Chapter 3

Moat and AI

FactSet's advantage is measurable in one place above all: subscription retention. Annual ASV retention has held above 95% every year from FY2020 through FY2025, with client retention in a 90–92% band through a rate shock and a growth slowdown [1]. That durability, plus pricing power and deep workflow embedding, is a real moat — but a narrow one built on switching costs, not scale. Generative AI, the risk the de-rating priced, now reads on the evidence as more tailwind-at-the-margin than existential threat, with early but small monetization.

The moat is retention, not size

FactSet ended FY2025 with 8,996 clients and 237,324 users, retaining more than 95% of annual subscription value and roughly 91% of clients [2].

Source: FactSet FY2025 Annual Report (Form 10-K), Item 1 Business — ASV retention stated as "greater than 95%" [3].

What makes the number a moat rather than a snapshot is that it barely moves. Across six fiscal years spanning the 2022 rate shock, the banking hiring slump, and the organic-growth trough covered in the Financial Record, client retention stayed within two percentage points of 91%, and ASV retention never fell below 95%.

Source: FactSet FY2021–FY2025 Annual Reports (Form 10-K), MD&A; ASV retention stated as "greater than 95%" in each year [4].

The mechanism behind the stickiness is embedding, not novelty. Management describes FactSet as running performance, attribution, and risk analytics for over 6 million institutional portfolios every night, and integrating more than 15 million wealth portfolios, used across clients' front, middle, and back offices [5]. Ripping that out means re-plumbing nightly production workflows, which is why the annual price increase has been a recurring driver of organic ASV in every 10-K — in FY2025, sales to existing clients and price increases lifted organic ASV 5.7% to $2,370.9 million [6]. By Q2 FY2026 management reported the annual Americas price increase contributed more than the prior year, attributing it to value, retention, and enterprise-agreement escalators [7].

The moat defends share; it does not confer scale

FactSet names its largest competitors as Bloomberg L.P., S&P's Market Intelligence division, and LSEG's Data and Analytics division (formerly Refinitiv), with BlackRock Aladdin, MSCI, and Morningstar as further competitive products [8]. Against that field, FactSet is small. Its $2.32 billion of FY2025 revenue is below Morningstar and MSCI, and a fraction of the diversified majors.

Sources: FactSet FY2025 10-K, competitor naming [9]; peer revenue per reported FY2025 annual results. Bloomberg is privately held and larger; LSEG's Data and Analytics division is also larger and is excluded from the bar because the indexed LSEG filing is a sub-entity.

Two cautions temper the chart. S&P Global and Moody's earn much of their revenue from credit ratings, not data workstations, so they are not clean comparables; the honest reading is that FactSet competes for the same data-and-analytics budgets as balance sheets several times its size, including Bloomberg's terminal franchise and LSEG's Refinitiv base. That gap is the strategic constraint behind the through-line: the moat has proven strong enough to defend share — retention above 95% — but not to out-grow a mid-single-digit market, which is why organic ASV growth sat in the 5–7% range even as retention held. Execution and breadth, not size, are what FactSet sells.

Generative AI: the threat the de-rating priced

The bear case is not speculative; it is written into the filings. FactSet first added a dedicated artificial-intelligence risk factor in its FY2023 10-K, warning that introducing generative AI could bring new liabilities [10]. By FY2025 the language had sharpened to name agentic AI and to concede directly that "third parties may be able to use AI to create technology that could reduce demand for our products and services," alongside a competition risk factor noting that many rivals "have significant AI capabilities and funding" [11].

Peer MSCI states the mechanism more bluntly than FactSet does: AI-enabled tools "may allow clients… to develop in-house capabilities to replace our products," and "large-scale data scraping and generative AI models trained on publicly available information could also diminish the perceived uniqueness and commercial value of our proprietary content" [12]. That is the disintermediation fear the market applied to FactSet when the multiple compressed. And FactSet's own early record fed the doubt: when it launched its GenAI assistant, FactSet Mercury, management said monetization "remains under discussion" and was "not included in this year's guidance" [13].

Generative AI: the tailwind now in the numbers

Two years on, monetization has moved from theory toward evidence, though the dollar base is still small. FactSet guided GenAI products to add only 30–50 basis points of ASV in FY2025, and its new CEO cautioned that firms "often underestimated the complexity involved in realizing" AI's value [14]. By Q3 FY2026 the tone had turned concrete: management said over 10% of the quarter's ASV growth came directly from AI SKUs, more than 20% of its top-100 clients were using its Model Context Protocol (MCP) data connectors on a paid basis, and one top-ten client "literally doubled their data subscriptions with us because of AI," and, on the moat, said the connected data and embedded workflows are "starting to see evidence rather than theory" [15]. Management added that MCP-linked deals improved contract value roughly 90% of the time [16].

The magnitude deserves plain framing. "Over 10% of ASV growth" is 10% of an annual growth increment of roughly $165 million on a $2.5 billion base — on the order of $16–17 million of incremental ASV, and management-sourced rather than an audited line item. It is early, not decisive. What makes it more than noise is that it aligns with independently visible trends: AI-ready data was cited as a demand driver in the Americas and Asia Pacific in Q1 FY2026, when wealth grew organic ASV 10% and off-platform data feeds and APIs became a rising share of expansion [17]. Management's competitive answer to the "point-solution" AI startups is integration — clients wanting one workflow instead of many tools — and it is applying AI internally as a margin lever, having captured more than half of a targeted 100 basis points of productivity gains this year [18]. AI is also one of the three core strategic pillars FactSet funds, alongside data and workflow depth [19].

Weighing the two readings

The same facts support both a tailwind and a threat interpretation; what separates them is which effect dominates over time.

Sources: FactSet FY2025 10-K risk factors [20]; MSCI FY2025 10-K [21]; Q3 FY2026 [22] and Q2 FY2026 [23] earnings calls.

The evidence points to a narrow-but-real moat and to AI as, so far, a net tailwind at the margin. Retention above 95% held through a full cycle, pricing power is intact, and AI monetization has begun to show up in ASV rather than only in slideware. The strongest fact against that read is the one MSCI names and FactSet concedes in its own risk factor: if frontier models and scraped public data erode the willingness to pay for aggregated proprietary content, a mid-sized specialist competing against far larger balance sheets is more exposed than the scale leaders. What would change the read in either direction is checkable in the filings: user and seat counts turning down (willingness-to-pay erosion), price realization reversing, or AI-SKU ASV stalling would confirm the threat; a broken-out AI revenue line growing to a material share of ASV, with retention holding above 95%, would confirm the tailwind.