Durability

Durability

Salesforce has grown revenue in each of the last ten fiscal years ($8.4B to $41.5B) and free cash flow in each of them ($1.7B to $14.4B), so the framework's structural-decline disqualifier is not tripped. But the conviction sources the year-10 gate rewards mostly do not apply here: the market is fragmented and low-barrier by the company's own account, there is no regulatory or capital-intensity moat, and the seat-based model faces a live AI-agent threat the 10-K itself concedes could cut demand. The gate turns on that doubt.

What the year-10 gate asks

Ruchir's one pure gate is binary by construction: year-10 revenue and adjusted free cash flow higher than today, held with very high conviction — and any proper doubt fails it. Conviction is supposed to come from structural sources (market structure, regulatory barriers, capital intensity, essentialness, long history), not from execution. This tab grades each source for Salesforce specifically, hunts the structural threats, checks the disqualifier flag, and states the read once. It builds on the market-position work in Business rather than repeating it.

The disqualifier check — revenue trajectory

The framework's one mechanical exclusion is revenue declining high-single-digit for three consecutive fiscal years after a long existence. Salesforce's revenue_trajectory feature records consecutive_decline_years of 0 and three_year_hsd_decline of false: revenue has risen every year in the corpus.

Source: FY2026 Annual Report (Form 10-K), Consolidated Statements of Operations, and prior-year 10-Ks; per fit_features.revenue_trajectory [1].

The disqualifier is not close to firing. What the chart also shows, and what matters for the gate, is deceleration: year-over-year growth fell from roughly 29% in FY2020 to 8.7% in FY2025 and 9.6% in FY2026.

Source: derived from reported revenue, FY2018–FY2026 10-Ks; per fit_features.revenue_trajectory [2].

The level is rising; the rate is compressing toward high single digits. Both facts belong in the year-10 read below.

The conviction sources, graded for this company

Salesforce is the #1 provider of CRM software globally and has been public and profitable in cash terms for years [3]. But leadership is not the same as the structural protection the gate rewards, and most of Ruchir's conviction sources are tuned for capital-heavy regulated essentials — the opposite archetype to asset-light software.

Sources: FY2026 10-K Item 1 Business [4]; Item 1A Risk Factors [5]; capex from the cash-flow statement [6]; customer concentration [7].

The one source that clearly applies is essentialness. Salesforce is a system of record: customers "have invested substantial personnel and financial resources to implement and integrate their current enterprise software" and are "reluctant or unwilling to migrate away," in the 10-K's own words [8]. Data gravity and workflow integration are real switching costs, and customer concentration is negligible — no single customer reached 10 percent of revenue in FY2026 or FY2025 [9].

The sources that do not apply are the ones the framework prizes most for surviving the AI world. There is no monopoly or regulated oligopoly: the company describes the market for enterprise applications as "highly competitive, rapidly evolving, fragmented and subject to changing technology, low barriers to entry" [10], and names competitors with "greater name recognition, longer operating histories, more significant installed bases" — the large incumbents (Microsoft, Oracle, SAP, Adobe, ServiceNow) plus more agile smaller players. There is no regulatory barrier to entry, and capital intensity — the framework's shelter for "capital-heavy essentials" that survive even when they are not the best businesses — runs the wrong way: capex was about 1.4% of revenue in FY2026 [11]. What is left is switching-cost strength, which sits close to product and execution — and the framework is explicit that execution is not a moat.

The structural threats, hunted

The threat that matters for the year-10 clock is the one Ruchir frames as "is anyone's margin here an Amazon opportunity?" For Salesforce it has a name: AI agents against the seat-based licensing model.

Salesforce sells "primarily on a subscription basis" [12], and much of that revenue is priced per human seat. The 10-K states the risk plainly: as AI is built into the market, "New AI offerings may disrupt our service offerings or transform workforce needs and negatively impact demand for our offerings, or our competitors may be able to incorporate AI into their offerings more efficiently or successfully than we are able to" [13]. The company reinforces the point in its dedicated AI risk factor, noting it is "increasingly building AI into many of our offerings, including generative and agentic AI," with attendant adoption risk [14].

The mechanism, quantified. If AI agents reduce the number of human service and sales agents an enterprise needs, a per-seat CRM faces volume, not just pricing, pressure: fewer licensed users in the service cloud. For the revenue level to keep rising, consumption revenue from AI work (Agentforce, priced per task rather than per seat) has to more than offset any seat erosion. External research puts numbers on the risk: Gartner has projected that roughly 35% of point-product SaaS tools could be displaced by AI agents by 2030, and in May 2026 a Bank of America analyst reinstated an Underperform rating with a $160 target, calling the shift an AI-driven structural reset (external analyst estimates, dated May 2026; not sourced to the filings). This is not a filings-visible fact, so it is carried as context, not as a cited number — but it is the fear repricing the stock, and it is structural rather than cyclical.

The threat is genuine and unresolved, but it is bounded by three facts. Salesforce is itself the largest vendor of the agentic tools doing the disrupting; its pricing is already migrating toward consumption; and the switching costs above make wholesale replacement slow. The plausible year-10 outcome is therefore compression of the growth rate and a mix shift in how revenue is earned, more than a decline in the revenue level — but the range of outcomes has widened, which is exactly what erodes "very high conviction."

Other threats are lower. Customer concentration is negligible [15]. Regulatory reversal is not a material lever for a horizontal software vendor. The subscription base must still be renewed — customers "have no obligation to renew their subscriptions… typically 12 to 36 months, and in the normal course of business, some customers have elected not to renew" [16] — but that renewal risk is the ordinary texture of the model, not a structural break.

FCF consistency (P2)

The deterministic adjusted-FCF stability series is not computable. fit_features.fcf_stability is empty and adjusted_fcf.latest_adjusted is null because stock-based compensation is absent from the cash-flow data feed for every year FY2017–FY2026, so the profile could not build the rolling five-year adjusted-FCF average (not_computable.fcf_stability: "fewer than five consecutive adjusted-FCF years"). That limitation is stated rather than filled by mental math.

On a reported free-cash-flow basis, the series is unusually consistent — it rose in every one of the last ten years, with no negative episodes.

Source: FY2026 Annual Report (Form 10-K), Consolidated Statements of Cash Flows, and prior-year 10-Ks; per fit_features.adjusted_fcf.series[].fcf [17].

This is a recurring-revenue subscription profile, not the underwriting-cycle pattern Ruchir tolerates in insurers and banks: there is no 5-to-8-year negative cadence to explain because there are no negative years. On the reported basis, predictability is high. The caveat is real and one-directional — the reported figure is before the roughly $3B-plus of annual stock-based compensation that the adjusted metric would subtract; the Yield tab carries the adjusted-yield arithmetic, and the true adjusted-FCF level is materially lower than the reported line. But nothing in the record suggests the adjusted series would be unstable; it would simply be lower and, given SBC's own steadiness, similarly smooth.

The year-10 case, both ways

The strongest case that year-10 revenue and FCF are higher. Ten consecutive years of rising revenue and rising free cash flow is a powerful base rate. Salesforce is the category leader in a still-growing market [18], sits inside its customers' systems of record with deep switching costs [19], carries no customer concentration [20], and is the largest supplier of the very agentic tools said to threaten it. Consensus, tracked in Yield, still has free cash flow rising through FY2029 (fit_features.consensus_forward_yield). Even at high-single-digit growth, revenue and FCF a decade out are, on the balance of evidence, more likely higher than lower.

The strongest doubt. The conviction the gate demands is supposed to be structural, and here it mostly is not. The company itself calls its market fragmented and low-barrier [21]; there is no regulatory shelter and, at about 1.4% of revenue in capex, no capital-intensity moat [22]; the operating history is shorter than the framework prefers. What protects the franchise is switching-cost strength, which is closest to execution — and execution is not a year-10 moat. Against that thinner structural base sits a live, named threat the filing concedes could "transform workforce needs and negatively impact demand" for the seat-based model [23].

The read. On base rates, year-10 revenue and free cash flow are more likely higher than not, and the structural-decline disqualifier is clearly not tripped. But the gate does not ask whether the outcome is likely; it asks whether it is held with very high conviction, and it fails on any proper doubt. That conviction is not available here: the moat rests on switching costs alone, without the market-structure, regulatory, or capital-intensity protection the gate rewards, and the AI-agent threat to the seat model is exactly the kind of "your margin is my opportunity" doubt the framework treats as decisive — genuine, current, and unresolved. On the framework's binary standard, there is a genuine doubt, and the gate does not clear.