Why VCs now fund what they once rejected

What if the reasons VCs rejected your startup five years ago are precisely why they'll fund you today?

The venture capital playbook has been turned inside out. For a decade, mainstream investors chased frictionless markets, viral growth loops, and zero marginal cost distribution. Hardware was "too capital-intensive." Regulated industries were "too uncertain." Complex institutional sales were "too slow." These weren't just seen as messy challenges - they were disqualifying factors that sent founders to the reject pile.

But when AI agents can rebuild entire startup portfolios in 24 hours, software alone offers no protection. The characteristics that made businesses unfundable in 2021 may be exactly what makes them defensible in 2026. As we explored in February's newsletter, hardware and proprietary data have emerged as two of the deepest moats in the AI era. Today, we examine the third and perhaps most underrated for DeepTech founders: the mess itself.


The 24-hour replication test

Mark Hazan's Feltsense experiment provides stark proof of software's new vulnerability. When his AI agents studied Y Combinator's latest demo day batch, they didn't just analyse the companies - they rebuilt them. Every startup. Full products. Ready for market. Within 24 hours.

Core technology that took founders months to develop was reconstructed from public specifications and deployed with new branding overnight. The exercise revealed something sobering: if your competitive advantage is code, you have no competitive advantage. Software replication is no longer a future threat - it's a present reality.

Most founders assume protection comes from human creativity and technical ingenuity. Hazan's data suggests otherwise. The real defence against AI replication doesn't come from the best parts of human nature. It comes from the messiest.


The funding tables turn

The reversal in VC criteria has been swift and decisive. The 2026 European Deep Tech Report shows that DeepTech captured a record 32% of total venture capital in 2025, more than double its share a decade ago - a complete inversion of the pattern that held from 2012 to 2024.

What mainstream VCs once considered disqualifying factors are now precisely the characteristics they seek. Capital intensity protects against competitors who can't match the upfront investment. Regulatory barriers create compliance moats that can't be downloaded. Long sales cycles reflect deep institutional embedding that takes years to replicate.

The SaaS era trained investors to think in terms of viral coefficients and network effects. The AI era demands they think about barriers that no amount of code can overcome. Physical infrastructure, regulatory certification, institutional relationships - these can't be prompt-engineered into existence. For DeepTech founders who spent years being told their businesses were "too hard" for mainstream VC, this represents a profound shift in leverage.


The messy stuff advantage

The third protective barrier emerges from what Hazan terms "the messy stuff" - politics, broken trust, turf wars, bureaucracy. Markets where social dynamics complicate deployment more than technical limitations.

Analysis of enterprise software adoption shows that the majority of failed technology deployments in regulated industries cite organisational resistance rather than technical problems. This friction creates barriers no AI capability can overcome without deep institutional embedding.

Consider defence procurement, NHS trust networks, or utility regulation - sectors where success depends as much on navigating competing stakeholders as solving technical problems. A startup that has learned to work inside that dysfunction, earning trust across factions and understanding unwritten rules, has created something competitors cannot replicate quickly. The complexity isn't a bug - it's the feature that makes these markets defensible.


DeepTech's natural home field

DeepTech companies operate precisely where institutional complexity is highest. These companies disproportionately depend on government and regulated markets, where deployment depends on relationships as much as technology.

Unlike SaaS, where you can test with individual users and iterate quickly, DeepTech sells into systems where buyers aren't users, where procurement involves committees across competing departments, and where the person signing the cheque may never touch the product. Success requires fluency in institutional languages that take years to learn.

Duet's first-hand experience in working with startups addressing the defence sector for example, demonstrates the importance of investing considerable time in relationship building throughout the Ministry of Defence and other government departments before securing multi-million-pound contracts. This relationship-building can often take as long as the product development but creates a moat that technical capability alone cannot breach.


The domain expertise divide

The founder background patterns reveal a crucial distinction. Research by Euclid Ventures shows that founders with domain expertise capture over 80% of exit value, demonstrating that industry knowledge significantly outweighs pure technical or business credentials.

In institutional markets, this pattern intensifies. Defence companies gain real edge when led by ex-military officers who understand procurement cycles. HealthTech thrives under NHS clinicians fluent in trust politics. Industrial automation works when started by former plant managers who know supplier relationships.

Research confirms that DeepTech startups with very technical CEOs from industry backgrounds raise significantly more funding than those with purely business-focused leadership - the opposite of SaaS patterns where young, product-first founders could learn markets through rapid iteration.

SaaS founders could afford to be young generalists because they were selling into frictionless markets where you could test, learn, and pivot. DeepTech founders need institutional grey hair because they're embedding within systems where one wrong move costs years of relationship-building.


Strategic implications for founders

The reframe for DeepTech founders is profound: institutional friction should be viewed as a feature, not a bug. Markets with high institutional complexity often reward founders who can navigate them effectively.

This creates a strategic advantage for founders with domain expertise. If you understand how NHS procurement really works, how defence committees make decisions, or how regulated utilities manage supplier relationships, you possess knowledge that can't be acquired by reading TechCrunch. The messier the market, the more valuable that institutional fluency becomes.

For first-time founders entering these markets, the lesson is clear: hire the grey hair early. And there’s a kicker: The complexity that makes institutional markets harder to enter also makes them harder to leave once you're embedded.


A word of caution

While VCs now recognise the value of these three deeper moats, many are struggling with a fundamental transition challenge. Investors who spent a decade optimising for SaaS metrics often lack the frameworks to properly evaluate hardware intensity, data compounding effects, or institutional complexity. They get part way there - understanding that these characteristics are now valuable - but then let themselves down by insisting on other criteria that made sense in the software world but can be irrelevant in DeepTech.

The result is a mismatched due diligence process: VCs who say they want "hard tech" but then demand software-style user growth metrics, or investors who claim to value regulatory moats but expect SaaS-style sales velocity. Founders should be prepared for this disconnect and give such investors a wide berth. Their money may be the same colour, but such a mindset misalignment will almost certainly cause downstream fallouts.


The takeaway

The AI era has revealed a counter-intuitive truth: the characteristics that made businesses unfundable in the SaaS era may be precisely what makes them defensible today. Hardware creates physical barriers. Proprietary data compounds over time. And institutional complexity - politics, bureaucracy, broken trust - creates human barriers that no amount of technical capability can overcome without deep embedding.

For DeepTech founders, this represents an opportunity to reframe their positioning. The very complexity that used to justify lower valuations now justifies higher ones. The institutional knowledge that seemed like table stakes is actually the scarcity. And the relationships that took years to build create moats that competitors cannot prompt-engineer into existence.

The mess is the moat. For founders brave enough to wade into it, that complexity may be the most defensible advantage of all.

 
Let's talk.