In June 2025, Anduril Industries was worth $30.5bn. By March 2026 it was reportedly raising again at around $60bn, close to doubling in nine months. Helsing, the European answer, went from a roughly €12bn valuation in mid-2025 to a reported $18bn round a year later. A defence company is now a credible venture outcome. Five years ago that sentence read as a category error.

These businesses look nothing like the primes that came before them. They are software-first. They build fast, ship cheap and attritable hardware, and sell autonomy rather than airframes. And here is the part most people in Britain have missed: increasingly, they are building here.

Anduril has worked its way through Programme TALOS with UK Strategic Command on a force-protection contract worth £17m that can rise to £24m, with a larger award reportedly in the region of £100m expected later in 2026. Helsing has committed £350m to the UK and opened its first British “resilience factory” in Plymouth. Stark, the loitering-munitions company Sequoia backed at a $500m valuation, chose Swindon for its first UK drone line. Tekever, Europe's newest defence unicorn, is putting more than £400m and over a thousand jobs into British drone production, having already supplied the MoD with £270m of systems bound for Ukraine.

The factories are being built in Britain. The companies that own them are not British.

That is the asymmetry worth sitting with. Britain has become a place where the world's defence champions want to manufacture, recruit and test. It has not yet become a place that produces many of its own. The talent is here: DeepMind alumni, a deep aerospace supply chain, more ex-military operators than any market in Europe. The demand is here. The ownership is somewhere else.

What changed

For years the honest answer to “why isn't Britain building Andurils?” was procurement. The MoD bought slowly, in decade-long cycles, from a handful of incumbents, on terms that punished anyone without a balance sheet to wait. A startup could win a trial and die before the contract landed.

The 2025 Strategic Defence Review, NATO-first and with all sixty-two recommendations accepted, set out to change that. Defence spending is committed to 2.5% of GDP by 2027 and 3.5% by 2035, inside a wider NATO target of 5%. More important for founders is the machinery underneath. UK Defence Innovation launched with a ring-fenced £400m a year, folding in DASA and the scattered innovation units that came before it. And procurement is being cut into three tiers: major platforms inside two years, upgrades inside one, and commercial software on three-month contracting cycles.

Three-month cycles. For a software-first defence company, that single line is the difference between a viable business and a slow death.

The capital is arriving to match. The NATO Innovation Fund, a €1bn vehicle backed by two dozen nations, put three of its first four defence bets into UK-based companies. European defence, security and resilience startups raised a record $8.7bn in 2025. London now has dedicated dual-use funds, and names like Sequoia, Thiel Capital and In-Q-Tel are writing cheques into companies that build on British soil. The squeamishness that kept generalist investors out of the sector is breaking down quarter by quarter.

Where the work actually is

None of this makes building one easy. The lesson of Ukraine, that cheap, attritable autonomy beats exquisite platforms, has rewritten what the MoD wants to buy. But knowing what to build is not the same as being bought. The defence customer ladder is unforgiving: design partner, then a funded pilot, then a programme award, then prime. Each rung speaks a different language. Most founders, however good the technology, stall between the pilot and the programme. That is the point where the business has to prove it can carry weight, not just demonstrate it.

That gap is the whole game. It is the difference between a clever drone and a defence company. Closing it is commercial work, not engineering work: sharpening the proposition until a programme manager can say yes, building the case the procurement system can actually act on, and being in the room when the timing is finally right.

AlphaFold, the breakthrough that made protein structure predictable and won a Nobel Prize, was built in London. The company commercialising it, Isomorphic Labs, is British too, and in May 2026 it raised a $2.1bn Series B. State it plainly, because it matters: when Britain backs its own science at scale, it can produce a world-beater without anyone needing to move to California.

The harder truth is how rare that sentence still is.

Britain's research firepower is not in question. Oxford, Cambridge, UCL and Imperial; the Francis Crick Institute, the Wellcome Sanger, the MRC Laboratory of Molecular Biology. Pound for pound, this is one of the most productive discovery engines anywhere. The Life Sciences Sector Plan the government published in July 2025 put the sector at around £100bn and 300,000 jobs, and committed more than £2bn behind it. The science is real. The discovery genuinely happens here.

And then, too often, the value leaves. When CellCentric raised its Series C in 2025, not one UK investor was on the round. The pattern repeats: a company is discovered in a British lab, seeded by British angels, and then, at exactly the moment it needs to scale, finds the cheque it needs in Boston or the Bay Area, and frequently follows the cheque across the Atlantic.

Britain is brilliant at the discovery and the first cheque. It is structurally short of the cheque that builds the company.

The numbers bear it out. Britain's domestic share of late-stage funding fell to a low in 2024–25 (by one Royal Academy of Engineering measure, around 9% of funding in 2025), and closing that gap would take somewhere between $4bn and $11bn more a year. The shortage is specific. It is not seed. It is Series B and C: the rounds that turn a validated asset into a business with its own gravity. That is the valley of death, and in Britain it is unusually deep.

Two things are quietly fixing the front of the pipeline

The first is the spinout problem. For years British universities took punishing equity stakes in the companies built on their research, often more than 30%, which deterred founders and scared investors off before a company had begun. The 2023 Tracey–Williamson review recommended a floor of 10%. It has largely worked: the average stake fell to 16% in 2024, a decade low, with Cambridge now at a median of 8.8%. Spinouts raised £2.6bn in 2024, up 38% on the year. More fundable companies are being born.

The second is the convergence Britain happens to be unusually good at. AI and biology are merging, and the institutions doing it best (DeepMind, Isomorphic Labs, the Sanger, the EBI's data infrastructure) sit within an hour of one another. The Isomorphic round, which drew in the new UK Sovereign AI Fund alongside Thrive, Alphabet and Temasek, is proof that this can be financed at scale without leaving.

The missing middle is not money alone

More fundable companies at the front, and more capital arriving in the middle, is the right direction. But capital does not close the gap on its own, because the gap is not only financial. A brilliant chief scientist is not, by default, a fundable chief executive. A landmark paper is not a proposition a generalist investor, or a pharma partner, can act on. The translation from science to company is its own discipline, and Britain has historically under-invested in it.

That translation is the work. It is sharpening the story until the asset is legible to people who don't read Nature; building the commercial case a Series B investor or a licensing partner needs; and lining up the right rooms (research partners, pharma, capital) so that by the time the introduction is made, the conversation is half done. The ladder in life sciences is its own climb: research collaboration, then a pharma partnership, then a licensing deal or the round that takes it to the clinic. Most companies stall not because the science failed, but because nobody did that work.

Equities have an exchange. Bonds have a tape. Real estate, bigger than both, has a phone call. Comparables are gathered by relationship. Leases live in PDFs. Valuations sit in a spreadsheet on one analyst's laptop. The largest store of wealth in the world is run, to a startling degree, on hand-keyed data and institutional memory.

That is finally changing, and the reason is AI. For the first time, the unstructured sprawl of property (leases, planning histories, EPCs, title, transaction records) can be read at scale and turned into something a machine can reason over. JLL's 2025 survey found 61% of institutional investors using AI for market analysis, up from 22% two years earlier. Underwriting that took weeks is being compressed into hours. Roughly $16.7bn flowed into real-estate, construction and infrastructure technology in 2025.

The tools have arrived. The wins haven't. Not yet. That gap is the opportunity.

Because here is the catch. By one industry count, 88% of investors have started piloting AI and only 5% have hit their objectives. The models are not the bottleneck. The bottleneck is trustworthy, unified data, and a buyer institutional enough to transact on the answer. Plenty of companies can demo a clever valuation. Very few can produce one a fund will underwrite a nine-figure cheque against.

Why Britain is the place to solve this

This is where the UK has a structural edge it has barely begun to use. HM Land Registry is digitising ownership and title. The EPC database already covers most of the building stock. And planning, historically the messiest data of all, is being opened up fast: the government's “Extract” tool, unveiled at London Tech Week in 2025, uses AI to turn decades of planning maps and documents into standardised, machine-readable data, feeding a national platform that already publishes open records from more than a hundred councils.

Put together, that is something the United States simply does not have: a national, increasingly open, standardised property data layer. It is the “Open Banking moment” for real estate, and Britain is ahead of it.

British companies are already mining the seam. LandTech turned Land Registry and planning data into a site-sourcing tool used by thousands of developers. Searchland did the same for off-market opportunities. Plentific built an operating layer for property maintenance at scale. The Pi Labs ecosystem in London has seeded dozens more. The raw material is here, and the first generation of tools is working.

The prize is the institutional layer

But the real prize sits one rung up. Blackstone, M&G, Legal & General, Aviva (the institutions that own Britain's built environment) are still underwriting in Excel off CoStar PDFs, while their own LPs demand net-zero and ESG data they cannot reliably produce. The platform that turns Britain's messy property data into decisions an institution will actually transact on, and report to a regulator on, is not a feature. It is a category.

Winning that category is not, in the end, a modelling problem. It is a selling problem. Institutional landlords move slowly, buy carefully and trust almost nothing on the first demo. The proptech customer ladder runs from the independent operator to the mid-market platform to the institutional buyer, and each rung is a harder, longer sale than the last. The founders who win will be the ones who can climb it: those who do the commercial work, not just the technical work.