-07-19: a bounce toward a downward-sloping 200-day average that requires a business pivot to become durable.
• RSI 65.6 — approaching overbought, not oversold. MACD +5.08, positive. The short-term momentum is genuinely constructive.
• Relative performance is the worst in this batch. 3-month +3.0% vs SPY +7.6% (−4.6 points); 6-month −9.4% vs SPY +11.1% (−20.5 points); 12-month −23.8% vs SPY +24.3% and QQQ +30.8% (−48 and −55 points).
• Sentiment corroboration runs the OTHER way, which is the anomaly. 2 strong buy, 72 buy, 21 hold, 2 sell out of 97 ratings, and a target range of $215 to $325 with a $265.75 consensus. The chart says capitulation-then-bounce; the ratings say nothing has happened.

Today's move and what it does to the entry

Salesforce closed +2.71% at $190.99 on a day when this batch's semiconductor names rose 5% to 13%. That is the first thing to say: CRM did not participate in today's leadership — it lagged it badly. Intel +10.9%, Marvell +12.8%, Micron +7.6%, Lam +7.9%, AMD +7.0%. Salesforce +2.7%; Adobe +2.5%; Accenture +2.9%. The software complex moved as market beta while the semiconductor complex moved as a theme — which is, precisely, the rotation the knowledge base has been describing all year in the words "you want to be long semi, short software." Today's move is not evidence that the software narrative is turning. It is evidence that it is not.

The intraday shape matters more than the percentage, and it argues against buying the close. The stock opened at $181.24, which was also the day's low, and 2.5% BELOW the prior close of $185.95. It then rallied to close at $190.985 against a high of $191.285. That is a 5.4% intraday reversal finishing at the top of the range. Technically it is a constructive bar. Practically it means anyone buying at today's close is paying the single worst price available during the session, on a name that is already 27.2% off its low and 11.4% above its 50-day average with RSI at 65.6.

So: are we asking the reader to buy into a one-day move? No — and here the honest answer is easy. There is no thesis that requires today's price. There is a scheduled, binding disclosure in 29 days that contains the one number capable of resolving the argument. The stock is neither washed out (it has already bounced 27%) nor confirmed (it is below a declining 200-day average). Buying today captures neither the capitulation entry nor the confirmation entry.

The levels that matter, stated in advance. The 200-DMA at $204.56 is 7.1% above and is the level a named bear claim identified as the natural stopping point for this bounce — reclaiming and holding it on the September print would be genuine evidence. The June low at $150.12 is 21.4% below and is where the valuation case becomes emphatic rather than merely adequate (roughly 9.6x FY2028E non-GAAP EPS, close to our bear anchor). Between those two levels this is a Watch, and today's close sits squarely between them.

10. Insiders — all mechanical, no discretionary activity in either direction

Every transaction in this file is either an M-Exempt (restricted-stock-unit conversion) or an F-InKind (shares withheld to cover tax on vesting). There is not a single open-market purchase and not a single open-market sale.

The reading, stated carefully. This tells you almost nothing about conviction and one useful thing about price. Routine monthly vesting with statutory tax withholding is not a signal in either direction — these executives did not choose to sell, and they did not choose to buy. Post-transaction holdings are essentially unchanged (Tallapragada 71,067 and 70,167; Harris 163,288).

What it does establish is a price fix. The 2026-06-22 withholding price of exactly $150.12 matches the tech 52-week low to the cent, which independently dates the low to around 22 June 2026 and confirms that the stock has run +27.2% in roughly six weeks. That is why Section 9 declines to describe this chart as washed out.

Two absences worth noting. Marc Benioff does not appear in this file at all — no purchases, no sales, no vesting in the eight transactions captured. And no director or officer has bought a share on the open market during a period in which the stock halved from its high, while the company itself spent $25 billion doing so. Management's capital and management's own capital are behaving differently.

11. Verdict, kill-criteria and flip conditions

Watch.

The case for owning it. The de-rating is largely done: 13.5x FY2027E and 12.2x FY2028E non-GAAP EPS, 22.1x trailing GAAP, and ~16.7x corrected enterprise value to free cash flow with stock compensation charged as a cost — inside the 15-17x band the bears themselves named as the correct terminal multiple. The business is genuinely good: 95% subscription revenue, no customer above 10%, $67.9B of contracted backlog, free cash flow of $14.4B on capital expenditure of $594M, margins expanding about 100bp a year on the filing basis, and a cost programme that has already falsified the 2022 bear case about profitability. Attrition is "consistent", current RPO grew 14%, and no disclosure in any filing shows seats being lost.

The case against buying it today. Our base is +10% and requires no re-rating; the street's is +39% and requires a 26% multiple expansion it does not justify — and the Adobe comparison inside this batch (+1.4% to consensus on the same narrative) says the Salesforce targets are stale. Organic constant-currency growth is roughly 7%, dressed as 13.3% by an acquisition and currency. Current RPO growth is decelerating on a constant-currency basis even with the acquisition inside it. Net debt went from $7.6B to ~$30.2B in twelve weeks, tangible book equity is negative $31.7B, and the $25B repurchase that caused it was struck at $198.34 — above today's price — and is only clearly accretive on the adjusted earnings basis. The knowledge base is decisively negative: 89 usable claims, roughly 52 bearish, from 17 sources. And the single disclosure capable of resolving the argument arrives in 29 days.

A 1.4:1 payoff with a −32% bear case, no edge versus consensus in the direction that pays, and a binding catalyst four weeks out is a wait. Watch.

Pre-registered KILL criteria — what would remove this from the watchlist entirely:

• Current remaining performance obligation growth below ~10% year-over-year in the 2026-09-02 release, or below ~8% constant-currency ex-acquisition. This is the primary kill.
• Any change from the word "consistent" in the attrition disclosure, or the first appearance of language about customers reducing subscriptions at renewal.
• Any disclosure of declining seat, subscription or user counts, in any form.
• A second large debt-funded acquisition. The balance sheet has been levered once; doing it again while organic growth decelerates would confirm that growth is being bought rather than earned.
• Reported revenue growth falling below ~5% once Informatica laps in the quarter ending 2027-01-31.
• Goodwill impairment. $57.941B of goodwill against roughly $33B of equity means an impairment of any size is a solvency-adjacent event, not an accounting one.
• A decisive break below $150.12 (the 22 June 2026 low, ~9.6x FY2028E) on company-specific news rather than sector beta — at which point the bear case is being priced, not debated.

Pre-registered FLIP conditions — what would convert this to Buy — Tactical:

• Consumption revenue disclosed and sized. If Salesforce breaks out Flex Credits or any usage-based line as a separate, growing figure, the terminal-value debate changes shape immediately. This is the highest-value flip and the one the new two-category revenue presentation makes easiest.
• Current RPO growth of 14% or better on 2026-09-02 with LESS than one point of currency help, i.e. genuine constant-currency acceleration.
• Street price targets being CUT toward the price on an in-line print — counterintuitively bullish, because it removes the staleness that currently makes the +39% consensus uninformative, and creates a real consensus to be measured against.
• Reclaiming and holding the 200-day moving average at $204.56, which is the level the most specific bear claim in the knowledge base named as the ceiling on this bounce.
• A price near $150-160 on sector beta rather than company news — roughly our bear anchor, at which the asymmetry inverts and the position becomes worth taking without confirmation.
• Any open-market insider purchase, of which there have been none.

Where CRM would fit in the Synthos Framework Portfolio. Not a position today. If a confirming September print arrives, a 1.0-1.5% position in the de-rated quality-software sleeve, entered in two tranches rather than one — small relative to the 3-4% a genuine conviction name earns, because the knowledge-base lane is negative and because our own base case is only +10%. Explicit overlap warning: Salesforce and Adobe are the same trade. They are named together in roughly twenty knowledge-base claims, they moved within twenty basis points of each other today (+2.71% and +2.51%), and they are both the short leg of the rotation whose long leg — semiconductors — is what actually rallied on 2026-08-04. Do not hold both. If a software expression of this theme is wanted, it should be one name, sized once. Logged as a tracked Synthos call (Watch) as of 2026-08-04 at $190.99.

Single biggest risk: seat erosion appearing in current remaining performance obligation. Salesforce sells per seat; agentic automation reduces the number of humans who need a seat; the company's answer is consumption pricing whose revenue it has not disclosed. cRPO is the earliest place the damage would show, it is reported on 2026-09-02, and nothing in the vendor payload or in any of the six filings in this file can size it in advance. A print below 10% would not merely miss a quarter — it would validate a thesis held at conviction 75 to 90 by a large fraction of the tracked panel and would reprice the terminal value, not the year.

Most fragile assumption in the price: that the market keeps paying about 13x forward non-GAAP earnings. Everything else in the inversion can drift for a year without breaking the base case. That one cannot. 13.5x looks like a floor because it is so much lower than the ~30x this stock used to command — but a multiple is only a floor if the earnings underneath it are durable, and durability is the exact thing under dispute. If organic growth settles at 5% rather than 7%, 13.5x is not a floor; it is a waypoint. That is why the bear case is $130 rather than $165, and why this is a Watch rather than a Buy.

Provenance & disclosures

• Traceability: 120 raw knowledge-base hits for the terms CRM, Salesforce, Benioff and Agentforce across 51,928 distilled claims. 31 removed: 18 discarded as acronym/common-noun/metaphor/analogy collisions (the token "CRM" as the software category; "salesforce" as the literal common noun in a claim about Solana; "the Salesforce of security"; the Google/Salesforce dot-com analogy in five Bitcoin and macro claims; claims whose entity list names IQVIA, Veeva, CrowdStrike, Sierra, SAP or Adobe with Salesforce only as a comparator), 8 discarded as attribution-quarantined (thinker: quarantine_misattributed, excluded on principle regardless of direction), and 5 collapsed as duplicate distillations of the same source document. 89 usable claims across 17 distinct sources. Stance after reclassifying four long-Chevron/short-Salesforce pair trades tagged bullish: roughly 52 bearish, 27 bullish, 10 neutral — net negative-high. Concentration warning: approximately 31 of the 89 come from one speaker (Jordi Visser, skill 2.0). Where speaker and thinker differ the claim is attributed to the speaker and the channel is noted separately — the 2026-02-07 and 2026-01-31 Visser claims carry doc_id values on the Pompliano channel and are attributed to Visser. Two claim texts naming specific AI model developers were PARAPHRASED rather than quoted verbatim, under the house rule against printing AI model or laboratory names; both are identified in Section 7. The Jim Bianco claim has an EMPTY date field — a defect in our own store, disclosed, and the claim is excluded from any dated test. Management voice (Marc Benioff, 2025-04-24) and promoter-class sell-side voice (Dan Ives, three claims) are explicitly LOW-weighted. All quotes are verbatim from stored claim text except where paraphrase is disclosed.
• Data as-of: income statement 2026-04-30 (Q1 fiscal 2027, filed 2026-05-28) · vendor balance sheet 2026-01-31 and therefore one quarter stale, with the 2026-04-30 position reconstructed from the 10-Q (Section 4) · cash flow 2026-01-31 annual, with Q1 fiscal 2027 flows from the 10-Q · estimates 2026-08-04 · prices 2026-08-04, quote timestamp 1785873598 = 2026-08-04T19:59:58Z, $190.985 (+2.71%), 50-DMA $171.44, 200-DMA $204.56, RSI 65.6 · knowledge-base claims 2026-08-04. Sources: the Synthos vendor payload for CRM and six SEC filings (10-K filed 2026-03-02 for the fiscal year ended 2026-01-31; 10-Q filed 2025-12-04 for the quarter ended 2025-10-31; 10-Q filed 2026-05-28 for the quarter ended 2026-04-30; and three 8-Ks dated 2026-05-27, 2026-06-01 and 2026-06-02). No figure comes from memory, recall or external retrieval.
• Fiscal-year convention: Salesforce's fiscal year ends 31 January and the label runs one year ahead of the calendar. Fiscal 2026 = year ended 2026-01-31. Fiscal 2027 = year ending 2027-01-31. Q1 fiscal 2027 = the three months ended 2026-04-30. Every fiscal reference in this dive is stated with its calendar end date at least once.
• Data rejections (detailed in Section 8): bal_a rejected as one quarter stale, missing a $25.0B note issue and a $25.0B accelerated share repurchase; netDebt corrected from $9.849B to $7.611B at fiscal year end for omitted short-term investments; enterpriseValueTTM corrected from $189.366B to ~$186.6B and rebuilt from the 10-Q ($39.3B of borrowings plus ~$2.7B of finance leases less $11.8B of cash and marketable securities = ~$30.2B of net debt); vendor opinc rejected for fiscal 2026 and Q1 fiscal 2027 as excluding restructuring charges that fiscal 2025 includes, cutting stated margin expansion from 250bp to 100bp; est.ebitdaAvg and est.ebitAvg rejected as unusable, and evToEBITDATTM, enterpriseValueMultipleTTM and netDebtToEBITDATTM rejected with them; est.niAvg rejected beyond FY2028E as internally inconsistent with epsAvg; profile.lastDividend (2.128) rejected in favour of dividendPerShareTTM ($1.712); priceToBookRatioTTM and priceToFairValueTTM (4.842x) rejected as uninformative against negative tangible book; payablesTurnoverTTM and daysOfPayablesOutstandingTTM rejected as unpopulated; quote 52-week extremes ($269.11/$146.32) rejected in favour of tech ($266.23/$150.12), the latter corroborated by an insider transaction at exactly $150.12; seg_prod labels flagged as superseded twice (renamed to Agentforce in Q3 fiscal 2026, re-cut into two categories in Q1 fiscal 2027) although the arithmetic ties exactly and contains no double-count; the 2026-05-27 8-K flagged as cover page only, with the earnings release and all management guidance ABSENT; prior-year headcount stated as missing, so revenue per employee cannot be trended.
• Basis note — the most important one in this dive: analyst estimates are NON-GAAP; reported EPS is GAAP, and the gap is $5.22 a share on a trailing basis ($13.85 non-GAAP against $8.63 GAAP; $12.55 against $7.80 for fiscal 2026). The bridge is principally stock-based compensation of $3.509B, 8.3% of revenue, plus $586M of restructuring and amortisation of acquired intangibles. Forward multiples in this dive are stated on the non-GAAP basis and labelled at every use, with GAAP equivalents given alongside; every free-cash-flow multiple charges stock compensation as a cost, because it is one. Trailing free cash flow of ~$14.66B becomes ~$11.15B after that charge, and the corrected enterprise-value multiple is ~16.7x rather than the ~12.9x the vendor's evToFreeCashFlowTTM implies.
• Estimate coverage: excellent through the year ending 2028-01-31 (32 analysts on EPS, 38-39 on revenue) and thin beyond (12, 6 and 11 analysts on FY2029E-FY2031E EPS). The FY2031E revenue of $70.505B implies +16.5% growth, faster than anything achieved in three years. No conclusion rests on any row beyond FY2028E.
• Peer note: the vendor peer set mixes genuine comparables (SAP $227.9B, IBM $221.5B, ServiceNow $122.1B, Intuit $88.5B) with irrelevant ones (Micron, Lam Research, Uber). No peer multiples are supplied, so no peer-multiple comparison is drawn. The one cross-reference used is Adobe from this batch's own table ($257.64 close, $261.16 consensus target, +1.4%) against Salesforce's +39.1% — same sector, same narrative, same trading day, and the basis for concluding that the Salesforce targets are stale.
• Fair-value caveat: the $130 / $210 / $275 anchors are 8.3x / 13.4x / 17.6x FY2028E consensus non-GAAP EPS of $15.627, cross-checked against corrected enterprise value on FY2028E revenue (2.7x / 4.0x / 5.1x) and on free cash flow after stock compensation. Stated arithmetic, not a discounted cash flow. The base case assumes no multiple re-rating whatsoever — it is today's multiple rolled forward one year. The bull case requires a 26% multiple expansion and is flagged as the fragile leg; it is essentially the street's own case.
• Non-equity tripwire: checked and passed. Common stock on the NYSE, IPO 2004-06-23, $190.99, beta 1.178, variable dividend, 6.76M shares traded, 52-week range $150.12-$266.23. Not a preferred, not a baby bond.
• Timing: the most recent quarter (ended 2026-04-30) was reported 2026-05-27 and filed 2026-05-28, sixty-eight days before this dive, so the data is one full quarter old. The next print is 2026-09-02, 29 days away, and its current-RPO disclosure is treated throughout as the deciding evidence — which is the explicit reason this is a Watch rather than a Buy.
• Not investment advice. Independent research, educational and informational only, never personalised. No recommendation to buy, sell or hold any security is made to any person.
• Version: 2026-08-04-full.