$ARX Thesis | Disguised Growth AI Play Within Insurance (116% Upside)

Accelerant Holdings ($ARX) – Long @$13 per share

NFA – DYOR

Theme: "Invest where the puck is going" – AI buildout -> undiscovered AI beneficiaries

TLDR

+ >30% rev/ EBITDA growth
+ AI beneficiary w/ strong data-driven MOAT in small specialty insurance
+ very likely to beat-and-raise several times over upcoming quarters
+ <8x fwd EBITDA (v 25x IPO)
+ 25% mgmt ownership & insider buys
+ $200m buyback
+ low float (this cuts both ways, but I think explosive setup with upcoming catalysts)

Introduction

Accelerant operates the only two-sided specialty insurance exchange at scale, connecting 280 specialty underwriters with 95 institutional risk capital partners across 22 countries and 600+ products. Exchange Written Premium reached $4.2bn in CY25A, up 35% year-over-year and entirely organic. Revenue was $913m, adjusted EBITDA $237m (+158%), FCF conversion (ex-underwriting) 87%. The stock IPO'd at $21 in July 2025, peaked at $31.18, and trades at ~$13 today, a $2.9bn market cap and a 58% decline from peak. I think the market is applying a carrier multiple to a platform business that is compounding through the cycle and inflecting toward capital-light economics. It fully neglects Accelerant's unique data asset: years of specialty insurance data for small businesses that further compounds in an AI native/ agentic world and already allows Accelerant to vastly outperform the market on underwriting (gross loss ratio >15 ppt below market in low fifties). My probability-weighted 1-year target is $27.09 (+107%), supported by a conservative DCF range of $31-38.

Dislocation

The Q3 2025 selloff was triggered by concerns over an affiliated party relationship with Hadron, a fronting carrier backed by Altamont Capital, Accelerant's controlling shareholder. The relationship was disclosed in the S-1, but the market treated it as a governance red flag as it learned that Hadron was the largest third-party insurer on the exchange in Q3 2025. The concern is mostly narrative. Hadron's 5.5% fronting fee is the most expensive of any third-party partner, making it the least favorable relationship for Accelerant in the value chain. Management is actively mixing it down: 47% of third-party premium in Q4 2025, guided below one-third by Q4 2026. The selloff was amplified by a first-time public company management team that failed to communicate the equity story clearly enough for the market to absorb, but has beaten or met guidance every quarter since the IPO.

Underneath the governance noise, consolidated accounting obscures where economic value actually originates. Accelerant reports three segments: Exchange Services (67% EBITDA margin), MGA Operations (45%+), and Underwriting (13%). The Underwriting segment writes roughly 60% of EWP on Accelerant's own paper before ceding approximately 90% to reinsurers. The intercompany eliminations that result compress the visible contribution of the fee platforms on the consolidated income statement, blending what looks like an insurer with what is in reality high-margin, capital-light infrastructure. Stripping out standalone Underwriting EBITDA reveals the hidden growth engine: the fee-based segments accounting for 76% of consolidated EBITDA in CY25A, growing at a roughly 32% CAGR in CY29E forecast. A structural separation would surface that clarity instantly. In the meantime, management has started revamping the narrative. The hire of Linda Huber as CFO is a robust signal: 13 years at Moody's during its run from $30 to $170, then MSCI and FactSet. As third-party premium scales (already 40% in Q4 2025, reaching 50%+ in my forecast by Q2 2027, the convergence of standalone and consolidated accounting will naturally become increasingly clear. Q4 2025 was the first quarter in which Exchange Services contributed positive EBITDA after consolidation eliminations.

Layered on top is sector drag. The quasi-peer basket, mostly brokers, has sold off on AI-disintermediation fears that are legitimate for distribution layers but do not apply to infrastructure. Soft-cycle concerns miss the point that Accelerant's growth is volume-driven: of CY25A's 35% EWP growth, only 3 percentage points came from rate. Further, I believe a softening market counterintuitively benefits Accelerant, because when carriers pull MGA capacity as a first line of defense, the five-year guaranteed capacity becomes disproportionately valuable.

In short, the market is valuing Accelerant as a specialty carrier, not as a platform. Every operating metric speaks against that and management keeps outperforming its own guidance.

Investment case

The thesis rests on three pillars. First, a data moat: Accelerant has built the only centralized specialty insurance loss dataset at scale. It cannot be purchased or replicated unless you operate an equivalent exchange for a comparable period. The dataset produces a 51% gross loss ratio, vastly below the roughly 66% industry average. Second, structural lock-in on both sides of the exchange. MGAs stay because the five-year guaranteed capacity bundle is unmatched: 83 NPS, 126% net revenue retention, and only one voluntary departure in platform history. Risk capital partners stay because the curated portfolio delivers returns they cannot assemble bilaterally. Institutional investors accessing through Flywheel Re, Accelerant's collateralized reinsurance sidecar, earn net IRRs exceeding 35% with floors around 20% even under loss ratio deterioration. That is why the partner base has grown from 2 to 95 in six years and third-party premium has surged from 15% to 40% of EWP in a single year. Third, AI acceleration. AI commoditizes distribution and processing, the layers Accelerant does not operate in. It compounds the value of what Accelerant does own: accumulated loss data, committed capital, and regulatory authority across 50 states. Infrastructure becomes more valuable in an agentic world, not less. Accelerant is uniquely positioned to win the "AI agent wars" as it becomes the infrastructure upon which specialty insurance can be autonomously written. Management understands this, but again, fails to communicate the AI beneficiary story digestibly to the market.

Growth and platform inflection

These pillars translate into a growth trajectory that is compounding and broadening. EWP has compounded at a 66% CAGR since inception in 2018. The platform added 63 net new members in CY25A, a record, with the pipeline standing at over $4bn of annualized premium, nearly matching the entire current book. Existing risk capital partner capacity alone supports a doubling of EWP from here; growth is not capital-constrained. Four new Mission members launched in Europe and the UK in early 2026, and the oldest member cohort (pre-2020) reaccelerated from 3.9% growth in CY24A to 11.0% in CY25A, confirming the platform effect compounds rather than plateaus. Captive insurance, an entirely new channel, contributed $40m in CY25A with management guiding over $100m in CY26E against a multi-billion-dollar addressable market hinting at the next S-curve Accelerant can capture with a strong right to win.

Two curves define the platform inflection that will force the re-rating. Third-party premium moved from 15% of EWP in CY24A to 30% in CY25A to 40% in Q4 2025 translating into fee income from third-party insurers deploying their own paper through the exchange. The Exchange Services take rate expanded from 6.8% in CY23A to 8.4% in Q4 2025, rising despite platform scaling. This is not a negotiated brokerage commission subject to competitive bidding. It is the take rate on infrastructure with no substitute at scale.

The setup structurally favors repeated beats. The founding team holds roughly $725m of economic interest with options struck at the IPO range of $20-21, none exercised. Management raised both EBITDA and EWP guidance at the Q4 print and has, I believe, deliberately set a (very) conservative bar for CY26, acutely aware of the market's sensitivity post-Hadron. The buyback pace since the March 18 authorization tells a supporting story: roughly $50m deployed in weeks against a $200m program, a signal that near-term prints are expected to beat. At 8.6x, the market is pricing a carrier fully neglecting that this business is quickly becoming the infrastructure upon which specialty insurance is written.

Valuation

There is no direct public comparable for Accelerant, but every imperfect one implies a materially higher multiple. The quasi-peer basket (Trisura, Kinsale, Neptune, AJG, BRO, Goosehead, Baldwin, Aon, WTW) trades at 16.0x CY26E EV/EBITDA on average. Kinsale, the most commonly cited specialty comp, trades at 13.9x with weaker fundamentals on every measure and taking full balance sheet risk. Neptune, the only public pure-play MGA, trades at 41.5x, illustrating where the market pays for a platform model. I exclude Ryan Specialty (8.2x, 3.2x net leverage) because roughly half its revenue is large-account E&S wholesale brokerage, the layer most exposed to rate-cycle softening and AI disintermediation. A broader basket of fintech rails, data platforms, financial exchanges, and marketplaces, which I believe should be applicable, trades at 16.9x. On forward Rule of 40, Accelerant scores 57.6% versus 48.6% for quasi-peers and 53.1% for the platform basket. In private markets, well-run MGA platforms transact at high-teens EBITDA multiples. Accelerant itself traded at 20-25x NTM EBITDA for its first month post-IPO. Every reference point lands well above current trading of 8.6x, while the company has a $478m net cash position. 

My base case (60% probability) applies 15x to CY27E EBITDA of $378m, yielding $27.35 (+109%) as a 1-year price target. Downside (20%) at 8x yields $13.57 (+4%). Upside (20%) at 20x yields $39.85 (+204%). Probability-weighted: $27.09 with a very asymmetric RR. DCF corroborates at $31-38.

With $725m of founder equity on the line, $200m buyback at 30% of float, $2m of management open-market purchases at $13 in November 2025, and aligned incentives in the cap table, the story has potential to quickly move from "show me" to pricing the infrastructure for specialty insurance narrative, especially if the company beats-and-raises repeatedly. 

Risks

Soft market GLR drift is the primary risk. The 15 percentage point gap to the industry average is the clearest measure of moat durability; mid-50s does not break the thesis, approaching 60% does (read-through from $ASIC's Q1 2026 ER invalidates this concern, also because $ASIC's average written premium is significantly larger than $ARX's). Take rate compression is the second risk: historically it has expanded from 6.8% to 8.4%, I reassess below 7%. Altamont's 79% voting power creates governance overhang, though Fund III's 2034 lifetime means no forced-liquidation timeline. Hadron concentration is being actively mixed down, and its premium fronting fee means the mix-down improves unit economics. Member attrition has no historical precedent: one voluntary departure. Prior-year development was $6.5m in CY25A, down from $15.1m, all from discontinued 2022-vintage books with an LPT backstop.

Catalysts

+ Q1 2026 beat-and-raise (May 13): EWP and EBITDA deliberately conservative for CY26E; EWP, DWP and EBITDA raises all more likely than not

+ Buyback execution and likely upsize: ~$50m deployed in weeks against $200m authorization

+ Hadron mix-down below one-third of third-party DWP by Q4 2026, removing the primary bear narrative

+ Narrative simplification: shift toward fee-based segment reporting, Rule of 40 framing, and revised communication; structural separation of Underwriting segment, exposing fee-based economics currently hidden in consolidation

+ Altamont Class A/B dissolution by July 2028 at latest, removing dual-class discount