Atoms are back. Are hardware startups suddenly more fundable?

Few reports this year have travelled as fast through founder conversations as Andreessen Horowitz's State of Markets, published at the end of September. Its verdict on hardware fits in four words: atoms are so back. After a decade in which software took the money and the multiples, chips, power, robots and rockets are suddenly where the returns are.

Many UK founders have quoted those four words back to us, usually with the same hopeful conclusion. If hardware is back in fashion, surely the investors who once winced at capital intensity will now take the meeting, and the deck that struggled last year will land this year.

Some will, and we made our own case for hardware as a moat in February. But that reading misses what the report actually measured, and it misses what is happening on this side of the Atlantic.

So, is hardware back in Europe? Not quite. In Europe, the money is not following hardware per se. It is following the buyers who have decided certain hardware is strategic: governments rebuilding defence and industry worried about resilience. Where such a buyer stands behind the demand, cheques are larger than ever. Where none does, the renaissance is much harder to find.

That makes the useful question for a hardware founder a different one. Not whether atoms are back, but whose demand underwrites the cheque.

What a16z actually measured

Read the full deck rather than the headline and its subject becomes clear: American public markets. The engine it describes is the AI buildout, which has in effect converted the free cash flow of the hyperscalers into the free cash flow of chipmakers. Capital-heavy US stocks have overtaken capital-light ones for the first time in a decade, and memory makers have been the biggest winners of that trade.

That is a real shift, and it has a single paymaster. a16z is candid about the risk: chip demand is "highly contingent on hyperscaler capex", and seasoned investors in cyclical industries are always wary of the cycle's peak.

US venture capital tells the same story. The deck shows 86% of US venture deal value going to AI-related companies this year, while noting a quieter shift of venture money towards semiconductors, power and defence.

So in the US, the answer to our question is simple. Hyperscalers underwrite the cheque. For a UK founder whose customers really are those hyperscalers or their suppliers, the story fits. The risk is for everyone else. A founder who cites the new enthusiasm for hardware, when no hyperscaler is underwriting their own demand, risks borrowing someone else's buyer.

The line the headlines skipped

For European founders, one sentence in the deck deserves more attention than it received. a16z writes that it is "not only AI-infra" driving the rotation to atoms, and lists defence spending, grid electrification, rockets, robots and the rebuilding of an industrial base.

That list is a good description of where European venture money is actually going. Dealroom's State of Defence Tech report compares sectors across Europe for 2026 to September. Defence has raised $7.4 billion and is projected to grow 309% on last year. Space ($3.7 billion) is projected to grow 284%, robotics ($9 billion) 263% and semiconductors ($4.7 billion) 122%. European venture as a whole is projected to grow 38%. The categories overlap, so the totals cannot be added together, but the direction is unmistakable.

For the most part, these are not sectors where AI spending is creating the demand. They are the sectors where governments and industry have decided they cannot afford to depend on someone else. In defence, the buyer could not be more explicit. Lakestar's Klaus Hommels, quoted in the same report, says Europe's military leaders have been clear that building defence capability at home also needs "sovereign European financing". The demand, he says, "is there and now the capital must follow."

Who is writing the strategic cheques

The British state is now an investor in this story as well as a customer. In May, the National Wealth Fund made its first defence investment, £25 million into Bristol's Rowden, explaining that "building and sustaining our sovereign capabilities is critical". UKI2S, the government-backed seed fund, has grown to £150 million with money from the Ministry of Defence, the UK Space Agency and the UK Atomic Energy Authority, according to Defence Online.

Welcome as they are, set those sums against Europe's. The EU's SAFE scheme lends €150 billion to governments for defence procurement, mainly from suppliers in the EU, the wider European Economic Area and Ukraine. Since Britain's talks to join it collapsed last November, outside suppliers, UK firms included, can provide no more than 35% of any SAFE-funded project.

So for a UK founder, the strategic buyer that matters most may well sit in Berlin or Paris rather than Whitehall, and access to that buyer may come with a ceiling that no amount of good engineering will lift. That shapes which European contracts a UK company can realistically win, and therefore which demand it can credibly put in front of investors.

Industrial automation, through a different door

This is not only a defence story. Drumbeat Capital's Deep Tech Report, which counts defence robotics separately, finds European robotics funding doubled in 2025. It describes the sector's purpose as automating physical work to address labour shortages, a strategic need of a different kind.

But look at who funds it. Neura Robotics' round of up to $1.4 billion brought in Nvidia, Qualcomm and Amazon alongside Bosch and the European Investment Bank. Neura calls its product "physical AI", a label now spreading across robotics, drones and anything else that moves. The labels blur in the data too: two of Europe's three largest robotics rounds this year went to military drone makers, per Tech.eu.

Labels can be changed. The buyer cannot. Industrial automation wins when it answers a need someone has decided is strategic, whether labour shortages, reshoring or dual use. And robotics is still where Europe trails the US furthest outside AI, raising 9.6 times less in 2025, per Drumbeat.

The sectors left off the map

Now look at what is missing from that list. Climate, medtech and mobility hardware are long-standing European strengths. In the same Dealroom comparison, energy is still raising serious money, $7 billion so far, but at 25% growth it is losing share of a market growing at 38%. Health is down 6% and transportation down 8%. Novel energy funding fell 41% in 2025, according to the European Deep Tech Report.

The difference is not the quality of the technology. It is that these sectors are not where this year's new strategic urgency has landed. Governments and large industrial customers are not yet treating these products with the urgency they now bring to defence or chips, so investors see less committed demand behind them. The technology may be excellent. What investors cannot yet see is a customer committed to buying it at scale.

There is a way back in, and a16z points to it. Its own deck rates power and grid equipment as very tight bottlenecks in the AI buildout, with lead times for transformers of up to five years. Several of the suppliers it names, Siemens Energy, Schneider and ABB among them, are European. An energy founder who can show they relieve that bottleneck is no longer selling climate. They are selling resilience.

Bigger cheques, fewer of them, earlier than you think

Even where the money is flowing, most of it goes to very few companies. Across European venture as a whole, rounds of €100 million or more took 55.4% of deal value in the first half, while making up just 1.8% of deals, per PitchBook's Q2 European Venture Report. In defence, Dealroom finds such mega-rounds made up over 85% of this year's funding. US investors took part in only 21.8% of European deals, but those deals accounted for 69.2% of the value.

Last week brought a clear example. Universal Quantum, the UK quantum hardware company, raised over $100 million at Series A, co-led by Silicon Valley's DCVC and a sovereign-backed quantum fund. Its anchor customer is a $77 million contract with Germany's aerospace agency, DLR.

In July we argued that Britain funds scaling badly, strong at Seed and Series A but thin beyond. Universal Quantum suggests the line is moving earlier: overseas customers and overseas capital now matter at Series A.

Our view is that this concentration raises the bar for everyone else. When the largest Series A rounds go to companies with a committed strategic customer already in hand, investors begin to expect the same evidence from every hardware company at that stage, including those raising a tenth of the amount.

Before you quote a16z in your next pitch

The a16z deck is worth reading, and worth quoting. Just ask yourself three questions before your next investor meeting rather than during it.

Who, specifically, needs your product to exist, and is that need backed by a buyer with real budget and urgency, whether a hyperscaler, a government or a large industrial customer?

Can you sell to that buyer on equal terms, or is your access capped?

And if you removed the word "hardware" from your story, or the words "physical AI", would the demand still stand on its own?

None of these questions is answered by the hardware itself or by the sector it belongs to. Whatever the a16z headlines may tempt founders to believe, a fashionable sector may win you the meeting, but only a real customer with a real need will close the round.

Atoms are back. Some hardware startups are now more fundable than ever. They are the ones that can name who will pay for what they build.

Let's talk.