Salesforce

The thesis in brief#

Salesforce is the world's largest customer-relationship-management (CRM) software company. It sells businesses the software they use to find, sell to, serve, and stay in touch with their customers, delivered over the internet as a subscription. It more or less invented that model, and it has compounded revenue from roughly $21bn (FY21) to $41.5bn (FY26).

The thing to hold in your head, as a first-time reader, is that two stories are running at once:

• The profit story is genuinely strong. Margins are expanding fast, free cash flow is enormous (close to 60% of revenue in the latest quarter), and net income has gone from a near-zero trough in FY23 to $7.5bn in FY26.
• The growth story is maturing. Organic revenue growth has drifted down toward 12% and below. Management is asking investors to trust a second-half re-acceleration that has not happened yet.
• The new bet is AI agents. "Agentforce" is the product Salesforce is staking its next decade on. It is growing very fast (over 200% year on year) but is still under 3% of revenue, so it is real momentum off a tiny base.
• The balance sheet just changed character. Salesforce borrowed $25bn to fund a record share buyback, moving from almost no net debt to roughly $27bn of it in a single quarter.

If you remember one line: a high-quality, cash-rich software compounder, getting much more profitable, whose growth durability and reliance on financial engineering are the open questions.

The business model: how the products connect#

Most write-ups stop at "Salesforce sells CRM software." That misses the point. The reason the company is hard to dislodge is that its products are not a shelf of separate tools, they are layers of one system that pass data to each other. To see why that matters, start with what the company actually does, then walk up the stack.

What a CRM is, in one sentence. A CRM (customer-relationship-management) system stores everything a company knows about its customers, who they are, what they bought, every email and support call, in one place, and gives the sales, service, and marketing teams the tools to act on it. Think of a bank: before CRM, your account history sat in one department's files, your last complaint in another, and your loan application in a third, so no single employee saw the whole picture. A CRM puts all of it on one screen. Salesforce pioneered delivering this over the web instead of installing it on each company's own computers, and that is how it grew into a giant.

The business model is software-as-a-service (SaaS, meaning you rent access online rather than buy and install the software once). Salesforce charges recurring subscriptions, usually priced per employee who uses it ("per seat"). This single fact drives how the market values the company. Recurring revenue is predictable: because customers keep paying and rarely leave, this year's revenue more or less locks in a big chunk of next year's. Investors pay a premium for that predictability, in the same way a landlord with signed leases is worth more than a shop relying on walk-in traffic.

The money arrives two ways:

Stream	Latest quarter	Share	What it is
Subscription & support	$10,593m (+14%)	~95%	Recurring fees to use the software. Sticky, predictable, the engine.
Professional services & other	$540m (+1.5%)	~5%	One-off fees to set up and customise. Low margin and lumpy; kept small on purpose.

The product stack, as one connected system#

Salesforce sells across six product families, but the right way to picture them is as a stack where each layer feeds the one above it. Imagine an airline that runs on Salesforce, and follow one passenger through the layers:

1. The application clouds (Sales, Service, Marketing, Commerce Cloud). These are the screens employees actually work in. The booking sits in Commerce Cloud, the loyalty offer is sent from Marketing Cloud, and when the passenger's bag is lost the agent handles the complaint in Service Cloud. Sales Cloud and Service Cloud are the two mature engines, each roughly a $10bn business on its own. This is where most revenue is earned today.
2. The data layer (Data 360, formerly Data Cloud). Each cloud above would normally keep its own copy of the customer, so the airline would "know" the same passenger three different ways. Data 360 unifies all of it into one profile, so the booking, the loyalty status, and the lost-bag complaint describe one person, not three. This layer is the quiet linchpin, because the layer above it is only as smart as the data it can see.
3. The intelligence layer (Agentforce, built on the older Einstein AI). This is where AI agents read the unified profile and act: an Agentforce agent can see the passenger is a frequent flyer whose bag was just lost, apologise, rebook the bag, and offer miles, without a human touching it. This is the layer Salesforce is staking its next decade on.
4. The collaboration layer (Slack). When the AI cannot finish a task, it hands off to people. Slack is the chat app where the human agent, the AI agent, and the operations team coordinate the fix. Salesforce is repositioning Slack as the "front door" where humans and AI agents work side by side.

Underneath all of this sits the plumbing, MuleSoft, Tableau, and Informatica (acquired integration, analytics, and data-management tools), which connects Salesforce to the airline's other systems, cleans the data, and turns it into dashboards. Unglamorous, but it is what lets the data actually flow between the layers.

Layer	Product	How it makes money	Why it is in the stack
Apps	Sales Cloud, Service Cloud, Marketing, Commerce	Per-seat subscriptions	The mature engines; ~$10bn each for the two biggest, most revenue today
Data	Data 360	Subscription to unify data	The linchpin; makes the AI monetisable by giving it clean, joined-up data
Intelligence	Agentforce (on Einstein)	AI subscriptions + usage credits	The central growth bet: agents that act, not just suggest
Collaboration	Slack	Per-seat subscriptions	Where humans and AI agents hand work back and forth
Plumbing	MuleSoft, Tableau, Informatica	Integration / analytics / data subscriptions	Moves, cleans, and visualises the data between layers

Why this design is the whole argument. Growth runs on a "land and expand" playbook: a customer buys one cloud, then over years adds more layers, more seats, and higher-priced tiers. Once an airline runs sales, service, data, and AI on Salesforce, ripping it out means re-plumbing the entire company, so customers stay and buy more instead. A telling data point for the AI story: more than half of recent AI bookings came from existing customers, which is exactly this expansion machine doing its job. A standalone AI tool, however clever, has to recreate all four layers plus the plumbing before it can compete, which is far harder than building one neat feature.

The bear view on the same design. The connected-stack story is not free of holes, and a skeptic would press three points. First, much of the stack was bought, not built (Slack, MuleSoft, Tableau, Informatica, Data Cloud lineage), so the "seamless system" is in places still being stitched together, and integration debt is real. Second, bundling can mask weakness: Sales and Service are healthy, but Marketing, Commerce, and Tableau have been visibly soft, and selling them inside a bundle can hide products that would struggle to win on their own. Third, the same data-gravity that locks customers in also draws the sharpest competitor: Microsoft sells a rival stack (Dynamics plus Teams plus Power Platform) bundled into software many of these companies already pay for, so Salesforce's integration moat is being attacked by an even larger integration moat. The investor's question is whether owning every layer is a durable advantage or an expensive way to defend a core (Sales and Service) that two products already provide.

Financial performance#

Salesforce has roughly doubled revenue in five years while transforming its profitability. The chart below is the full income statement, in US dollars. Use the toggle to switch between the annual picture (the long compounding story) and the quarterly picture (the recent trend, where you can see growth maturing and margins lifting). Each cell in the table also shows the year-on-year change.

A few things to read for in the data. Revenue growth has slowed in percentage terms as the base got huge, which is normal. The profit lines tell the more dramatic story: operating income was almost nothing in FY21 to FY23 (FY23 was a restructuring year, with GAAP net income of just $208m), then stepped up sharply from FY24 as the company cut costs and held headcount flat. Net income in FY21 looks high relative to operating income because it was flattered by large one-off gains on strategic investments, a reminder to read the operating lines, not just the bottom line.

The latest quarter in focus#

Numbers age, so treat this as the snapshot at the time of writing rather than a live figure. It is here because comparing the latest quarter with the same quarter a year earlier is the fastest way for a new reader to understand what is actually changing at the company.

The competitive moat#

Salesforce's durability comes from several reinforcing layers, not one feature:

1. Switching costs. Once sales, service, and customer data run on Salesforce, replacing it is expensive, slow, and risky. Customers stay and expand rather than rip and replace.
2. The system of record. Salesforce holds the authoritative customer data. AI is only as good as the data it can see, so owning the data layer (Data 360) is what lets Salesforce charge for AI rather than be disintermediated by it.
3. The integrated chain. Apps plus data plus the Slack interface make the whole more useful than the parts. A stand-alone AI tool has to recreate all of that to compete.
4. Ecosystem and custom apps. A large network of system integrators and customer-built apps deepens reliance and turns partners into a distribution machine.

The threats are real, though. The central one is that AI lets companies build their own tools and lean less on Salesforce. Management's counter is that customers consume more, not less, because Salesforce holds the trusted data and workflows the AI needs. There is also product softness today in Tableau, Commerce Cloud, and Marketing Cloud, and bigger rivals (Microsoft above all) push hard on the same customers.

Key risks and what to watch#

The honest risk list, grounded in the latest results:

• Growth durability (high). Organic growth (revenue growth stripping out acquisitions, so the true underlying pace) is drifting toward 12% and below, and next quarter's guide implies a slowdown to roughly 6-7% organic before the promised second-half pickup. The whole bull case rests on a re-acceleration that has not arrived.
• Cash-flow growth cut (high). Management lowered full-year operating and free-cash-flow growth guidance to about 4-5%, because interest on the new debt eats cash. This is the clearest genuine negative, easy to miss behind the strong headline cash number.
• Leverage (medium). Long-term debt jumped to $39.3bn from about $10.4bn to fund the buyback. Debt adds fixed costs and reduces flexibility in a downturn.
• AI monetisation unproven (medium). Agentforce is under 3% of revenue, the growth is off a tiny base, and the link from soaring AI usage to actual dollars has not been fully shown.
• Earnings quality (watch). A one-off $558m gain on strategic investments flattered headline GAAP earnings in the latest quarter. Focus on operating income, which is the truer guide.

What to watch next: does cRPO (current remaining performance obligations, the value of signed contracts due to be billed within twelve months, so a leading demand signal) hold around 14%; is there real evidence of organic acceleration rather than just the promise repeated; does the strong free cash flow get used to pay down the new debt; and do the weak products stabilise or drag harder.

Scenario analysis: bull, base, and bear#

Three coherent ways to read the same company. None is "the answer"; they map the range a reasonable investor could land on.

Linked Notes#

• Microsoft is the most direct large-cap rival across the enterprise stack.
• ServiceNow is the other enterprise-workflow compounder in the book, with a similar switching-cost moat.
• See 03 Themes for the AI-software theme and 04 Portfolio Construction for sizing.