What we said in July — and what happened
Right on the price, wrong on the balance sheet
On 2 July we published Salesforce at $156.66 under the title "Buy the Fear". We said the shares had fallen 43% on the fear that AI would hollow out seat-based software; that the business was a wonderful franchise with a deep switching-cost moat; that at about ten times free cash flow it was cheap; and that readers should begin buying and add with conviction below $145. The shares never reached $145. They rose to $264.43 on 3 September and closed at $194.52 on 17 November. Here is the account, and it contains an error larger than any we have had to correct in this series.
| What we said on 2 Jul | What happened by 17 Nov | Grade |
|---|---|---|
| 'Buy the fear' at $156.66 | +24.2% to $194.52; +69% at the 3 September peak. | ✓ Right |
| 'Add with conviction below ~$145' | Never reached: the lowest close after publication was $156.93 (23 July). | ◆ Not triggered |
| Growth decelerating towards high single digits | Reported growth 10.8% in Q2, but $456m came from Informatica. Without it, about 6.4%, down from ~8.7% in Q1. | ✓ Right, and lower |
| 'The M&A itch is not cured' | About $15.5bn of deals since July: Contentful, Fin, Listen Labs, Braze and Cresta (two at reported prices). | ✓ Right |
| Stock pay ~8% of revenue, 'a real cost' | Guided at 9.0% of revenue this fiscal year. | ✓ Right, worse |
| 'Net debt ~$10bn; interest covered ~21×' | Wrong when written. In March 2026 Salesforce issued $25bn of notes and a $6bn term loan to fund a $25bn accelerated buyback. Net debt was about $28–30bn; operating profit covers interest about 4.9 times. | ✗ Wrong |
| 'Share count falling to ~950m' | Stale. The buyback delivered 103m shares up front; there were about 819m shares in May. | ✗ Wrong |
| Market value ~$149bn; yield ~1.4% | On the right share count, about $128bn at $156.66, and a yield of 1.12%. | ✗ Wrong |
| Agentforce 'unproven at scale' | Agentforce annual recurring revenue above $1.5bn (+240%), but the definition was widened in Q2 to include Slackbot and other products, so the growth rate is not like-for-like. | ◆ Partly |
| 'AI likelier a tailwind than a wrecking ball' | Contracted revenue for the next twelve months grew 14%; no fall in seats disclosed. But the market fell 9.45% on 17 November on exactly this fear. | ◆ Too early |
The error, plainly. Our July report took its balance sheet and share count from a snapshot in our data feed dated 31 January 2026, before the March debt issue and buyback. It therefore described Salesforce as a company with about $10bn of net debt and interest covered 21 times, when it had roughly three times that debt and a fifth of that cover. Two of the reasons we gave for calling the shares cheap and safe, the light balance sheet and the high interest cover, were not true on the day we wrote. The call still made money, but readers deserved the right numbers, and we did not check the filings that would have shown them. We have corrected our process: balance sheets in follow-ups are now taken from the latest 10-Q, not from the feed.
★★ The round trip
From $157 to $264 and back to $195, in four and a half months
| Date | What happened | CRM that day |
|---|---|---|
| 27 Aug | Second-quarter results: contracted revenue for the next twelve months +14%, the full-year guide raised, and adjusted EPS of $5.90 against about $3.27 expected, most of the beat from a gain on Anthropic shares. | +22.6% |
| 3 Sep | Peak close. | $264.43 |
| 16–18 Sep | Investor day: the fiscal-2030 target of $63bn+ of revenue reiterated, no fiscal-2028 guide. | −3.1%, −2.0% |
| 29 Sep – 2 Nov | Listen Labs (reported ~$2bn), Braze ($5.66bn) and Cresta (reported ~$2.7bn) agreed. | drift lower |
| 4 Nov | S&P reported to have cut Salesforce's rating to A from A+, citing acquisitions and shareholder returns (TIKR). | $227.11 |
| 17 Nov | No company news. Software shares sell off on the eve of Anthropic's expected IPO pricing and on fears that AI agents shrink per-seat pricing; Salesforce is the S&P 500's worst performer. | −9.45% |
Two lessons for an owner. The first is that the fear we bought in July was not resolved; it was suspended. The second-quarter report gave the market permission to stop worrying for two months, and the worry came back as soon as something new, an AI company's enormous listing, reminded investors what they were afraid of. Anyone who bought in July for the AI question to be answered by now was early by years, not months.
The second is that the best day in the period was not as good as it looked. The 22.6% jump on 27 August followed earnings per share far above expectations, and most of the excess was a non-cash gain on Salesforce's stake in Anthropic, not the business. The market eventually worked that out.
It is also worth seeing that most of the rise came before the results. The shares had touched $146.32 during trading on 22 June, rose about 31% to $205.62 by 26 August with no single cause we could verify, and only then jumped on the report. A reader who waited for proof that the business was sound paid about 60% more than one who bought the fear in July. That is the usual price of certainty, and the reason we write about companies when they are disliked.
★ The balance sheet we missed
A $25 billion buyback on borrowed money, and the deals that followed
In March 2026 Salesforce issued $25bn of bonds maturing between 2028 and 2066 and a $6bn term loan, and used the bond money for a $25bn accelerated share repurchase: it paid the banks $25bn up front and received 103 million shares at once, with the rest due when the programme settled in October. Management's presentation of late August expected an average price of about $182 and a reduction in the share count of at least 14%, and described the year as returning 190% of free cash flow to shareholders. The final number of shares will be in the third-quarter 10-Q in early December.
The result is a different company from the one we described. At 31 July Salesforce had $39.3bn of debt and $11.4bn of cash and securities: net debt of about $27.9bn, against about $4.9bn in January. Interest expense rose from $67m in the second quarter of last year to $473m; operating profit now covers it about 4.9 times. Since July it has paid about $5bn for Contentful and Fin and agreed to pay about $10bn more for Listen Labs, Braze and Cresta, most of it in cash, so net debt is heading towards $35bn before the year's cash flow. S&P was reported to have cut its rating to A in November, and Moody's to A2 in March; both are still comfortably investment grade.
| Deal (since July) | Status | Price |
|---|---|---|
| Contentful | Closed, ~1 September | ~$1.5bn net of the stake already owned |
| Fin (from Intercom) | Closed, 10 September | ~$3.6bn |
| Listen Labs | Agreed 29 September; appears closed | ~$2bn (reported) |
| Braze | Agreed 20 October; closes in fiscal 2028 | $5.66bn ($47.50 a share, cash) |
| Cresta | Agreed 2 November | ~$2.7bn (reported) |
The Anthropic stake. Salesforce owns a stake in Anthropic, valued in its books at about $5.1bn at 31 July after $3.0bn of unrealised gains in the first half. Anthropic was expected to price its listing on the evening of 18 November, after our price date. The listing will set a market price for the stake and will produce a non-cash gain or loss in the fourth quarter, depending on where the shares trade. The chief executive said in September, according to a transcript, that Salesforce would probably sell the stake over time to pay down the buyback debt. That would be a sensible use of a lucky investment.
What the buyback did for an owner. The banks delivered about 103 million shares in March at a reference price of $198.34, the close on 16 March. If the final average is the $182 management expected, the programme will retire about 137 million shares in all, some 14% of the company. Our arithmetic on what that trade means: interest now runs at about $1.9bn a year, against about $0.3bn a year ago, so the year's borrowing (the buyback and the Informatica loan) costs about $1.3bn more after tax; the shares retired would have earned about $2.5bn of free cash flow a year at today's level. On paper, Salesforce borrowed at about 4% after tax to buy a business yielding about 10% on its cash, and that is accretive as long as the cash holds. The cost is of another kind. Before March, the AI question was a question about how fast Salesforce would grow. With $28bn of net debt, it is also a question about how much room the company has if it is wrong.
Two growth rates
11% on the headline, about 6% underneath
Salesforce reported second-quarter revenue of $11.35bn, up 11%. Informatica, bought in November 2025, contributed $456m of it. Without Informatica, revenue grew about 6.4%, down from about 8.7% in the first quarter. The core clouds, now reported as "Agentforce Apps" (sales, service, marketing, commerce and Slack), grew 8% in constant currency; the faster growth came from the data and integration business, much of it acquired. Professional services shrank 4%, as customers bought fewer large transformation projects.
The forward-looking numbers are better. Contracted revenue for the next twelve months (cRPO) grew 14%, up from 13% in the first quarter and 10% a year earlier, and management said new bookings grew at the fastest pace in four years. It promised that organic growth would re-accelerate in the second half; the third-quarter report on 2 December is the first test. Gross margin slipped to 76.7% from 78.1%, as the costs of running AI and the amortisation of acquisitions rose faster than revenue.
| Q2 FY2027 | Value | vs a year earlier |
|---|---|---|
| Revenue | $11.35bn | +11% (~+6.4% without Informatica) |
| Contracted revenue, next 12 months (cRPO) | $33.5bn | +14% (constant currency) |
| Agentforce + Data 360 annual recurring revenue | ~$3.9bn | +210% (definition widened in Q2) |
| Operating margin, reported / adjusted | 20.5% / 34.1% | 22.8% / 34.3% |
| EPS, reported / adjusted | $4.29 / $5.90 | Incl. $2.43 / $2.53 of investment gains |
| FY2027 guidance | $46.1–46.4bn revenue | +11–12%, of which >3 points Informatica |
★ The AI question, again
What Salesforce concedes, what it claims, and what nobody yet knows
On 17 November Salesforce fell 9.45% with no news of its own. Software shares as a group fell; ServiceNow and HubSpot lost about 7% each. The reasons given in the press were two: investors raising cash to buy into Anthropic's listing, and the old fear, sharpened by that listing, that AI agents will do the work of the employees whose seats Salesforce charges for. One analyst was quoted saying AI companies would try to "destroy Salesforce entirely".
What Salesforce concedes. Its own investor-day deck describes the "agentic" stage of a customer as adopting Agentforce "with some seat optimization", offset, it says, by 1.5 to 2 times more recurring revenue from consumption. That is the company telling you, in its own slide, that some customers will buy fewer seats. What it claims. An executive said in September, according to a transcript, that there was "no decline yet" in seat counts for the sales and service clouds; pricing is moving to premium editions (one at $550 per user per month), usage credits and enterprise-wide agreements, which made up seven of the ten largest deals in the second quarter. What nobody knows is whether the revenue from agents will be larger than the revenue from the seats they replace, and whether customers will buy those agents from Salesforce or from the AI companies themselves. Anthropic is a Salesforce partner and investee; it may also become a competitor.
| The bull case | The bear case |
|---|---|
| The data is the moat. Agents need the customer records, permissions and workflows that already live in Salesforce; contracted revenue is growing faster than it was a year ago. | Fewer people, fewer seats. The company itself expects 'some seat optimization'; the core clouds grow 8%; gross margin is slipping as AI costs rise. |
| Consumption pricing. If an agent replaces a seat and is billed by use at a higher price, revenue per customer rises. | The AI labs move up. Anthropic's listing gives an AI company more capital than most software firms to build customer tools of its own. |
| Cheap enough to be wrong. At about 14 times free cash flow after stock pay, the price assumes little growth. | Growth bought, not grown. About $15.5bn of acquisitions since July makes the organic rate hard to see and easy to flatter. |
The moat, rechecked
Still deep, a little narrower than we scored it
| The claim (July) | Evidence now | Verdict |
|---|---|---|
| Switching costs of the system of record | cRPO +14%; no disclosed fall in seats or renewal rates. | ✓ Intact |
| Pricing power | Premium editions and credits sell; but gross margin fell from 78.1% to 76.7% and seats face 'some optimization'. | ◆ Under pressure |
| The ecosystem | Partners and developers still build on Salesforce; Anthropic is both partner and possible rival. | ◆ Contested |
| The data | Data 360 and Informatica growing 20%; the company is betting the moat moves from seats to data. | ✓ The right bet |
We lower the moat from 8 to 7. Nothing in the numbers shows customers leaving; the switching costs of a system that holds a company's customer records are as high as ever. But the company itself now describes a future with fewer seats per customer, and its gross margin is falling as AI costs rise. A moat that has to change its shape to survive is not narrower today, but it is less certain than the one we scored in July.
Where the moat is moving. Read the acquisitions as a map and the strategy is plain. Informatica cleans and connects a company's data; Data 360 holds it; Contentful manages content; Braze sends the messages; Fin, Listen Labs and Cresta are AI agents for service, research and contact centres. If the seat shrinks, Salesforce wants to be paid for the data an agent reads and the actions it takes, wherever the agent comes from. That is a sound idea, and the data business growing 20% says customers are buying it. But it is being built at speed, with borrowed money, out of companies bought at full prices, and the reported growth of the combined business will not tell an owner for some years how much of it the old Salesforce would have earned on its own.
- That owner earnings (free cash flow after stock pay) grow about 8% a year to 2030, from about $14 a share, through modest organic growth, the shrinking share count and the acquisitions, despite higher interest costs.
- That AI changes how Salesforce charges more than how much it earns: fewer seats, more consumption, similar revenue per customer.
- That management stops at Braze and Cresta, and uses the Anthropic stake to repay debt rather than to fund the next deal.
The competition
The rest of seat-based software is being asked the same question
| Rival | Where it meets Salesforce | The latest | Read |
|---|---|---|---|
| ServiceNow | Customer service and workflow. | Q3 subscription revenue +21%; Q4 contracted-revenue guide 18.5%; shares fell 27% in a day on 29 October. | ◆ Same fear, faster growth |
| HubSpot | Smaller businesses' CRM and marketing. | Q3 revenue ~$948m, +17%; Q4 guided ~+12% (as reported); shares fell. | ◆ Slowing |
| Microsoft Dynamics 365 | CRM bundled with Office and Copilot. | Reported to have slowed to about 13% growth in the June quarter; no September figure found. | ◆ Bundled rival |
| AI laboratories | Agents that could sit on top of, or replace, CRM workflows. | Anthropic is a partner (Claude inside Salesforce and Slack), an investee and, potentially, a disintermediator. | ✗ The open question |
Across our board. Yesterday we wrote a postscript to our ServiceNow letter after its shares fell a third: nothing on our list for leaving had happened, and we said to add in stages. Salesforce sits in the same weather with a weaker position on two counts, slower organic growth and far more debt, and a stronger one on a third: at about 14 times free cash flow after stock pay, it is much cheaper. The two companies are being priced as one bet on whether AI shrinks the seat. They are not the same business, and the debt is the difference an owner should weigh most.
Management & capital allocation
A buyback on borrowed money, a deal every few weeks, and an accounting chief from the auditor
Capital allocation. In July we scored management and capital allocation at 7, praising the margin discipline that activist investors forced in 2023. We now score it 5. A buyback at about $182–198 a share, funded with $25bn of new bonds, was a large bet that the shares were cheap; at $194.52 it has not lost money, but it turned a company with almost no net debt into one with about $28bn of it, and a rating agency has reportedly marked it down. Then came about $15.5bn of acquisitions in four months, including Braze, which management itself describes as diluting profitability in the near term. The policy that won the activists' approval, margins first, has given way to a policy of buying both shares and companies at the same time.
The institutional imperative. There is an old pattern in which a company whose core business slows begins to buy growth, and the buying itself becomes the strategy. We cannot say Salesforce has fallen into it; each deal has a logic, and the core still produces fifteen billion dollars of cash a year. But the pattern has its signs: deals announced weeks apart, a target ("$63bn+ by fiscal 2030") that acquisitions make easier to reach, and an organic growth rate that each new purchase makes harder to see. A management that wanted to prove the pattern wrong would stop after Braze and Cresta, report organic growth every quarter without being asked, and say how much of the 2030 target it expects to buy.
People. Srini Tallapragada, president and chief engineering officer, stepped down in August and is an adviser until August 2027. The new chief accounting officer was, until 2021, the partner in charge of auditing Salesforce at EY. That is legal and not unusual, but an owner who values an arm's-length audit should note it. Insiders: one director, David Blair Kirk, bought about $1.0m of shares at about $239 in September; another sold about $1.2m at about $260.
The numbers
Earnings flattered by Anthropic, cash still strong, and six faults in our feed — one of which misled us
Two kinds of profit. Salesforce's reported earnings per share over the last twelve months, about $10.97, include roughly $4.7bn of non-operating gains, mostly the rise in the value of its Anthropic stake. Strip them out and the shares trade at about 26 to 29 times earnings, not the 17.7 a screen shows. Adjusted earnings for this fiscal year are guided at $16.67–16.71, of which about $2.98 is investment gains from the first half. Cash is the honest measure, and it is strong: free cash flow of $15.15bn over the last twelve months, about $18.41 a share, or $13.96 after stock pay. At $194.52 that is 10.6 times free cash flow and 13.9 times free cash flow after stock pay.
Why we subtract stock pay. Salesforce guides stock-based compensation at 9.0% of revenue this year, about $4bn. It does not cost cash, which is why free cash flow looks so large, but it is paid in new shares that would dilute owners if the company did not buy them back, and the buybacks do cost cash. The difference between free cash flow ($18.41 a share) and owner earnings ($13.96) is therefore not an accounting nicety; it is about a quarter of the cash, every year, that goes to staff rather than to shareholders. A software company that wants to be valued on its cash should be shrinking this number, as ServiceNow has begun to, not growing it.
| What the screen shows | What is wrong, and the number to use |
|---|---|
| Valuation snapshot: 950m shares, $17.2bn of debt | Dated 31 January 2026, before the March buyback and bonds. <b>This is what misled our July report.</b> Use about <b>823m</b> shares and <b>$39.3bn</b> of debt (31 July). |
| Desk: net debt/EBITDA 0.8×, interest cover 31× | Net debt is about <b>$27.9bn</b>; operating profit covers interest about <b>4.9 times</b> on the latest quarter. |
| Non-operating income −$2,694m (Q2) | The sign is flipped: it was a <b>gain</b> of $2,694m, mostly the Anthropic stake. |
| Revenue by product, FY2025: $55.0bn | Duplicated rows. Actual FY2025 revenue was <b>$37.9bn</b>; FY2026 $41.5bn. |
| Desk: 'cash conversion 899%'; share count −1.8% in five years | Free cash flow was about <b>1.9 times</b> net income in fiscal 2026; diluted shares fell about <b>16%</b>, from 974m to 821m. |
| Three quarterly dividends marked 'Irregular' | All regular quarterly payments; $0.44 a quarter in fiscal 2027. |
The dividend
Small, covered ten times, and growing about 5% — tested
Salesforce pays $0.44 a quarter, $1.76 a year, a yield of 0.9%. It began paying a dividend in 2024 and raised it by 4% in 2025 and 5.8% in 2026; the next raise is customarily announced in late February or early March. It is a small part of what Salesforce returns; the buyback is the large part.
For an income investor the dividend is not the reason to own Salesforce, and we would not pretend otherwise. Its value is as a signal. A company that has promised a rising cash payment every quarter has to keep enough cash free to make it, and that sits awkwardly beside a habit of borrowing for buybacks and paying cash for acquisitions. At about a tenth of free cash flow the dividend is safe under any outcome we can imagine. The test we care about is the next raise in late February or early March: a raise of 5% or more, after a year of heavy borrowing, would tell us the board sees the cash as secure; a token raise or none would tell us the deals come first.
| Test | Evidence | Result |
|---|---|---|
| 1. Cover on free cash flow | $1.5bn of dividends against $15.15bn of free cash flow: about 10%. | Pass |
| 2. The trend of that cover | Too young to have a trend: the dividend began in 2024. | ◆ Young |
| 3. From operations or from paper? | The dividend from operations; the $25bn buyback from new debt. | ◆ Dividend yes, buyback no |
| 4. Balance-sheet room | Net debt ~$28bn, rising towards ~$35bn with the deals; interest covered ~4.9×; rated A (reported). | ◆ Adequate |
| 5. What would force a cut | Free cash flow would have to fall by about nine-tenths. | Very unlikely |
| 6. The growth rate and its direction | +4% (2025), +5.8% (2026): modest and rising. | Pass |
★ Valuation
Cheap on its cash, with the debt now counted
Our method. Software companies pay a large part of their staff in shares, so we value Salesforce on owner earnings: free cash flow minus stock pay, per share. We grow that figure to 2030, apply an exit multiple, discount 3.2 years at 10%, and add the dividends received on the way (about $5 a share). The share count and debt are now taken from the latest 10-Q.
The debt, per share. Owner earnings are measured after interest, so the debt is already inside them; but an owner should see its size directly. Net debt of about $27.9bn is about $34 a share, a sixth of the price. At $194.52 the shares are worth about $160bn; add the net debt and the whole business is priced at about $188bn, or about 12.4 times free cash flow rather than the 10.6 a share-price screen implies. That is still cheap for a company of this quality, but it is less cheap than it was in July on the numbers we then used, and the difference is the error we described in Part I.
| Case | Owner earnings a share, 2030 | Multiple | Value today |
|---|---|---|---|
| AI shrinks the seat | $13.96 (no growth) | 12× | ~$130 |
| Slow and steady ◆ | $17.86 (+8% a year) | 16× | ~$215 |
| The 2030 plan | $20.80 ($63bn revenue) | 20× | ~$310 |
The cases. "AI shrinks the seat" freezes owner earnings at today's level: seats lost, consumption not yet replacing them, interest and stock pay rising. "Slow and steady" grows them 8% a year: about 6% organic revenue growth, a smaller share count and the acquisitions, less higher interest. "The 2030 plan" takes management's $63bn revenue target at a 35% free-cash-flow margin, less stock pay, on about 790m shares.
The middle case moves to about $176 at 13 times and $255 at 19 times; to $191 at 4% growth and $241 at 12%. What is priced in? At $194.52 and 16 times owner earnings in 2030, Salesforce must grow them about 4.5% a year: less than half of what it has done, and about what a seat-shrinking but still growing company would manage. The price assumes a modest outcome. That is the good news.
Margin of safety. The price is about 10% below our central value. For a business we now score lower on management and balance sheet, that is a discount, not a cushion. We would continue to buy in smaller amounts than we suggested in July, because the company is more indebted and the AI question is less settled, and we would add more readily below $165: about 12 times owner earnings and an 11% free-cash-flow yield, close to where we first recommended it.
Risks, lawsuits & controversies
Verified 18 November 2026 against the Q2 FY2027 10-Q and attributed reports
Legal. Salesforce's 10-Q of 27 August 2026 describes securities litigation over Slack's 2019 listing (a state-court stay was partly lifted in November 2025) and several suits that allege Salesforce software facilitated the Backpage website; one claim was dismissed in January 2026 and another case is stayed pending criminal proceedings. Salesforce does not expect a material adverse effect. Authors have filed class actions alleging that pirated books were used to train Salesforce's AI models (Northern District of California, as reported). Class certification was denied on 4 September 2026 in a chat-privacy suit. None of these has been quantified.
Ratings. S&P was reported to have cut Salesforce to A from A+ on 4 November 2026, and Moody's to A2 in March, citing acquisitions and shareholder returns; we did not confirm either at the agencies.
| Test | Today | What would change our mind |
|---|---|---|
| Q3 results, 2 Dec | Organic ~6.4% (Q2) | Organic growth back above 8% would support the middle case; below 6% would lean towards the AI-shrinks-the-seat case. |
| cRPO growth | +14% | Holding at 13–14% says the contracts are still growing; a fall below 11% would be the first sign of seats leaving. |
| Buyback settlement (Q3 10-Q) | ~103m shares delivered | A final average price near $182 would mean the debt bought shares cheaply; much higher would weaken the case for it. |
| Anthropic stake | ~$5.1bn in the books (July) | Selling part of it to repay debt would restore some of the balance sheet we wrongly described in July. |
| More acquisitions | ~$15.5bn since July | Another large deal before Braze closes would lower our management score again. |
| The price | $194.52 | Below $165 with cRPO intact: add more readily. |
In July I told you to buy the fear in Salesforce at a hundred and fifty-seven dollars. You would have made sixty-nine per cent by early September, and you have made twenty-four per cent today. I would like to take the credit. I can take only part of it, because the report I wrote had a hole in it that I must show you before anything else.
I described a company with about ten billion dollars of net debt and interest covered twenty-one times. In March, three months before I wrote, Salesforce had borrowed twenty-five billion dollars and spent it on its own shares in a single transaction. I took the balance sheet from a snapshot in our data that predated it. The company I praised for its light balance sheet in fact owed about twenty-eight billion dollars net, and covered its interest about five times. The shares rose anyway, but you were entitled to the right numbers, and I should have read the quarterly filing instead of trusting the screen. I have changed how we work so that it does not happen again.
Now the business, which has been better than its critics say and worse than I said. Contracted revenue is growing faster than a year ago, customers are not leaving, and the company produced fifteen billion dollars of free cash in twelve months. But growth without acquisitions is about six per cent, not eleven; since July management has agreed about fifteen and a half billion dollars of further purchases; one rating agency is reported to have marked the company down; and Salesforce's own presentation now describes customers using fewer seats as they adopt its AI agents. On 17 November the market remembered all of this at once and took nearly a tenth off the price in a day.
At a hundred and ninety-five dollars you pay about fourteen times the cash the business produces after paying its people in stock. That price assumes owner earnings grow only four or five per cent a year, which a company shrinking its seats but selling more to each customer could do. I value it at about two hundred and fifteen dollars on steady progress, and about three hundred and ten if management's 2030 plan comes true. The discount is real. It is smaller than the one I thought I was buying in July, because the debt is real too.
So: keep buying, but in smaller amounts than I suggested in summer, and buy more readily below a hundred and sixty-five. Watch the third-quarter report on 2 December for organic growth and contracted revenue, and watch what Salesforce does with its Anthropic shares. If it sells them to pay down the debt it took on in March, it will have turned a lucky investment into a stronger balance sheet, and I will raise my management score. If it uses them to buy another company, I will lower it again.
— The Buffett Lens. Read the latest filing before you praise a balance sheet; the screen is a summary, not a witness.
Add Below $165A deeply embedded cash machine, cheaper on its cash than in a decade, now carrying debt and an acquisition habit our July report understated. ★ At $194.52, 13.9× owner earnings, the price is about 10% below our central value of ~$215, within a range of ~$130 to ~$310. Keep buying in smaller amounts; add more readily below ~$165. Q3 results on 2 December.
