A net-cash SaaS network at 12x EV/EBITDA vs peers at 16–33x — a mispriced compounder, or a toll road the AI agents are about to drive around?

**Quick background**: SPS Commerce (SPSC) runs the largest retail-focused trading network in North America — the cloud plumbing that lets \~3,000 retailers and tens of thousands of suppliers swap purchase orders, invoices and shipping docs through one connection. FY25 revenue was $752M (+18%, but M&A-flattered), \~69% gross margins, $152M free cash flow, and a net-cash balance sheet (\~$144M, effectively no debt). The stock round-tripped off its 52-week low near $49 to \~$61 now (24 Jul 2026), still down \~57% from its 52-week high near $142. At $61 it trades \~12x EV/EBITDA and \~2.86x revenue vs Descartes 16.5x, Manhattan \~30x, Kinaxis \~33x.

**The interesting part of the bull case**: the argument isn't "it's cheap." It's that what customers actually pay for is the labour-intensive onboarding/mapping/compliance SERVICE and the retailer mandate that funnels suppliers in — not the raw EDI bytes. If that's the moat, the fashionable "APIs will kill EDI" worry is aimed at the wrong target, \~7% growth is a trough, and a live sale process (Morgan Stanley engaged, activists Anson + Irenic pushing) gives you a catalyst.

**Where I kept poking holes**:

* The re-rating math is partly circular. SPS's own history (32–63x EV/EBITDA, 2018–24) and today's peers both belong to the same zero-rate / AI-mania regime. Normalize for a \~7% grower and the defensible multiple is mid-teens — so from 12x the multiple-only upside is maybe +15–35%. The rest of the path to $100+ needs margin expansion or an acquirer's premium, not the multiple re-inflating.
* The AI risk isn't "EDI vs API" — it's the SERVICE margin. AI auto-mapping compresses the exact manual work that justifies >$13k ARPU, and it's invisible in a 69% gross margin until renewal pricing rolls. On 21 July, Morgan Stanley's research desk downgraded to Underweight (PT $70→$57) citing agent-to-agent protocols doing exactly this — while MS's banking arm runs the sale. Same firm, opposite signals.
* Growth is already soft: Q1 2026 revenue missed ($192.1M vs \~$197.6M). Q2 lands 30 July.
* Capital allocation wobble: Carbon6 (bought Feb 2025) blew up within weeks on an Amazon policy change and a customer-count miss (\~8,500 vs \~6,500 underwritten), unwound in 2026 at a \~$20M loss. SBC \~7% of revenue, buybacks mostly just offsetting dilution.

**One genuinely interesting angle**: the whole thing leans on the network-hub analogy (Descartes, WiseTech, card networks) — businesses that got more valuable as they got denser. The question is whether retail EDI is still densifying, or whether it's a mature utility where AI makes the "complexity you pay us to absorb" progressively cheaper to absorb yourself.

**Closing question:** for those who've owned or diligenced SPSC (or Descartes/TrueCommerce) — is the onboarding/compliance service genuinely sticky enough to survive AI auto-mapping, or does ARPU quietly compress at renewal even while EDI volumes keep growing? That's the variable I'm least sure about.

More details at my full write-up: [https://fmarinisecondopinion.substack.com/p/spsc-sps-commerce-inc](https://fmarinisecondopinion.substack.com/p/spsc-sps-commerce-inc)