Lovable’s $1.8B Series A wasn’t an AI premium. It was a retention bet.

The market still talks like Series A valuations are a growth multiple plus an AI uplift. But the deals that clear in Western Europe are priced like insurance: investors pay up for revenue that looks durable and scalable under European underwriting constraints.

On 17 July 2025, TechCrunch put a number on Lovable.

Lovable is a B2B AI company from Stockholm, Sweden. TechCrunch reported a $200m Series A led by Accel at a $1.8b valuation.

TechCrunch also reported Lovable’s own earlier claims: $17m ARR by February 2025, and 30,000 paying customers.

That last detail is the one that should make you sit up.

Because it is also a trap.

If you have 30,000 paying customers, you almost certainly have 30,000 reasons for someone to leave.

Imagine you are Accel, looking at that spreadsheet.

You are not just seeing growth. You are seeing a future problem hidden inside it.

How many of those customers are real users, building a habit?

And how many are just trying the new thing because it is the new thing?

That diligence question is the whole round.

Lovable’s scale story only works because investors believe the revenue will stick after the early-adopter rush. Photo: Andrew Neel / Pexels, Pexels licence (free commercial use).

So here is the simple question Lovable forces onto every European Series A founder.

What are investors really buying when they pay an ARR multiple?

Founders talk about “multiples” like a scoreboard.

Investors treat them like insurance pricing.

An ARR multiple is a bet on what your current recurring revenue becomes after renewals, expansions, and the ugly work of delivery.

Will customers stick around?

Will they buy more?

Will each new customer cost you more people?

And will your margins survive once the easy customers are gone?

The story we all want to believe

The old model is tidy.

Series A valuation is mostly a growth multiple. If you are in a hot AI category, you get a narrative premium.

Retention and efficiency are nice, but secondary.

It is comforting.

It says the future is mostly about persuasion.

Here is the twist: in Europe, “AI” isn’t the premium. Underwriteable revenue is.

In Western Europe’s 2024-2025 Series A market, the evidence in this brief points to the same driver again and again.

Not “AI versus non-AI”.

Durable, scalable revenue versus fragile, labour-heavy revenue.

Alvarez & Marsal, in its Q4 2025 European software commentary, describes a repricing that widens the gap between resilient software platforms and the rest.

That is the setting.

The plot is what investors do inside it.

They pay up when retention is credible and margins can scale.

They mark down when delivery looks like services, or when data rights are ambiguous.

Here is the Netflix moment.

This is why two AI companies with the same growth rate can get wildly different outcomes.

One looks like a machine.

The other looks like a magician.

“Tourist churn”: the one churn story that matters in AI

Andreessen Horowitz, the US venture firm, put language to a pattern on 10 September 2025: “tourist churn”.

They meant the early wave of users who try an AI product because it is novel, then leave.

Think of a simple cohort curve.

Month 0: 100 customers sign up.

By Month 3: 40 have churned. The tourists have gone.

By Month 12: the remaining 60 are still there, and some are paying more.

That second slope is the valuation story.

If the line keeps falling, your ARR is a leaky bucket.

If it flattens, you have a retained core. That core can compound.

That is why cohorts beat topline charts.

Not because investors love spreadsheets.

Because they hate paying for revenue that disappears.

Lovable: the conditional AI premium in the wild

Lovable’s headline is a unicorn valuation.

The quieter message is distribution.

TechCrunch reported 30,000 paying customers with low spend. That implies the product can be adopted without an army of people installing it.

In other words, it can behave like software.

Now return to the diligence question.

If you are Accel, you do not need perfection on day one.

You need evidence that the paying base is not just a flash crowd.

This is where the “AI premium” becomes conditional.

AI can justify a high price when investors believe the cohort curve will flatten into a stable, expanding core.

AI loses that premium when the same growth is powered by novelty that fades.

Lovable is the case study founders misread.

They copy the word “AI”.

They miss the underwriting logic.

JUPUS: the European advantage nobody brags about

Now shift to Cologne, Germany.

JUPUS is a legal-tech company. It sells into workflows where mistakes can be existential.

In a Series A announcement published by HTGF in July 2026, JUPUS raised €13m. The announcement also said that in 2025 the company quadrupled ARR.

But the detail that matters is not the money.

It is what JUPUS chose to build around.

GDPR, professional conduct rules, and attorney-client privilege.

That sounds like legal overhead until you see it through retention.

In legal workflows, trust is not a marketing claim.

It is the reason renewals are boring.

This is a different route to the same destination as Lovable.

Make the revenue durable.

Windsor Drake’s Q4 2025 vertical SaaS valuation report is consistent with that broader point: vertical workflow depth can support stronger multiples than people expect, because durability shows up in GRR and NRR.

The surprise is emotional, not technical.

The “slow” market can be the safer market.

JUPUS shows how compliance-grade design can become a retention moat investors will pay for in Europe. Photo: KATRIN BOLOVTSOVA / Pexels, Pexels licence (free commercial use).

A UK public company says the quiet part out loud: 109% NRR

Private investors like to pretend public markets are irrelevant.

Then they use them to anchor reality.

Alfa Financial Software is a UK-listed software company. In its full-year results for the year ended 31 December 2025, it disclosed ARR of £43.9m, up 15%.

It also disclosed NRR of 109%.

And it pointed to cloud-native architecture and profitability, explicitly stating it beat the Rule of 40.

This is not a Series A pitch deck.

It is a public company telling you what it believes investors reward.

Retention that expands.

A model that scales without collapsing margins.

Alvarez & Marsal’s Q4 2025 commentary links the same repricing logic back to European software more broadly.

The message is painfully simple.

If growth is high-quality, it compounds.

If it is low-quality, it leaks.

Palantir: the delivery model that decides your multiple

There is one modifier that keeps ruining “AI premium” stories.

Delivery.

Palantir Technologies is a US public company known for high-stakes enterprise deployments. In its 2025 Form 10-K, Palantir reported gross margin rising to 82% from 80%.

The filing also explains how deployments, professional services, and cloud or on-prem subscriptions interact.

That matters because many AI startups pitch “Palantir-like outcomes”.

They want the credibility, and the valuation.

But investors now obsess over whether forward-deployed engineering, the high-touch team that helps customers implement, is scalable software delivery or disguised services.

A dedicated diligence and valuation playbook has emerged around this, as described by M Search’s guide to FDE.

Here is the twist.

This is not an operational footnote.

Your deployment model is your unit economics.

If every deal requires months of high-touch work, your margins will tell the truth before your pitch does.

Europe’s “constraints” are not vibes. They are underwriting inputs.

You may have noticed what we did not do.

We did not claim Europe is slower because someone “feels” it.

The brief is tighter than that.

It says European underwriting constraints shape what looks durable and scalable: procurement reality, GDPR and data rules, services-heavy deployment, and more conservative capital.

Those constraints act like a filter.

They force investors to ask:

Can this scale like software in this environment?

Can we underwrite renewals when buyers demand proof?

Do data rights make defensibility clear, or dangerously fuzzy?

This is why the same pattern shows up in wildly different stories.

Lovable is the productised adoption route.

JUPUS is the compliance-grade retention route.

Alfa is the public-market mirror.

Palantir is the warning label about delivery.

Different paths.

Same underwriting instinct.

The dinner-table version

A Series A multiple is not a prize.

It is a price for uncertainty.

Lower uncertainty about renewals, higher multiple.

Lower uncertainty about margins, higher multiple.

Lower uncertainty about delivery scaling, higher multiple.

Lower uncertainty about data rights and defensibility, higher multiple.

“AI” can reduce uncertainty.

Or it can add it.

That is the point most people misunderstand.

When the opposite is true

Sometimes you cannot show the evidence yet.

If you are extremely early, cohort history is short.

Then investors will lean more on category, team, and speed of adoption.

But notice what even that relies on.

A belief that the adoption you see today will become durable revenue tomorrow.

Lovable itself is proof that markets will pay very high prices.

Yet even in that story, the bet is not “AI”.

The bet is that the cohorts will mature into something stable.

What founders should do before a Series A

Most founders cannot change their category quickly.

They can change what feels underwriteable.

First, instrument cohort retention in a way that makes tourist churn visible.

Do not hide it in blended averages.

Second, map delivery to margin.

If you use forward-deployed engineering, show the path to repeatability so software margin dominates over time.

Third, treat compliance and data rights as revenue drivers, not legal footnotes.

JUPUS made that explicit because privilege and GDPR constraints change willingness to renew.

If you are negotiating terms, remember that price is not the only lever. Some clauses can matter more than headline valuation, as we explored in The term sheet that lets you keep shares, then stops you buying machines.

For investors: what the money is really afraid of

Investors are not allergic to risk.

They are allergic to hidden risk.

In 2024-2025 Western Europe, the hidden risks that keep showing up in the brief are consistent.

Retention that looks fine in aggregate but falls apart in cohorts.

Gross margin that looks fine until services and implementation costs expand.

Defensibility that depends on data you do not clearly control.

So experienced investors ask for proof that survives diligence.

Cohort curves, because tourist churn is common in AI.

Delivery detail, because forward-deployed teams can become permanent.

Data rights clarity, because ambiguity turns differentiation into hope.

This is why fundraising sometimes stalls even when a company is “growing”. It is often about what cannot be underwritten, a theme we cover in Why Fundraising Stalls, and What Serious Companies Do Differently.

GI Network's view: Founders think they are selling a story. Series A investors in Europe think they are buying a renewal stream. Make the renewal stream legible, and the story starts working again.

What GI Network would do in this situation

GI Network would turn the valuation conversation into an underwriting pack before investor outreach.

We would rebuild the narrative around cohort retention so “tourist churn” is quantified rather than hand-waved. We would map gross margin and any services mix to the deployment model, so investors can see whether you are scaling software or scaling headcount. And we would pressure-test data rights and dependencies so the defensibility story holds up in diligence.

Then we would rehearse the investment committee objections that repeatedly appear in this brief’s 2025 Western Europe lens: renewal durability, payback, implementation intensity, and compliance constraints.

If you need a broader reset on how trust gets built before price gets discussed, see Capital moves on trust: what funders expect before they look at anything.

The takeaway tool: the Durable ARR Multiple checklist

Use this before you pick a number for your Series A.

Four questions, answered with evidence.

1) Is retention legible?

Can you show cohorts that separate tourist churn from a retained core, as a16z’s framework demands?

2) Is efficiency backed by gross margin?

Does scaling increase software margin, or do services and implementation expand with revenue?

3) Is deployment repeatable?

Are you closer to Lovable’s productised adoption, or closer to a forward-deployed motion investors will discount?

4) Are data rights clear?

Can you explain defensibility and availability in a way that survives diligence?

Answer “yes” to all four, and the premium can appear.

Answer “no” to two, and you will feel the repricing Alvarez & Marsal described.

That is the real lesson of Lovable’s $1.8b moment.

The multiple was not a compliment.

It was a bet that the revenue would behave.