The Barrier and the Bill

The Barrier and the Bill

The FY2025 Form 20-F makes the case for the moat in one sentence. The expansive network, the data-collecting infrastructure and the scale of two decades of historical data are, in the company's own words, "a significant barrier to our competitors" [1]. A page later it adds the second half of the claim: clients are "deeply integrated with us from an operational and technology perspective, making it difficult for them to switch providers and serving as a strong barrier to entry" [2].

Both halves are testable against numbers the company publishes itself. FY2025 is the year several of them moved the other way.

Sources: revenue growth derived from reported financials, FY2023–FY2025 revenue as filed [3]; Customer Net Retention Rate and sport rights share from the FY2025 Form 20-F [4] and Note 6 [5]; licence fee payables from Note 22 [6].

A step down in the top line

Revenue grew 16.6% in FY2025, against an average of 25.5% across the three preceding fiscal years. That is the largest single break in the multi-year record, and it is the number the rest of this chapter is built to explain.

Source: derived from reported revenue, FY2020–FY2025 as filed [7]; [8].

Two things sit inside that 16.6%, and they push in opposite directions.

The first is acquisition. IMG ARENA closed on November 1, 2025 and contributed €28.0 million of revenue and €1.3 million of net profit in the two months Sportradar owned it [9]. Strip that out and the underlying rate on the prior-year base is 14.0%. Content growth in the fourth quarter was explicitly attributed in part to the acquired rights: Betting and Gaming Content rose 29% in Q4 against 16% for the full year, with IMG ARENA named among the drivers [10]. So the reported figure flatters the organic rate by roughly two and a half points, in the same year that the acquisition brought its own obligations onto the balance sheet.

The second is currency, and it runs the other way. Sportradar reports in euros and sells a material share of its product in dollars; management describes the dollar-euro move as a headwind through FY2025 and quantifies it for the following quarter, where 11% reported revenue growth was 16% on a constant-currency basis [11]. The corpus does not give the equivalent full-year FX bridge for FY2025, so the constant-currency organic rate cannot be pinned. What can be said is that the FY2025 deceleration is partly translation and partly real, and that the acquired-content contribution offsets some of the translation drag in the reported number.

The comparison that gives the deceleration its edge is not against Sportradar's own history. It is against the one supplier that collides with it across the whole stack.

Sources: Sportradar revenue as filed [12]; Genius Sports revenue, growth and the cited market growth rate from its Q4 FY2025 earnings call [13]; [14].

Genius Sports reported group revenue of $669 million for FY2025 at 31% growth — "our strongest annual increase since 2021" — with a full-year 20% EBITDA margin its management called "our highest annual margin" [15]. It benchmarked that against "the 24% growth of global online sports betting GGR" in 2025 [16]. On the numbers each company published, one supplier in this layer grew above the end-market rate it cites and the other grew below it.

That comparison needs three caveats stated at full strength, because it is doing a lot of work. The two companies report in different currencies and neither discloses a like-for-like constant-currency figure the other can be measured against. Genius Sports' figures here come from earnings-call transcripts rather than an audited filing — the two annual reports filed under the GENI ticker in this corpus belong to a UK cardiovascular-genetics company that shares the ticker, so no Genius balance sheet or audited income statement exists in the record. And the mixes differ: Genius Sports' media business grew 37% to $144 million on a second half that nearly doubled against a soft comparison period, with management warning that rate would not continue [17]. The gap is directional, not decomposable.

The one expansion metric on the page

The company discloses exactly one measure of how much more its existing customers buy from it each year. Customer Net Retention Rate takes the trailing-twelve-month revenue of the top 200 clients as of twelve months earlier, recalculates the same cohort's revenue at the current period end — including upsells, net of contraction and attrition, excluding any revenue from new customers — and divides one by the other [18]. A reading of 109% means the base grew by nine points on its own, before a single new logo.

Sources: FY2021 Form 20-F [19]; FY2024 Form 20-F [20]; FY2025 Form 20-F [21]; Q1 FY2026 earnings call [22].

The 109% recorded for 2025 is the lowest of the six years the company has disclosed, and the 108% reported for the first quarter of 2026 is lower still [23] [24]. The previous trough, 111% in 2023, was followed by a recovery to 127% [25], so a single weak reading in this series has not previously been the start of a trend.

Four qualifications belong with the number, and each of them cuts in a different direction.

The metric has been renamed twice. It was the "Dollar-Based Net Retention Rate" through FY2021 [26], the "Net Retention Rate" from FY2022 to FY2023 [27] and the "Customer Net Retention Rate" from FY2024 onward. The company states at each rename that the calculation is unchanged, and the overlapping years in successive filings carry identical values — 125% for 2021, 119% for 2022, 111% for 2023 — which is the strongest available confirmation that the series is continuous [28] [29].

The cohort it is measured on is shrinking as a share of the business. The top 200 clients were approximately 76.4% of revenue at the end of 2022, 77.6% at the end of 2023, 83% at the end of 2024 and approximately 79% in 2025 [30] [31] [32]. Growth outside the measured cohort is real revenue that the metric is built to exclude, so a falling cohort share and a falling retention rate are not two independent pieces of bad news — they can be two views of the same shift toward newer and smaller accounts.

Both recent readings are stated on a basis that removes the year's largest content addition. The 109% "excludes any contribution from IMG," and management says the 108% likewise excludes existing customers' use of IMG content while including the currency headwind [33] [34]. The company does not publish an including-IMG figure, so the size of that adjustment cannot be checked from outside — and the same quarter's Betting and Gaming Content line grew 20% on "strong demand for IMG content across our client base" [35] [36]. On the disclosed basis, in other words, the cross-sell that most visibly worked in the period is the one the retention metric leaves out.

And the decline cannot be attributed. Sportradar has not disclosed a customer count since FY2022, when it reported 1,790 customers [37], and no filing or transcript in the corpus names an account won from or lost to a competitor in any period. Whether 109% reflects pricing, contraction inside accounts, or outright attrition is not determinable from the record.

The nearest peer number is not a like-for-like substitute. Genius Sports states that net revenue retention "remains in the 120% to 130% range across our Sportsbook customers year after year" across "circa 500 licensed Sportsbook brands" [38]. That is a single-product cohort of a different size, measured on a different definition, disclosed on a call rather than in a filing. It establishes that a rival claims a higher expansion rate; it does not measure the difference.

The bill compounding faster than the book

The barrier and the largest claim on future cash are the same line item, and both are visible in the same filing.

Sport rights expense — non-capitalized rights plus the amortization of capitalized licences — was €214.2 million in FY2023, €352.4 million in FY2024 and €404.3 million in FY2025 [39]. Against revenue that is 24.4%, 31.9% and 31.3%. The re-basing happened in FY2024 and has not reversed; management attributes the further FY2025 increase to the ATP partnership, the renewed MLB agreement and the addition of IMG ARENA rights [40].

The forward version of the same picture is the comparison between what Sportradar owes rightsholders and what customers have contracted to pay it. Both are disclosed, in different notes, on the same balance-sheet date.

Sources: licence fee payables from FY2025 Note 22 [41] and FY2023 Note on trade payables [42]; unsatisfied performance obligations from FY2025 Note 24 [43], FY2024 [44] and FY2023 [45].

Licence fee payables rose 42% during FY2025, from €1,074.0 million to €1,529.2 million [46]. Contracted unsatisfied performance obligations — the revenue customers have committed to and Sportradar has not yet delivered — rose 14%, from €1,854.6 million to €2,122.6 million [47] [48]. Three times the rate, on the face of it.

That single-year ratio does not survive contact with the acquisition, and the honest version is more interesting than the headline. IMG ARENA arrived with €122.4 million of current and €350.1 million of non-current trade payables — €472.5 million in total, assumed rather than incurred [49]. Sportradar's non-current trade payables consist entirely of licence fee payables for capitalized sport rights [50], so at least €350.1 million and at most €472.5 million of the €455.2 million increase came in through the acquisition rather than from new rights bought with the company's own signature. Removing that band leaves an ex-acquisition payables balance between €1,056.7 million and €1,179.1 million — somewhere between 2% lower and 10% higher than the year before. The company discloses no roll-forward of the payable, so the split cannot be closed further than that band. The backlog is contaminated the same way: IMG's customer contracts sit inside the €2,122.6 million and are not broken out.

The multi-year view is the one that survives the acquisition, and it says the same thing more slowly. Between the end of 2022 and the end of 2025, licence fee payables went from €413.2 million to €1,529.2 million, a factor of 3.7 [51] [52]. Contracted customer backlog went from €943.9 million to €2,122.6 million, a factor of 2.2 [53] [54]. Expressed as a coverage ratio, what the company owes rightsholders was 44% of what customers had contracted to pay it at the end of 2022 and 72% at the end of 2025. The obligations compounded at roughly 55% a year over that span against roughly 31% for the contracted book — not three times, but consistently and over three years, and through a period in which one full rights cycle was renewed.

The renewal clock

The exclusivity is concentrated, and it runs off on a schedule the company publishes. Six individual sport rights licences each exceed roughly 5% of the licence balance; together they carry a net book value of €1,020 million and constitute 70% of the total [55].

Source: FY2025 Form 20-F, Note 13 Intangible assets and goodwill [56].

The weighted-average remaining life of those six is 5.6 years. The rest of the licence portfolio averages 3.7 years [57]. The whole exclusivity position, in other words, comes up for renegotiation inside the decade, and the smaller thirty percent of it comes up first. The price of the next round is set by the counterparties, not by the buyer — and the last full round is the one that took rights expense from 24.4% of revenue to 31.9%.

Some of those counterparties are also shareholders. The eight-year NBA agreement grants warrants exercisable, once vested, for Class A ordinary shares equal to 3.00% of the total outstanding on a fully diluted, as-converted basis at an exercise price of $0.01 [58]. The eight-year MLB agreement signed in February 2025 issues MLB equity with a cash value of $35.5 million, up to 1,855,724 Class A shares vesting to July 2032 [59]. Part of the rights bill is settled in ownership, which means the leagues that set the largest cost line hold a claim on the equity that line is levered against.

What can be repriced, and by whom

Against the compounding claim sits genuine contractual visibility. Of the €2,122.6 million of contracted backlog, €1,357.0 million is scheduled for recognition in 2026 [60]. FY2026 revenue guidance is €1,557–1,582 million [61]. Roughly 86% of the guided year was already under contract at the point the guidance was set. That is the switching-cost claim showing up in a number rather than in an adjective.

The repricing surface is correspondingly narrow in any one year. The CFO puts the shape of it plainly: "about 2/3 of our revenue are fixed fee, 1/3 are variable[;] out of the 2/3 that are fixed[,] traditionally, about 1/3 comes up every single year" [62]. That is roughly 22% of the book renegotiated annually, spread through the year rather than at a single renewal date. The counterparty side of that negotiation is fragmented, on the client concentration figures established in Rights and Data Layer.

What the record does not contain is a price-versus-volume bridge. Sportradar has published no decomposition of revenue growth into price and volume in any period covered here. The clearest evidence that data prices are moving comes from outside the company, from a firm that is simultaneously a Sportradar customer and a competitor: Kambi guided its 2026 cost of sales higher on "an increase in recharged data supplier and other supplier costs, which are charged through to customers," alongside "the impact on commission rates of certain key partner renewals" [63]. Data-supplier costs in this layer are rising and are being passed down the chain rather than absorbed. That is third-party corroboration of direction, from a party with no incentive to flatter its supplier, and it stops short of quantifying what any one supplier captured.

Where the contest is actually being fought

The competitive language in the peer shelf is not about headline price. It is about the cost and the speed of turning a live event into a priced market — and both principal peers position against the human-operator model rather than under it.

Genius Sports: "Legacy manual data capture, where humans key in events from television feeds is obsolete. Leagues are transitioning towards automated AI-driven solutions, and we are winning that transition." It expects that automation to span its entire data rights portfolio "by the end of next year" and describes "a meaningful opportunity to take market share and drive incremental revenue with limited additional costs" [64].

Kambi has taken the same argument into the odds layer, which is Sportradar's. It reports that 49% of all bets across its network were fully AI-traded in 2025 and that it passed 50% in January 2026 [65]. Its Odds Feed+ pricing service — sold, in its CEO's words, against "established incumbents in the odds feed space" — added FDJ UNITED, Hard Rock Digital, LeoVegas, Superbet and Rei do Pitaco during 2025, reaching seven odds-feed partners at year end [66] [67]. After the 2026 World Cup its CEO put the operating-leverage claim directly: "others still reliant on manual trading will need to scale back down now… We will not have this need to scale anything down" [68].

The scale of that particular threat should be kept in proportion. Kambi's Turnkey Sportsbook still generated 87% of its total revenue in 2025 [69]; seven odds-feed partners is a beachhead, not a share shift. And Kambi's own European commentary describes a mature market where the emphasis falls "on winning business from competitor suppliers" [70] — which is the condition under which a beachhead matters more than its current size.

Sportradar identified this contest before either rival made it a slogan. The FY2021 Form 20-F states the risk in its own words: if a competitor "replaces the need for data journalists before we do, our business could be materially harmed" [71]. The FY2025 filing shows the company building toward that objective: automated collection and production of live events using computer vision, and a proprietary transformer-based foundation model of basketball trained on tracking data and 3D body-pose sequences [72].

The human cost line has nonetheless grown faster than revenue. Data journalist and freelancer fees were €24.1 million in FY2023, €22.4 million in FY2024 and €31.7 million in FY2025 — up 42% in a year when revenue rose 16.6% [73].

Source: FY2025 Form 20-F, Note 6 Purchased services [74].

There are two readings of that line and the filing supports the benign one. Management attributes the rise in purchased services partly to "additional scout costs driven by the expansion of product offerings and data collection" [75] — more events covered requires more people at more venues, whatever the automation rate on the events already covered. The line is not a per-event unit cost and cannot be read as one. But it is the only human-capture cost the company discloses, and in the year two rivals declared the manual model finished, it grew two and a half times as fast as revenue. The metric that would settle it — cost per collected event, or the share of events captured without a human present — is not disclosed by any party in this record.

What the exclusivity has established, and what it has not

The company's self-description has moved once, and in one direction. The IPO-year filing headed the section "The only end-to-end data and software solutions provider with a global footprint" and stated: "We are the only company providing software solutions that address the entire sports betting value chain" [76]. From FY2022 onward the identical passage reads "a leading provider" [77], and it still reads that way in FY2025 [78].

Set against the evidence assembled here, the two halves of the barrier claim stand on very different ground.

The input side is established in numbers. Assembling an exclusive rights book at this scale costs €404.3 million a year in expense and €1.5 billion in accumulated obligations, and the clearest external mark on what that costs a sub-scale operator is that a rival's entire rights portfolio changed hands with money moving toward the buyer — the transaction and the market structure it produced belong to Rights and Data Layer, and the obligations that travelled with it to Rights Bought on Credit. Contracted backlog covering 86% of the guided year [79], and a client base with no account above 10% [80], are integration and diversification showing up as arithmetic rather than adjectives.

The output side is not established. The measure of pricing and expansion power the company chose to publish is at its lowest reading in six disclosed years and lower again in the following quarter; the cohort it is measured on has shrunk; the record contains no price-versus-volume bridge, no customer count since 2022, no named account won or lost in five annual reports and every transcript, and no market-share estimate for any participant anywhere in the corpus. Delivered FY2025 growth ran below both the closest rival's and the end-market rate that rival cites. On this evidence the moat is narrow rather than wide, and it is asymmetric: demonstrated on the cost of assembling the input, not demonstrated on the ability to charge for the output.

Two things would change that read, and both are checkable on a date. A recovery in Customer Net Retention Rate through 2026 on an including-IMG basis — or the disclosure of that basis at all — would show the expansion engine restarting rather than the metric excluding the thing that worked. And the next rights round, whose first renewals fall due against a portfolio averaging 3.7 years outside the big six, will price what two decades of network and archive are actually worth to the counterparties who set the bill.

The rights book is the asset and the liability schedule, and FY2025 is the first year the record lets a reader watch both move at once. What it does not yet let anyone judge is whether the people running the company have been telling investors what to expect from it.