Somewhere between 2016 and 2021, “hardware startup” became a phrase you didn’t say out loud in a pitch meeting. It signalled long timelines, tooling costs, inventory, customs, a bill of materials that only goes up when your volumes go down. Founders learned to bury the physical product three slides deep and lead with the platform.
That taboo is over. Investors did not suddenly develop a taste for injection moulding. What changed is that the categories now pulling in the most capital in Europe (defence, robotics, semiconductors, energy, space) are all, structurally, hardware businesses. The money came back under a different name, with different expectations attached, and if you pitch it the way people pitched hardware in 2015 you will still get a polite no.
The rebrand is real, and the numbers are behind it
European deep tech took $20.3bn in VC funding in 2025, 32% of all European venture investment, up from 15% a decade earlier, and within a few percent of the 2021 peak. Read that share number twice. A third of European venture capital now goes to companies that build physical or physics-adjacent things. Nobody can call that a niche.
Within that, defence, security and resilience went from 20% of European deep tech funding in 2022 to 43% in 2025. The sector raised a record $8.7bn across the year, up 55%. Helsing’s €600m Series D and Quantum Systems’ €340m are the headline rounds, but the early stages are more instructive. The 2026 cohort is full of companies that would have struggled to raise a euro five years ago: Munich’s TYTAN Technologies took a €30m Series A for autonomous interceptor drones with the NATO Innovation Fund participating; Estonia’s Frankenburg Technologies raised €30m for interceptor missiles; Amsterdam’s Onodrim Industries raised a €40m seed for defence manufacturing and sensing. A €40m seed for a company that makes things.
Robotics tells the same story louder. Robotics startups raised $18.8bn in the first half of 2026 alone, more than all of 2025 ($15bn) and comfortably past the previous annual record set in 2021. Germany’s Neura Robotics closed a Series C reported at up to $1.4bn. These are capital-intensive businesses that depend on factories and exposed supply chains, and they are being funded at software valuations.
The lesson is not “call yourself a defence company.” Investors have re-learned that some physical products have durable moats and buyers with real budgets in government and enterprise, and that their pricing power does not evaporate the moment a Shenzhen competitor lists a lookalike. If your product sits in one of those categories, say so plainly and early. If it doesn’t, forcing the label will fail the first technical diligence call.
The AI build-out made atoms strategic again
A second engine sits underneath all of this, and it runs whether or not you have anything to do with AI.
The five largest US cloud providers are planning something in the region of $660-690bn of infrastructure capex in 2026: Amazon around $200bn, Alphabet $175-185bn, Microsoft north of $120bn, Meta $115-135bn, Oracle roughly $50bn. That money does not stay in software. It lands as concrete, transformers, switchgear, cooling loops, cabling, power electronics and the people who install them.
And it is hitting physical limits. Microsoft has disclosed an $80bn backlog of Azure orders it cannot fulfil because it cannot get power. The IEA expects global data centre electricity consumption to roughly double between 2022 and 2026. When the binding constraint on the most capitalised industry in the world is grid interconnection and heat rejection, hardware companies in power conversion, thermal management, sensing, grid equipment and industrial automation become an easy sell.
If your product touches any part of that chain, even indirectly, even as a component, that framing is worth more to you than any amount of polish on your consumer story.
What a hardware seed has to show now
The bar is legible, which is the good news. It is also higher than founders expect, and higher than it was in the last cycle.
First, a working prototype. Not a render, not a dev kit with your logo on it. Something you can put on the table and switch on. In 2026, with the tooling that exists, an inability to produce working hardware reads as an execution signal, not a funding gap.
Second, unit economics you can defend, modelled honestly and backed by real quotes. Nobody expects gross margin at seed; investors expect you to know your numbers and have a credible path. Benchmark against actual hardware companies rather than SaaS: mature differentiated hardware generally lands somewhere in the 40-60% range, and the businesses that do better than that usually do it by attaching services. Apple’s product margin sits in the thirties; its blended margin is pulled up by services running above 70%. That is the entire argument for a recurring layer in one sentence.
Third, that recurring layer, or a serious plan for it. Subscription, consumables, data, service contracts, licensing: the mechanism matters less than the fact that revenue does not stop the day the box ships. This is the single most common gap in first-time hardware decks, and the one that most reliably converts a maybe into a no.
Fourth, and this is the European-specific one, a manufacturing answer. Who builds it, at what volume, with what lead time, and what happens to your landed cost when tariffs or shipping move. A founder who can talk fluently about their contract manufacturer, their second-source strategy and their certification path is treated as an operator. Say “we’ll figure out manufacturing after the raise” and you are a science project. This is the part we spend most of our time on with clients at RMBG, and it is consistently underweighted in first decks.
Consumer gadgets are still, essentially, unfundable
Founders lose years to this one.
US consumer electronics startups raised under $300m across all stages in 2024, the slowest pace in at least a decade, and roughly one dollar out of every three hundred of US venture capital. Even at the top of the last bull market, in 2021, the whole category took $2.2bn, under 1% of US VC. There was never a boom, just a trickle, and then less than a trickle.
The reasons are not mysterious. Consumer hardware is deflationary: your product gets cheaper every year while your costs don’t fall as fast. Customer acquisition is paid and rising, and retail takes margin and dictates terms. The category’s canonical outcomes are Magic Leap ($3.5bn raised) and Essential ($330m raised), and every generalist partner remembers them.
If you are building a consumer gadget, the honest read is that venture capital is not your funding source, and structuring your company as if it will be is the mistake. The business can still be a good one. It just has to be financed like a consumer products business: pre-orders, distribution partners, working-capital debt, revenue-based facilities, patient equity from people who understand physical goods. Plenty of profitable independent companies get built that way. Very few of them raise a Series A.
Crowdfunding didn’t die. It got taken over.
The received wisdom is that Kickstarter is finished as a hardware launch channel. The data says otherwise.
Kickstarter had its biggest year ever in 2025, and Design & Technology had the biggest year in the category’s history. But the winners were not newcomers. The all-time funding record, $46.7m, went to the eufyMake E1 UV printer, from a sub-brand of Anker, a company with a mature supply chain, existing distribution and a large owned audience. The other standouts were Snapmaker, already an established 3D printing manufacturer, and Peak Design, a brand with more than a decade of campaigns behind it.
So the platform is healthier than ever and simultaneously worse than ever for the first-time founder. What used to be a discovery channel, where an unknown team with a good video could find ten thousand strangers, has become a launch and pre-sell channel for companies that already have brand, audience and factories. The crowd doesn’t discover you anymore; you bring the crowd.
Treat it accordingly. Crowdfunding in 2026 is a demand-validation and working-capital tool for a product you can already build, run on an audience you already own. It will not replace your financing or find you product-market fit. If you cannot fill the first day yourself, the platform will not fill it for you.
The European stack that isn’t venture capital
This is where European founders have an advantage they routinely fail to use.
Start with the non-dilutive R&D money. The EIC Accelerator carries a €414m budget for 2026, offering up to €2.5m in grant plus up to €10m in equity through the EIC Fund; the minimum blended-finance ticket rose to €1m and there are six bimonthly cut-offs. It is competitive and the 2026 process added technical due diligence at full-proposal stage, so budget real time for it. Five thematic challenges account for €220m of that, so check whether your work maps onto one before you write a general application.
Then there are the national schemes. In the Netherlands, WBSO is the workhorse: a payroll-tax credit of 36% on the first €391,020 of qualifying R&D basis, 50% for starters, 16% above that, against a 2026 budget of €1.817bn. You can apply through the year. It is not glamorous and it never shows up on a cap table, but for a small engineering team it is often worth more than a bridge round. Regional and MIT-type schemes sit alongside it. Check current terms with RVO rather than trusting a blog, including this one.
State-backed deep tech capital is the third layer. The Dutch Deep Tech Fonds received another €360m in June 2026, taking it to €610m, targeting semiconductors, photonics, quantum and AI, with Axelera AI, Nearfield Instruments and QuantWare already in the portfolio. The context is sobering: Techleap has found European deep tech companies raise 3.4 to 4.2 times less per round than US counterparts, and 71% of first investments in Dutch deep tech come from public money. That is a structural weakness for the ecosystem and a practical fact for your company. In Europe, public capital is frequently your first cheque, and there is nothing second-rate about that.
Debt comes into play once you have something to lend against. The EIB writes venture debt tickets of €10-50m for commercial-stage innovators who have already raised professional equity, typically bullet-repaid. Below that, purchase-order financing against signed enterprise or government orders, inventory facilities, and revenue-based financing on a recurring layer all do work that equity does badly. The rule of thumb: equity funds uncertainty, debt funds certainty. Paying 20%+ dilution to buy components against a confirmed order is a bad trade.
Last, strategic and corporate investors. Harmattan AI’s $200m Series B came with Dassault Aviation. Quantum Systems’ trajectory is inseparable from German government orders. In industrial and defence hardware, the strategic investor often brings what you need most, whether that is qualification, a channel or a first serious customer, and the governance cost is worth reading carefully rather than reflexively avoiding.
Milestones, and what they cost you
Rough shape, and treat it as orientation rather than gospel: current market benchmarks put median seed dilution and median Series A dilution both around 18%, on medians near $4.1m raised at $24.3m post at seed and $14.4m at $80m at A. Those figures come from a predominantly software dataset, and deep tech seeds have been running larger, around $5m on average in Europe versus $4m for regular tech, because you cannot bootstrap a tooling run. Expect two rounds to cost you roughly a third of the company before an A closes, plus option pool.
What that capital has to buy is specific. By seed: working hardware in the hands of real users, a bill of materials with real quotes, a named manufacturing partner, a clear certification path, and either early revenue or signed intent from buyers with budget. By Series A: pilot deployments converting to repeat orders, unit economics measured rather than modelled, a functioning recurring revenue line, and evidence you can build the hundredth unit as well as the first. The transition that kills hardware companies is ten-to-a-thousand, not prototype-to-product. Investors at A are underwriting that transition specifically, so structure your seed milestones to de-risk it visibly.
Position first, then raise
The 2026 market rewards founders who are honest about which business they are in.
If you are building infrastructure, industrial, defence, robotics or energy hardware, the capital is there, European institutions are looking, and the constraint is your ability to show working technology and a manufacturing answer. Lead with the category, not with the enclosure.
If you are building a consumer product, stop optimising your deck for VCs who will never invest, and go build the financing stack that fits: grants for the R&D, pre-sales for demand, debt for inventory, strategic partners for distribution. You will keep more of your company, and you will spend your months on customers rather than on Sand Hill Road or Zuidas.
Either way, the order is the same: position, then prove, then raise.
Sources
- The 2026 European Deep Tech Report (Dealroom)
- Europe’s defence and resilience startups hit $8.7B in 2025 — top deals in early 2026 (Vestbee)
- Europe’s Defence Tech Gold Rush (deeptech.build)
- Robotics Startups On Fire As Venture Funding Surges To Record Numbers In 2026 (Crunchbase News)
- AI Capex 2026: The $690B Infrastructure Sprint (Futurum)
- Consumer electronics funding slump (Crunchbase News)
- A Year in Review: 2025 Kickstarter Highlights
- eufyMake E1 breaks the all-time Kickstarter funding record
- EIC Accelerator 2026: key updates, deadlines and funding (Zabala)
- WBSO (RVO)
- Government puts €360 million more into high-risk tech fund (DutchNews)
- Invest-NL Deep Tech Fund
- EIB Venture Debt
- VC startup fundraising benchmarks (Carta)
- Gross margin for hardware startups at seed stage (SeriesOps)