Independent research on European small caps
Long-form, primary-source research on listed small caps in Switzerland, Italy, Austria and Germany
The Disruptor's Price: Eat Yourself, or Buy Your Executioner
When AI coding made custom software nearly free, every horizontal SaaS company became its own most dangerous competitor's first target. The incumbents have answered in exactly two ways: disrupt yourself before the market does, or buy the disruptor and absorb the blast. monday.com is the purest case of the first. Wix is the purest case of the second. Both are expensive, both are honest, and neither one fixes the seat.
Framework · AI & Business ModelsThe Cement Curve: AI Coding and the End of the Managed Margin
In two earlier notes we argued that AI deflates capability and concentrates value onto proprietary substrates — and that the market's error was filing every software business under capability. The same framework, run on the compute stack itself, produces a harder conclusion. A cloud provider's product splits in two: raw infrastructure, which is physical, and the managed services layered on top of it, which are software — and the margin lives in the software. The premium of a SageMaker over the EC2 instance underneath it, of DynamoDB over a database you run yourself, was always a labour arbitrage: you paid the markup because doing it yourself cost more in engineers than the markup cost in dollars. AI coding collapses the labour side of that arbitrage. When an agent can assemble the open-source stack — the orchestrator, the inference server, the IAM policy, the patch cycle — the managed margin deflates toward zero, and what remains is a commodity business bounded by energy, with the economics of a cement plant. The hyperscalers survive this, because they sell something an agent cannot assemble: liability, brand, and the contract. The neoclouds are caught precisely where the framework bites — renting bare GPUs to customers who want exactly the commodity. This is a framework note, not an initiation; the companies named are illustrations. No rating.
Where attention is scarce, mispricing is not.
No coverage, no scrutiny
Below €500m of market value, sell-side coverage thins; below €150m it has all but vanished. Prices there are set by fund flows, family holdings and habit — rarely by anyone who has read the accounts.
A few names, fully understood
We would rather know a handful of businesses completely than skim hundreds. An initiation is months of filings, registries, site visits and interviews — and most companies we examine, we decline.
Paid to wait
We look for cash-generative businesses priced to hold up even if nothing goes right. The safety is in what we pay; a margin recovery, a re-rating or first coverage is upside we haven’t paid for.
Research in your inbox
New initiations, updates and flash notes on Europe’s uncovered small caps. No noise, no spam.
The Winner's Share: Who AI Actually Pays in European Software
Software survives AI by not losing — but survival is only half the trade. The other half is capture: if generative AI deflates the cost of capability while concentrating value onto proprietary substrates — licensed data, systems of record, liquidity pools, settlement rails — then someone owns those substrates, and the repricing that punished the whole software column has mispriced them twice. Europe is where the mispricing is most extreme, because Europe's software champions are, by an accident of industrial history, disproportionately substrate businesses rather than capability businesses. The market imported an American category fear onto a different asset mix. This is a framework note, not an initiation; the companies named are illustrations. No rating.
Framework · AI & Business ModelsThe Loser's Game: Why Software Survives AI by Not Losing
Most of the software complex was repriced in 2025–26 on a single fear: that generative AI collapses the value of capability. The fear is right about capability and wrong about which businesses sell it. The durable franchises in an AI world are not the ones that hit the most brilliant AI product — they are the ones that avoid the one unforced error that ends them. We borrow a fifty-year-old idea from tennis to explain why. This is a framework note, not an initiation; the companies named are illustrations, not recommendations. No rating.
Equity Research · Frankfurt Stock Exchange (Prime Standard); SDAXMedios: Six Quarters of Margin Expansion, Then a Margin Miss
Medios is Germany's largest specialty pharmaceutical distributor and a top-three European specialty pharma platform, with FY2025 revenue of €2.08bn (+10.4%) and EBITDA pre of €93.1m (+17.8%) at a 4.5% margin, expanded 30 basis points year over year. The 2026 guidance — revenue €2.0-2.12bn, EBITDA pre €94-102m at ~4.8% margin — implies continued mid-single-digit organic EBITDA growth and a credible multiple-expansion thesis against Fagron at ~9x forward EV/EBITDA versus Medios' ~4.5x on FY2026 guidance today. The bear case rests on Q1 2026's -7.9% EBITDA pre print, which arrived with a new CEO and CFO on the watch and a capital-markets narrative that the previous regime had not solved. We publish no rating; the question is whether the multiple rerates or the operating trajectory disappoints first.
Equity Research · Euronext Growth MilanCyberoo: The Deadlines Are the Demand
A fiscal year stretched to fifteen months makes every headline growth rate incomparable, and the extra quarter added almost nothing: on the company's own like-for-like figures, the January–March 2026 stub contributed roughly €1.7m of cybersecurity revenue against a 2025 monthly average of about the same. Underneath, monthly EBITDA fell by roughly 30% while margin dropped eleven points. The genuine news is working capital — receivables from the largest shareholder fell from €11.5m to €2.6m — but €5.3m of that left by offset rather than cash, and an unconsolidated leasing subsidiary with a €15m facility now sits beside the group's receivables machinery. No rating.
Equity Research · Euronext Growth MilanTECMA Solutions: An Italian Proptech Sold to the AI Woodchipper
TECMA Solutions is a €17m Italian proptech that builds software, configurators and digital marketing infrastructure for residential real-estate developers. It lost €1.06m on €12.4m of revenue in FY2025, down 11% on the year, has accumulated losses larger than its equity, and trades on EGM (Euronext Growth Milan), the segment most exposed to small-cap flight. None of those facts is the story. The story is that TECMA is the natural counter-example to our own SMG thesis: a company that sells capability to real-estate developers, not liquidity between them, and is therefore the kind of name where the AI-disintermediation argument actually has purchase. We initiate coverage at Marketweight. The market is roughly right on the company; we want to own it only on a substantial derating or a tangible proof that its international and recurring-revenue shift survives the cycle.
Equity Research · Vienna Stock Exchange (direct market plus); Munich m:accessWolftank Group: Remediation, Above and Below the Line
The 2026 turnaround narrative rests on a 2025 base year whose reported profit was produced by an internal revaluation; the operating business lost roughly €3m, minority shareholders own the profitable half of the group, and €27.7m of bank debt matures within twelve months. Against that: a genuinely strong Q1 and an order backlog whose arithmetic raises a €40m question.