The most expensive word in your pitch deck

Somewhere in the UK this week, a founder building something genuinely hard - a quantum computer, a new semiconductor process, a fusion reactor - is rewriting their pitch deck to sound more like an AI company.

You can hardly blame them. AI took 74% of all venture capital invested in the UK in the first half of 2026, some $12.6 billion of $17 billion, per Dealroom. Across Europe, PitchBook's Q2 European Venture Report puts AI at 60.2% of deal value, up from 37.8% last year. Numbers like these exert a gravitational pull on positioning. If three-quarters of the money is flowing to one theme, surely the rational move is to stand in the flow?

So, the temptation whispers: add the word. Reframe the roadmap. Let the algorithms take top billing and move the engineering to slide nine.

Our argument this week is that this instinct, understandable as it is, now points in exactly the wrong direction. The AI label has quietly repriced. What once functioned as a fast pass into investor meetings now routes you into the most scrutinised queue in venture. And beneath the headline, the data shows capital rotating hard into engineering-led sectors that carry no AI badge at all.

The founders reading the 74% as an instruction are misreading it. Here is what the number actually says.

A value story, not a volume story

Start with what the headline conceals. AI's dominance is a story about deal value, not deal count. On PitchBook's European data, AI accounts for 60.2% of the money but only 38.4% of the deals: six in every ten European venture deals this year have no AI tag at all. The gap between those two figures is explained by round size, not sector lockout. Mega-rounds of €100 million or more now constitute 55.4% of European deal value while representing just 1.8% of deal volume.

In other words, the AI supremacy that dominates the headlines is overwhelmingly a late-stage, mega-round phenomenon: a handful of enormous cheques written to a handful of companies. At Seed and Series A, where most readers of this note are competing, the market looks nothing like the headline. The capital chasing frontier AI labs was never available to an early-stage engineering business anyway, whatever its deck says.

Nor does the 74% describe what investors will fund. PitchBook's regional figures are based on where companies are headquartered, not where the money comes from, and capital in European venture is mobile: US investors alone participated in 21.8% of European deals in the first half. That same mobile pool stands behind the DACH region's top rounds this quarter, which spanned robotics, cleantech, space and mobility, with only one AI company among the top five. Our read for a British founder: the UK's concentration reflects what British companies raised, not a limit on what carefully selected international investors will back.

The quiet re-ranking

Now look at where the money is actually moving, because the most interesting story in the European data has nothing to do with AI's share.

PitchBook ranks twenty verticals by annual deal value. In 2026, Advanced Manufacturing has surged to fifth, from twelfth last year, having sat between fourteenth and eighteenth for nearly a decade. Its €6 billion in the first half already exceeds its entire 2025 total by half as much again. Robotics and Drones sits twelfth, from eighteenth as recently as 2022, with first-half value already within touching distance of last year's full-year figure. SpaceTech, bottom of the table for a decade, is already nearly 60% ahead of its 2025 total. CleanTech ranks third, pacing 59% above last year. Meanwhile Fintech has fallen from fourth to tenth, and Mobile from third to fifteenth.

One caution on reading the table: PitchBook allows a single company to carry several vertical tags, so some of these rising deals are counted under an engineering vertical and AI at the same time. But many notable others carry no AI moniker. STARK, a German robotics company with no AI tag, closed a €500 million round - at early stage - that sits in Europe's top ten for the quarter. What unites the risers is engineering: capital is following hard engineering whether or not AI appears on the tin, while the verticals losing rank are software and consumer.

Why the rotation? Our read is that it is partly a hedge. Nobody yet knows where the value in the AI stack will ultimately settle - models, infrastructure, applications, or somewhere else entirely - and PitchBook itself flagged concentration around AI valuations as a watchpoint at the start of the year. Investors whose portfolios are already full of richly valued AI positions have every reason to want ballast: real engineering, real customers, revenue that does not depend on where the stack shakes out. Non-AI conviction bets are becoming portfolio construction, not contrarianism.

The label has been repriced

If the label no longer explains where the money goes, what does wearing it actually buy you? Increasingly, scrutiny.

In April 2025, the US Securities and Exchange Commission charged the founder of shopping app Nate with fraud, alleging he raised over $42 million on claims of AI-driven automation while transactions were allegedly processed manually by contract workers. A criminal indictment was filed in parallel. Whatever the outcome, we doubt a single diligence team missed the message: AI claims are now checked, as the consequences of stretching them may have a courtroom attached.

The behavioural shift is visible in investor language. KPMG's Q2 Venture Pulse reports European investors increasingly prioritising AI-native companies over software businesses bolting AI onto existing offerings. Oxx's Mikael Johnsson predicts AI investments will face the same scrutiny as any other software investment this year. Index's Nina Achadjian warns of false positives of product-market fit in AI: revenue from enterprise customers rushing to trial the latest tools, without true return on investment for those customers. None of this means the pull isn't real; the share figures above are real money. It means the label now raises the evidential bar rather than lowering it. You are volunteering for the queue where investors have learned to disbelieve.

The company that never wore it

Consider a company that faced this choice and declined the label. A declaration of interest: Greenjets is a client of ours, and we have known the business since its earliest days.

On 17 July, the British propulsion business announced a $40 million Series A led by Blossom Capital, with participation from the NATO Innovation Fund, the National Security Strategic Investment Fund and existing investors. The announcement came days after the Ministry of Defence selected Greenjets under the LCADE programme to develop a British low-cost drone interceptor. Read the announcement and you will find no AI story: the round rests on propulsion engineering, a growing IP portfolio, sovereign supply chains, and the brutal new economics of defending against cheap drones.

Look at that syndicate. A generalist Series A firm known for high-conviction bets, sitting alongside strategic, state-adjacent capital. Each backed the same thing: engineering depth located in a theme with genuine strategic weight, proven by a paying government customer. We suspect a round like this would have fought hard for attention in 2021, when capital chased software narratives above engineered products. In 2026 it is the shape of deal the market is rotating towards - which is precisely what the vertical rankings, and STARK's raise, and the repricing of engineering depth we examined three weeks ago, have been saying all along.

Positioning honesty as strategy

We argued in February that hardware was becoming the defining moat of the AI era, and in December that AI-native startups were raising less capital than the noise suggested. This week's argument completes the thought from the other side: for the founder who isn't building AI, the label is no longer a costume worth renting.

The question to sit with is not "how do I sound more like AI?" but "which conviction theme does my engineering genuinely serve, and can I prove it?" Capital is not buying labels this year. It is buying engineering depth located in themes with real strategic weight - defence, energy, manufacturing, space - and it is checking the claims.

The crowded queue is the slow queue. The scrutinised label is the expensive one. And the most valuable word in your deck may be the one you nearly deleted to make room for "AI": the plain one - quantum, semiconductor, fusion, propulsion - that names the hard thing you have actually built.

 
Let's talk.