What it does · how it got here · where it operates
Let me tell you what Salesforce actually is, because the name undersells it. Salesforce runs the software that companies use to keep track of their customers — every lead, every sale, every support ticket, every marketing email. When a salesperson logs a deal, when a call-centre agent pulls up your account, when a marketer sends you an offer, there is a very good chance it all runs through Salesforce. It is the system of record for how the world's businesses manage the people who pay them — and once a company builds its operations on top of it, ripping it out is like re-plumbing a skyscraper while people are still working inside.
Founded in 1999 by Marc Benioff with a then-radical idea — that business software should be rented over the internet like a utility, not installed on your own computers — Salesforce more or less invented the software-as-a-service industry. It grew into a suite of "clouds" (Sales, Service, Marketing, Commerce, Data, plus Slack, Tableau and MuleSoft, mostly bought along the way). In the year to January 2026 it booked $41.5 billion of revenue and around $7.5 billion of profit, at a gross margin near 78%. And here is why this report exists: after a run to $277, the stock sits near $157 — roughly 43% below its high — because the market is afraid that artificial intelligence will unravel the very thing it sells.
"The most important thing to do if you find yourself in a hole is to stop digging." — which is roughly what Salesforce's board finally did in 2023, when the activists arrived
Two things happened at once to bring the stock here. First, growth slowed — from the 20–30% of its youth to about 9–10% today, as the business matured. Second, the arrival of powerful AI made investors ask an uncomfortable question: if software can now do the work of the humans who buy Salesforce seats, does the world need as many seats? That fear — not any collapse in the business — is why a franchise this good trades at a price this modest. The whole of this report is an attempt to weigh it honestly.
| Founded | 1999 · San Francisco, USA · by Marc Benioff |
| Sector / Industry | Technology · Application Software (SaaS) |
| What it is | The #1 CRM — the enterprise system of record |
| Key products | Sales, Service, Marketing & Data Clouds · Slack · Tableau · Agentforce |
| Revenue (FY2026) | $41.5 B |
| Market capitalisation | ~$149 B |
How subscription software actually makes money — and whether you'd understand it
Software is nearer the edge of my circle than a bottle of Coke — but the money here is simple, and worth walking through end to end:
Here is the beauty of the model and the reason cash flow runs ahead of reported profit: customers pay for a year up front, so the cash lands before the revenue is "earned," and the company sits on a growing cushion of deferred revenue — a bit like insurance float. Add negligible cost to serve one more customer, and you have a machine that, once mature, gushes cash. That is exactly what Salesforce has become. The question an owner must answer is not "how does it make money?" — that is plain — but "will the customer still be locked in, and still adding seats, ten years from now?"
The clouds · geographic mix · the size of the pond
Revenue comes from a family of "clouds," each a subscription business, sold to the same customers again and again. Here is the mix (fiscal 2026).
Salesforce is comfortably the #1 CRM in the world — by most counts larger than its next several rivals put together — in a market that still grows in the mid-single digits as businesses digitise. The pond is not draining like tobacco's, nor rising like Visa's; it is a large, mature, still-growing lake in which Salesforce is the biggest fish. The genuinely new water is AI-agent software, which could either be a fresh river feeding the lake or, if it lets rivals rebuild CRM from scratch, a leak in the dam. I treat that squarely in Part V.
Switching costs, an ecosystem, and the data itself
Salesforce's moat is one of the clearest examples of a switching-cost fortress in all of business — the kind Buffett prizes because the customer wants to stay put. It is filled from three springs.
Switching costs — the deep one. When a company runs its sales, service and marketing on Salesforce, its most precious asset — its customer data — and its daily workflows live inside the system. Migrating to a rival means re-entering years of data, rebuilding hundreds of custom processes, and retraining thousands of employees, all while risking a stumble in the business that pays the bills. Most firms would sooner pay the annual bill and grumble. That is why renewal rates run high and customers keep adding seats and clouds — the "land and expand" that compounds revenue.
The ecosystem. A vast network of certified administrators, developers, consultancies and third-party apps (the AppExchange) has grown up around Salesforce. Careers are built on knowing it; partners make their living implementing it. That human and commercial web is a moat no newcomer can conjure.
The data. The system of record accumulates a company's entire customer history — and in an AI era, whoever holds the proprietary data is best placed to sell the intelligence built on top of it. That is the bull case for AI as friend, not foe.
How durable? Wide, but — for the first time — contested. The switching-cost moat is genuinely deep and not going away next year. But the same AI that could let Salesforce sell more could also, over a decade, let a rival rebuild a leaner CRM natively around AI, or let customers need fewer human seats. The moat is wide and, for the first time in its history, being tested. Watch the trend, not just the width.
Agentforce · is artificial intelligence Salesforce's friend or its undertaker?
This is the whole ballgame, and the reason the stock is cheap, so let me put both sides plainly rather than cheerlead. The fear can be stated in one sentence: Salesforce charges by the human seat, and AI's promise is to need fewer humans. If artificial intelligence lets one support agent do the work of five, a company might buy one Service Cloud seat where it once bought five. That is a real threat to a per-seat model, and it is why the shares de-rated so hard.
Which is right? Honestly, nobody knows yet — and any analyst who claims certainty is selling something. My own read, held loosely, is that the switching-cost moat and the data advantage make Salesforce far more likely to capture the AI wave than be drowned by it — you do not casually rebuild the system that runs a Fortune 500's customer base, AI or no AI. But this is precisely the kind of ten-year durability question that keeps a business at the boundary of my circle of competence, and it is why I demand a cheap price rather than a fair one before getting interested. The good news, which the next parts show, is that the price is cheap.
Benioff the visionary · the M&A habit · the activists · the turnaround
Here the story is genuinely interesting, and I must be even-handed: Salesforce's management is both a visionary founder and a recovering practitioner of the one sin Buffett warns about most — the empire-building itch.
By early 2023, after the $27.7 billion Slack deal and years of "growth at any cost," Salesforce earned a GAAP operating margin of just 3% despite $31 billion of revenue — a franchise being run for size, not for owners. This is the institutional imperative Buffett describes: the herd expansion, the size-worship, the sense that a CEO must always be doing a deal. It took an extraordinary gathering of activist investors — Elliott, Starboard, ValueAct, Third Point and others, all at once — to force the correction. And correct it did: costs were cut, the workforce trimmed, buybacks ramped, a first-ever dividend was declared in 2024, and the GAAP operating margin climbed from 3% to ~21% in three years. It is one of the more dramatic disciplinings of a great business in recent memory.
So how do I score it? Integrity: no red flags — Benioff communicates openly, though the heavy reliance on "adjusted" non-GAAP numbers earns a raised eyebrow. Owner mentality: high — a founder with his fortune on the line. Capital allocation: improving, but watch it like a hawk. The margin turnaround is to their credit; the M&A history, and a fresh ~$8 billion deal for Informatica in 2025 just as the activists eased off, are the reasons I would never assume the empire-building itch has been permanently cured. The best thing management can do for owners now is exactly what it is finally doing: run the wonderful business it already owns, and return the cash.
Sourced from your endpoints · FY ends 31 Jan · linked to the data pages
| Metric | Value | Read |
|---|---|---|
| Revenue growth (FY2026) | ~10% | ◆ decelerated from 20-30% |
| Revenue (FY2017 → FY2026) | $8.4B → $41.5B | ▲ ~5× |
| Gross margin | 77.6% | ▲ classic software |
| Operating margin (GAAP) | ~21% | ▲ up from 3% in FY23 |
| Free cash flow / yield | ~$14B · ~10% | ▲ gushes cash |
| FCF vs. net income | $16.9 vs $9.2 / sh | ▲ cash beats profit (D&A, deferred rev) |
| Stock-based comp / revenue | ~8% | ◆ a real cost — dilutive |
| Return on equity (ROE) | 14.9% | ◆ solid |
| Return on invested capital | 9.2% | ◆ dragged by M&A goodwill |
| Balance sheet | net debt ~$10B | ▲ low · interest covered ~21× |
| Capex / revenue | ~1.3% | ▲ wonderfully asset-light |
The numbers tell a story of a business that grew up. Growth slowed, yes — but in exchange the company learned to convert that growth into profit and cash: a gross margin near 80%, a GAAP operating margin that quadrupled to ~21% in three years, and free cash flow of roughly $14 billion on almost no capital. Note the honest wrinkles I have flagged in amber: free cash flow (~$16.9/share) runs well above reported earnings (~$9.2/share) because of heavy amortisation of acquired intangibles and that upfront subscription cash — flattering, but partly real; and stock-based compensation of ~8% of revenue is a genuine cost to owners that the cash-flow figure ignores. Strip both out and the underlying cash economics are still excellent — and, as the next part shows, cheaply priced.
Cheap on its own history · DCF and the Street both well above · margin of safety
| Yardstick | Today | Its own history | Read |
|---|---|---|---|
| P/E — reported earnings (TTM) | ~17x | often 40-60x | cheap vs. its past |
| P/FCF — free cash flow | ~10x | 20-30x+ | strikingly cheap |
| Free cash flow yield | ~10% | — | rich for a leader |
| Price / sales | ~3x | 8-10x | a fraction of its bubble multiple |
| PEG (P/E ÷ growth) | ~0.5 | — | cheap vs. growth |
| Dividend yield | ~1.4% | new (2024) | small, plus buyback |
Look at what the fear has done to the price. A business that spent most of its life at forty to sixty times earnings and eight to ten times sales now trades at about seventeen times earnings, ten times free cash flow, and three times sales. On my preferred honest measure — free cash flow less the real cost of stock compensation — you are paying roughly twelve times owner earnings for the #1 franchise in enterprise software. A conservative DCF lands near $253, and the analyst crowd (rarely a brave bunch) sits at ~$266 — both roughly 60–70% above today's price. When the market offers you a wonderful business at a price this far below every reasonable estimate of value, it is usually paying you for a specific fear. Here the fear has a name — AI — and the honest investor's job is to decide whether ~40% is enough discount to cover it.
AI disruption · deceleration · the M&A itch — with sources
The dominant risk is the one that made the stock cheap: AI. If autonomous agents compress the number of human seats companies buy, or let a rival rebuild CRM natively, the per-seat model is pressured — this is a genuine, decade-long structural question, not a passing worry, and I have marked it accordingly. Beneath it sit the ordinary ones: growth that keeps decelerating, a Microsoft that bundles a capable rival (Dynamics) with its Copilot AI and its Office grip, stock-comp dilution of ~8% of revenue, and a management M&A habit that a fresh ~$8B deal shows is not fully cured. None of these is likely to break the franchise this year — but together they are why a business this good is on sale, and why I insist on the cheap price rather than a fair one.
Every so often the market hands you a wonderful business at a distinctly un-wonderful price, and attaches to it a fear you must weigh with your own head rather than borrow from the crowd. Salesforce today is exactly that, and I want to separate — as always — the quality of the business from the price on offer, and then confront the fear head-on.
On quality, there is little to argue. Salesforce is the system of record for how the world's businesses manage their customers, and that is one of the deepest switching-cost moats in all of commerce. A company that runs its sales, service and marketing on Salesforce has poured its most precious asset — its customer data — and years of custom workflows into the system; asking it to leave is asking it to re-plumb the building while the tenants work. That is why customers renew almost automatically and keep adding seats, why the gross margin sits near eighty per cent, and why — now that management has been taught discipline — the thing gushes some fourteen billion dollars of cash a year. This is a genuinely wonderful franchise.
On management, I will be even-handed. Marc Benioff built this from nothing and owns billions of it — a real owner, not a hired hand. He also spent years doing exactly what I warn against: buying expensive companies because the institutional imperative whispers that a chief executive must always be doing a deal. It took a rare mob of activists, all arriving at once, to stop the digging — and to their credit the company listened, quadrupling its operating margin, buying back stock and paying its first dividend. I applaud the reformation. I also note the fresh eight-billion-dollar acquisition that followed the moment the activists relaxed, and I keep one hand on my wallet.
Now the fear, which I will not wave away, because it is the whole reason we are offered this price. Artificial intelligence threatens a business that charges by the human seat, for the plain reason that AI's promise is to need fewer humans. That is real. But set against it is this: Salesforce holds the enterprise's proprietary data and workflows — the very fuel the new AI agents run on — and it is selling those agents, priced by the job rather than the seat, on top of a moat no startup can cheaply rebuild. My honest judgement, held without false certainty, is that this franchise is far likelier to ride the AI wave than to drown in it. But I do not know that, and neither does anyone else, which is precisely why I will not pay a fair price — only a cheap one.
And cheap it is. A business that traded most of its life at forty-to-sixty times earnings now sells for about seventeen times earnings and ten times the cash it throws off — roughly twelve times what I'd call its honest owner earnings — while a sober appraisal of the cash flows, and even the timid analyst crowd, sit sixty to seventy per cent higher. That is a wide margin of safety wrapped around a first-class franchise, and the discount exists for a nameable reason rather than a rot in the business. So my counsel is this: for the investor willing to hold ten years and to keep a close eye on how the AI story actually unfolds, I would begin accumulating here, near the lows, and buy with more conviction on any further fright that drags it beneath the mid-hundred-and-forties. Size it as a good business bought cheap during a fog — not a certainty, but the kind of asymmetric bargain that, bought at the right price, tends to reward the patient. Remember the oldest rule: it is far better to buy a wonderful company at a fair price — and better still, on the rare day the crowd is frightened, to buy it at a cheap one.