---
title: "Nordic SaaS Valuation & M&A: How AI Reprices Software Companies"
description: "How software companies in the Nordics are valued and sold as AI moves inference into the cost of goods sold — where multiples stand in mid-2026, the metrics that get repriced, why vertical SaaS defends its premium, AI due diligence, and the shift from seats to outcomes. Grounded in Translink's SaaS valuation work and Neuvottelija conversations."
keyword: "Nordic SaaS valuation and M&A"
datePublished: 2026-07-18
dateModified: 2026-07-18
authors: ["Sami Miettinen"]
canonical: https://www.neuvottelija.com/guides/nordic-saas-valuation-and-ma/
---

# Nordic SaaS Valuation & M&A: How AI Reprices Software Companies

If you are valuing, buying or selling a Nordic software company, the ground has moved. AI doesn't just add a feature to SaaS — it changes the economics behind the numbers you already run. This guide lays out where multiples actually stand in mid-2026, how software is priced now, why the vertical/horizontal gap persists, what AI due diligence looks like, and where the business model is heading. It draws on Translink's quarterly SaaS valuation work and on conversations from the [Neuvottelija podcast](/podcast/).

## The thesis: SaaS does not disappear, seat pricing does

The fear that AI ends software-as-a-service is misplaced. What AI ends is the per-seat model. I gave the full argument as a keynote at Arctic15, delivered with DLA Piper and Fenerum: [AI Reprices SaaS](/events/ai-reprices-saas-arctic15/). The short version: as agentic work converts labour budgets into demand for inference tokens, inference stops being an IT line item and becomes a labour-substitution cost — and that cost lands inside SaaS unit economics.

## Where multiples stand in mid-2026

The market is not uniformly cheap — it is polarised, and the spread is the story.

| Segment (public markets, July 2026) | Median EV/Revenue (NTM) | Median EV/EBITDA (NTM) |
| --- | --- | --- |
| Horizontal SaaS | ~2.2x | ~9.4x |
| Vertical SaaS | ~2.4x | ~10.8x |
| Infrastructure software | ~2.8x | ~11.6x |

Public software medians have compressed hard from the 2021 peak and even from mid-2025, when the median briefly ran above 6x revenue on AI optimism; by spring 2026 it had fallen to the 3x range and continued drifting down (Aventis Advisors; Multiples.vc, July 2026). Private M&A medians tell the same story: roughly 3.1x revenue in early 2026 against a ten-year median of about 4.5x, with the first quartile nearer 2.4x.

Against that, the top of the market has detached: [at Translink's Nordic Tech Stars event](/podcast/episodes/633-translink-nordic-tech-stars-negotiator-356/), international investors described roughly 20x EBITDA or around 10x ARR for the very best assets, a flight to quality accelerating since 2023, and plenty of private equity dry powder waiting for strong-KPI SaaS. The valuation question in 2026 is not "what is the market multiple" but "which market are you in".

## The metrics that get repriced

When inference cost enters cost of goods sold, it reprices the traditional SaaS scorecard — because a cost that used to be negligible now scales with usage.

| Classic metric | What AI does to it | The AI-era companion metric |
| --- | --- | --- |
| Gross margin | Inference lands in COGS; 80%+ software margins are no longer automatic | Inference share of COGS (%) |
| Rule of 40 | Growth bought with heavy inference spend counts less | Rule of 40 with AI costs fully loaded |
| CAC payback | AI-assisted sales lowers CAC — for you and every competitor | Cost per acquired workflow |
| ARR per employee | Agentic work inflates it; buyers ask what is durable | Tokens per active user |
| Net revenue retention | Usage-priced AI features make NRR path-dependent | Cost per outcome delivered |

None of the classics die — buyers still open the spreadsheet at gross margin and the Rule of 40. Only about a fifth of actively traded SaaS companies cleared the 40% threshold in late 2025, with the median score around 28%, and each ten-point improvement has been worth roughly one extra turn of EV/Revenue in public markets (Aventis Advisors). What changes is what sits underneath: a 75% gross margin built on well-managed inference is a different asset from a 75% margin that a model-price change can erase. The [keynote writeup](/events/ai-reprices-saas-arctic15/) sets the metric pairs out side by side.

## Why vertical SaaS still defends its premium

Multiples compressed across software into 2026, but the spread between vertical and horizontal did not close. At Q1 2026, vertical software traded roughly 79% above horizontal on median EV/LTM revenue in Translink's Nordic data — about 3.4x versus 1.9x — and the same ordering shows up in public-market medians. The gap holds for structural reasons: vertical software owns the system of record, with domain depth, higher switching costs and stickier data. AI commoditises generic, horizontal workflows first — not systems of record. That distinction is where much of the value now sits, and it is visible segment by segment: design and engineering software and data infrastructure command multiples several times those of commoditising categories like adtech or undifferentiated cloud tooling.

## AI due diligence: the new workstream in every software deal

Two years ago software DD meant code quality, security and churn. Today every serious buyer of a Nordic software company adds an AI workstream, and sellers should run it on themselves first:

- **Inference economics.** What share of COGS is inference, who pays for it (you or the customer), and what happens to gross margin if model prices move against you — or for you.
- **Model dependency.** Single-provider risk, switching cost between models, and whether the product's quality survives a model swap. Local and open-weight fallbacks change this calculus, a theme I dig into in [Local Models and AI Sovereignty](/ai/insights/local-models-ai-sovereignty/).
- **Data rights.** Do your contracts give you the right to use customer data to improve the product? The proprietary-data moat only exists if the paperwork says so.
- **Code provenance.** AI-assisted codebases ship faster — [top engineers now orchestrate agent fleets](/podcast/episodes/666-tekoalykoodaus-huipputasolla-markus-hav-371-neuvottelija/), and the craft has moved [from vibe coding to agentic engineering](/ai/insights/vibe-coding-to-agentic-engineering/) — but buyers ask how the code was produced, reviewed and licensed.
- **Substitution exposure.** The honest question: could an agent with a model subscription rebuild your product's core workflow? For horizontal tools the answer is increasingly yes, which is exactly why the vertical premium persists.

Run this list eighteen months before a process and every item is a fixable operating decision. Run it for the first time in a buyer's data room and every item is a price adjustment.

## From seats to outcomes

The strategic shift reduces to three moves: seats decline (agents need data access, not user seats); usage and outcomes rise (customers pay for completed workflows, metered via API); and proprietary data wins (the data layer captures margin as the application tier commoditises). Pricing follows the same arc — from per-seat licences, through usage metering, to outcome-based pricing where the vendor is paid per completed workflow and owns enough of the data to underwrite that promise. The Nordic proof point is ICEYE — a large valuation built on a synthetic-aperture-radar sensor-data hypervertical with outcome-driven economics. The full argument, with sources, is in the [keynote](/events/ai-reprices-saas-arctic15/).

## What a Nordic founder should do 12–24 months before a process

Translate all of the above into preparation: instrument your inference costs so you can show them per customer and per workflow, not as one cloud line; restate gross margin with AI costs fully loaded before a buyer does it for you; renegotiate data rights into your customer contracts at renewal; reduce single-model dependency where quality allows; and reposition reporting around the metric pair the market now pays for — Rule of 40 with honest COGS, and revenue quality by segment. If you are deciding between growth and profitability on the way to a sale, remember the public-market arithmetic above: ten Rule-of-40 points has been worth about a turn of revenue multiple. Then read the [exit-process guide](/guides/how-to-sell-a-company-in-finland/) — everything about buyers, structure and shareholder terms applies doubly when part of your price is an AI story.

## Growth, capital and deep tech around the edges

Valuation is downstream of growth, and growth in the Nordics runs into a specific tempo problem. Mårten Mickos — who ran MySQL and HackerOne — argues that Silicon Valley operates at an exponential clock speed most of Europe doesn't share; I pull that together in [The Silicon Valley Growth Formula](/ai/insights/mickos-silicon-valley-growth-formula/). At the earliest stage, deep tech breaks the standard SaaS playbook entirely — different survival odds, a longer fund horizon, and a Nordic pre-seed gap that kills good science before it becomes a company, which Anssi Uimonen explains in [Investing in Deep Tech](/investment-banking/insights/deep-tech-investing-finland/). And if you want the sceptic's case on whether any of this AI value is real, [Is AI a Bubble?](/ai/insights/ai-bubble-inference-economics/) splits that into the questions of inference pricing, capability and adoption.

## Where to go next

- [AI Reprices SaaS](/events/ai-reprices-saas-arctic15/) — the full keynote, with slide deck
- [Investing in Deep Tech](/investment-banking/insights/deep-tech-investing-finland/) — the Nordic capital gap at pre-seed
- [The Silicon Valley Growth Formula](/ai/insights/mickos-silicon-valley-growth-formula/) — linear country, exponential world
- [Is AI a Bubble?](/ai/insights/ai-bubble-inference-economics/) — inference economics from first principles
- [How to Sell a Company in Finland](/guides/how-to-sell-a-company-in-finland/) — the exit process end to end
- All [SaaS & software episodes](/podcast/topics/saas_software/)

For advisory enquiries, see [Investment Banking](/investment-banking/) or [Contact](/contact/).

## Frequently asked questions

### How is SaaS valued now that AI is involved?

The classic metrics — gross margin, EBITDA, Rule of 40, CAC payback, ARR per employee — still matter, but they are now driven by a cost that used to be near zero: inference. As model calls land inside cost of goods sold, they reprice margins, growth and valuation at once, which is why new metrics like inference COGS %, tokens per active user and cost per workflow have become essential.

### What multiple can a SaaS company expect in 2026?

The market is polarised. In mid-2026 public SaaS medians sit around 2–3x forward revenue and private M&A medians around 3x revenue — but the best assets trade far above: investors at Translink's Nordic Tech Stars event described roughly 20x EBITDA or around 10x ARR for top-quality companies while ordinary ones lag. Quality of growth, profitability and AI exposure decide which side of that divide you are on.

### Why does vertical SaaS trade at a premium to horizontal SaaS?

At Q1 2026, vertical software still traded roughly 79% above horizontal on median EV/LTM revenue in Translink's Nordic data — about 3.4x versus 1.9x. The gap holds because vertical software owns the system of record: domain depth, higher switching costs and stickier data. AI commoditises generic, horizontal workflows first, not systems of record.

### Does the Rule of 40 still matter?

Yes — arguably more than ever, but with AI costs inside it. Only about a fifth of public SaaS companies cleared the 40% threshold in late 2025 (median score around 28%), and analyses of public multiples find each ten-point improvement in the Rule of 40 worth roughly one extra turn of EV/Revenue. The catch: growth bought with heavy inference spend and thin gross margin no longer counts the way it used to.

### What is AI due diligence in a SaaS sale?

Buyers now examine how exposed the product is to AI substitution and how well the company controls its AI economics: inference cost share of COGS and who pays it, dependency on a single model provider, rights to customer data for model improvement, code provenance in an AI-assisted codebase, and whether the product is agent-readable. Weakness here prices as risk; strength prices as moat.

### What does 'from seats to outcomes' mean for software pricing?

Agents need data access, not user seats, so per-seat licensing stops scaling with the value delivered. Software increasingly meters completed workflows via API, and the proprietary data layer captures margin as the application tier above it commoditises. In shorthand: software moves from selling access to selling work.

### Does AI mean SaaS is dying?

No. SaaS does not disappear — seat pricing does. The business model shifts from selling access to selling work, and the winners own the data and workflows that agents need to do that work.

---

Cite as: Sami Miettinen, Neuvottelija — Nordic SaaS Valuation & M&A: How AI Reprices Software Companies, https://www.neuvottelija.com/guides/nordic-saas-valuation-and-ma/, 2026-07-18.