The Blink Edition No. 3

The European private markets briefing you can't afford to miss. Deals, signals, and moves, decoded every Monday by SeedBlink.

Five years ago, you could raise money on a story. A strong technical team, a good deck, a nice idea. "Let's do it." That's over. We don't invest in stories anymore, and investors across the region are making the same call.

What replaced the story is validation. Not the promise of traction, but traction itself, ideally paying customers. And the goalpost moved again: it's no longer growth at all costs, it's profitability. That's a major shift from the years when growth was almost the only thing that mattered, and it changes what a fundable company looks like from both sides of the table.

There's an old line in tech: "if you build it, they will come." It isn't true. You still have to sell it. You have to market it. For founders, that means treating go-to-market as seriously as the product. For investors, it means the diligence question has changed: not "how big could this get?" but "who is actually paying, and why?"

None of this makes building or backing companies harder than before. It makes it clearer. The founders raising well right now aren't telling better stories. They're showing better evidence, and the investors doing well are the ones who know how to read it.

Validation is the new bar for raising.

But here's the uncomfortable part, and it matters as much for investors as for founders: even validation has a shelf life now.

The old logic was simple. Prove product-market fit, and you've earned a durable advantage. AI is quietly breaking that assumption. As software gets faster and cheaper to build, the advantages that once lasted years can now be matched in months. Morningstar put numbers to it: reviewing 132 companies with economic "moats" in early 2026, it downgraded 40 of them, with the heaviest pressure on internet software, enterprise software, and IT services, exactly the models built on repeatable, automatable work. For anyone holding a portfolio, that's a repricing signal.

The cost curve points the same way. Gartner projects that inference on a trillion-parameter model will cost over 90% less by 2030 than in 2025, a dynamic it calls a "commoditization trap." When the underlying capability gets cheap, the premium built on top of it erodes. A business model that's profitable today can be undercut within months, and product-market fit is no longer a fixed target you hit once, but a moving one companies have to keep re-earning.

So the question on both sides of the table has quietly changed. For founders: "will this traction still matter in two years?" For investors: "am I underwriting a durable advantage, or a temporary one?" The moats that survive aren't features, which get copied. They're the things that compound and resist automation: proprietary data, a trusted brand, a network, distribution, speed of execution.

Resilience keeps a company alive. Validation gets it funded. But in an AI market, the real question, whether you're building or backing, is whether what works today still works tomorrow.
‍
‍

“Can’t Touch This”

In almost every deal we see, the hardest part isn't raising, but simply existing across borders.

That's why we're behind Allied for Startups' Can't Touch This campaign, and the push for EU-Inc: a single European company form, usable across the Single Market, sitting alongside national systems rather than replacing them, with the same fast, digital incorporation everywhere. Our CEO Andrei Dudoiu joined their latest video to make the case.

The detail that matters: what's on the table right now risks losing what would make it useful. Swap a Regulation for a Directive, and it gets reinterpreted 27 times, quietly recreating the fragmentation it was meant to remove. From where we sit, running deals across borders, two things matter most: one legal form recognized the same way in every member state, and a framework fast enough to keep up with an actual funding round.

Watch the video.
‍
‍

France 🇫🇷

Gradium raises $30M, seed past $100M

Paris voice AI startup Gradium pushed its seed past $100M just seven months after launch, one of Europe's largest ever, betting that ultra-low-latency voice becomes a primary interface for AI agents rather than a feature bolted onto chatbots.

Read more
‍
‍

United Kingdom 🇬🇧

Fleek raises $25M Series B

London-based Fleek raised $25M, backed by eBay and Vinted's early investor Burda, to digitise a $200bn+ secondhand fashion trade still sorted by hand, betting its proprietary transaction data becomes the industry's intelligence layer.

Read more

‍
‍

Belgium Croatia 🇭🇷

Polysense raises $10.7M seed‍

Belgian startup Polysense raised an oversubscribed $10.7M to put AI-powered quality control on food production lines, targeting an industry where manufacturing accounts for 19% of all EU food waste, a reminder that AI's clearest ROI is often on the factory floor, not in chat.

Read more

‍

Investing beyond seed: Maria Koletsou on syndicate investing in the AI era

Part one of our conversation with Blossom Ventures' founder on why she treats the first check as an entry point, not the relationship, and why she's become "a more structural investor and a faster no." A sharp read on investing in lines, not dots, as AI shortens how long any advantage lasts. Part two coming soon.

Read more

‍

The AI wrapper is dead: 3 approaches to verticalization

NFX's Pete Flint on why "AI for X" turned out to describe a temporary advantage, not a business. The companies surviving the wrapper era own the whole workflow, not just the model, a useful lens on where defensibility actually lives as frontier labs productize everything.

Read more

‍

The CIO's choices are clear in 2026

Across 87 public SaaS companies, only two sectors are up over the year: infrastructure and security. The seat-priced application layer is down 36%. Tunguz shows how the market is already pricing the same question founders now face: is your advantage necessary in the AI stack, or exposed to agent substitution?

Read more