Anduril is reportedly in talks to raise at a valuation near $100 billion, more than triple where it stood a year ago. Either defence tech is one of the fastest-repriced asset classes in years, or investors are paying today for procurement the Pentagon has not yet confirmed.
PitchBook counted a record $19.8 billion deployed across 262 defence-tech deals in Q1 2026, against $5.7 billion in Q1 2024. The moment Anduril’s valuation talks leaked, everyone asked the same thing: is defence tech a bubble?
This article runs three tests: the multiples maths, the bull and bear cases from the people writing the cheques, and whether startups can displace Northrop Grumman and Lockheed Martin. You’ll leave with a way to judge any defence-tech price, set inside where the whole defence money story is mapped.
Why are early-stage defence startups raising at 17-50x revenue multiples, and what does that maths assume?
Early-stage defence startups raise at 17-50x revenue because the denominator is contracted forward revenue, not trailing sales, and investors are pricing future procurement, export markets and margin expansion ahead of current cash flow. The multiple assumes defence-budget growth and clean prototype-to-production conversion. If either slips, the valuation unwinds.
Defence tech went from pariah to asset class in under a decade, and the pricing followed the money.
EV/Revenue sets enterprise value over revenue, but in defence the denominator is contracted forward revenue or annualised run-rate, not the trailing twelve months a SaaS business reports. A 17-50x forward multiple can be single-digit on realised revenue, while public software trades at single-digit to low-double-digit trailing multiples. When you are shown a 30x number, the first question is what the denominator is.
Finro’s Q2 2026 study of 168 private defence companies shows how uneven the stage picture is: medians of 13.3x at seed, 19.7x at Series A, 12.7x at Series B, 16.6x at Series C, 35.5x at Series D and 24.7x at late stage. Use the median, not the average. Series C has a 16.6x median but an 84.2x average, because a few outliers drag the mean upward.
Every multiple bundles four assumptions: budget growth holds, prototypes convert to production, export markets open, and margins expand at scale. Miss one and the next round is a down round. At the frothiest edge, deals have priced around 100x forward revenue, a level Anduril’s own CEO has flagged.
Helsing is the clearest worked example. The Munich firm raised $1.8 billion at an $18 billion valuation, the largest round ever completed by a European defence startup, at roughly 32x forward revenue for a company supplying HX-2 drones to Ukraine. The same forward-revenue maths, playing out in Europe.
Is defence tech actually in a bubble, and what do founders and investors say?
Most founders and investors say it’s not a simple bubble. Anduril’s Brian Schimpf concedes “a bit of a bubble”, while Founders Fund‘s Trae Stephens warns that too much late-stage capital is chasing too few deals. The evidence splits the same way: froth is worst in battlefield AI and drones, yet contracted demand and structural budget growth are real.
The bull case rests on rearmament, and it has two legs. Demand is the first: in one four-week snapshot the US fired roughly 850 Tomahawk missiles while rebuilding its stockpile at about 90 a year, and the Pentagon already has framework plans for 10,000 cruise missiles by 2029 and a five-year hypersonic order. The second leg is proof that a tech company can scale in defence: Palantir’s market value exceeded RTX, Lockheed Martin and Northrop Grumman individually going into its public offering, and the same analysis assumes a proposed US defence budget increase toward $1.5 trillion.
The bear case is the gap between private pricing and procurement reality. The NatSec100 found the top venture-backed defence firms raised $86.3 billion but won only $4.3 billion in FY2025 federal obligations, a private-to-public gap of $102 billion. The valley of death between a working prototype and production is where most of that money stalls. When you see a defence multiple quoted, this is the gap to check against.
PitchBook’s Ali Javaheri says it’s too early to say, flagging battlefield AI and drones as saturated, while Vantor’s Peter Wilczynski calls it complicated. That the founders and investors disagree with each other is the point.
Financing paths differ too. Venture rounds impose private-market discipline, dilution and staged checkpoints; SPAC listings skip that. Space-Eyes agreed to go public through a $638 million SPAC merger on a sliver of annual revenue.
Real contracted demand sits alongside speculative multiples, and Crunchbase‘s tools count nearly four dozen defence IPO candidates lining up to test those prices.
Anduril vs Northrop Grumman vs Lockheed Martin: can venture-backed startups genuinely displace the defence primes?
Not outright. Venture-backed startups can win new programmes and force the primes to reprice, but displacement depends on winning a Program of Record, which takes years rather than months. Anduril at roughly $100 billion is the valuation where the neo-prime claim stops being rhetoric and becomes a numbers question.
“Neo-prime” is the label for venture-backed, vertically integrated defence builders that own hardware, software and manufacturing. Anduril, Saronic and Shield AI fit it: Anduril in autonomy and missiles, Saronic in uncrewed maritime systems at $9.25 billion, Shield AI in autonomy at $12.7 billion. Palantir and SpaceX were the earlier proof cohort.
Startups bring software culture, iteration speed and cost per unit; the primes bring backlogs, manufacturing and congressional relationships. Anduril’s estimated value sits in the market-cap range of the primes, roughly $114 billion against $78-122 billion for Lockheed Martin and Northrop Grumman. Startups win single programmes; primes win decades of multi-programme scale. If you are weighing a neo-prime against an incumbent, speed and cost sit on one side and backlogs and relationships on the other.
The gate is the Program of Record, the funded budget line for sustained procurement. Procurement reform, from Other Transaction Authority to middle-tier acquisition, compresses the path, but an OTA funds experiments and early buys, not production. Shield AI’s multibillion-dollar valuation sits in pre-Program-of-Record limbo.
Anduril versus Helsing shows how different the markets are. Europe’s defence startups raised $8.7 billion in 2025, with Munich leading, but Helsing’s $18 billion is an earlier-stage, institutionally backed bet against Anduril’s roughly $100 billion.
By the end, the bubble question turns into a screening question. That leaves three questions to ask about any price you are shown.
First, check the denominator. Is the multiple on contracted forward revenue or realised sales? Second, check the sub-sector and stage. Froth concentrates in battlefield AI, drones and late-stage rounds; use the median, not the average. Third, check the Program of Record path. Is there a confirmed route to sustained procurement, or is the company sitting in pre-Program-of-Record limbo?
The verdict is froth in places, structural repricing overall, with the Program of Record as the dividing line. For most businesses, the practical consequence is what the boom is doing to hiring and supplier costs, and the wider arc sits in the rest of the defence tech story.
Frequently Asked Questions
What does Anduril actually build, and why is it worth so much?
Anduril builds autonomous systems for the battlefield: counter-drone and surveillance towers, uncrewed submarines, long-range munitions and the Lattice software platform that ties them together. Founded in 2017 by Palmer Luckey and a group of Palantir alumni, it has moved from border surveillance into major Pentagon and allied-navy programmes. The ~$100 billion valuation it is reportedly targeting prices in years of future production and export growth, not the revenue it books today.
Where does a private valuation like Anduril’s actually come from?
It is negotiated, not discovered. In private markets there is no ticker: a valuation is the price agreed between the company and its investors in a funding round, set by comparing forward revenue projections against public peers such as Lockheed Martin and Northrop Grumman. Because the denominator is contracted forward revenue, a $100 billion price can rest on revenue the company has not yet earned. Secondary-share sales and media reports then harden the number into conventional wisdom.
What is a down round, and why do defence tech founders fear them?
A down round is a new funding round priced below the valuation of the round before it, and it is the failure mode defence founders most fear. It dilutes early investors, resets option strike prices and signals to the Pentagon and suppliers that momentum has stalled. Founders Fund’s Trae Stephens warns that too much late-stage money chasing too few deals ends in musical chairs, exactly the risk when one of the assumptions baked into a 30x multiple slips.
What is the valley of death in defence technology?
The valley of death is the gap between winning a prototype contract and securing a Program of Record, the formal budget line that funds sustained production. Demonstrating a working system does not create an order, and most startups never make the crossing. Shield AI’s multibillion-dollar valuation sits in exactly this pre-Program-of-Record limbo, and the NatSec100 data, which pits billions raised against millions obligated, is the valley measured in dollars.
What is an OTA, and how does it change the procurement game?
Other Transaction Authority, or OTA, lets the Pentagon prototype and buy technology outside the slow, regulation-heavy Federal Acquisition Regulation. OTAs are the reason a startup can go from pitch to fielded prototype in months rather than years, and Anduril, Saronic and Shield AI have all used them to get in the door. But an OTA is not a Program of Record: it funds experiments and early buys, and the production budget line still has to be won separately.
How long does it take a startup to reach a Program of Record?
Reaching a Program of Record takes years, not months, even with procurement reform. The traditional US acquisition path can stretch past a decade, which is why the Pentagon now pushes Other Transaction Authority and middle-tier acquisition to compress it. Anduril has moved faster than almost anyone, but the gate still requires a funded budget line, repeated budget cycles and congressional support. Every 30x multiple prices in that outcome before it is confirmed.
Is the Pentagon a reliable customer once a startup wins a contract?
A contract award is not the same as revenue, and that distinction is the whole game. Winning a prototype deal or an OTA gives a startup credibility but no guaranteed production budget. The NatSec100 found that top venture-backed defence firms raised $86.3 billion but won only $4.3 billion in federal obligations in FY2025. Congress funds programmes year by year, so orders can be delayed, scaled back or cancelled. Reliable in the long run, but lumpy in practice.
Can ordinary investors buy shares in defence startups like Anduril?
Not directly. Anduril, Helsing, Shield AI and Saronic are all private, and their funding rounds are limited to venture funds and accredited investors who meet strict wealth and income tests. Retail investors can get exposure indirectly through venture capital funds or by waiting for a listing, the route Palantir took. Until one of these companies files to go public, there is no ticker to buy.
What happens to employee stock options if a defence tech valuation falls?
When a startup raises a down round, the strike price on new grants falls below what earlier employees paid, leaving existing options underwater and their paper wealth largely gone. That matters doubly in defence tech, where retention of engineers with clearances is critical and the hiring boom has already pushed salaries up. Companies respond by repricing options or issuing fresh grants, but the reprice is effectively an admission that the previous valuation was not achievable.
Would a defence budget cut pop the defence tech bubble?
Only if the cut were deep and sustained. Uninterrupted defence-budget growth is the first of four assumptions baked into every 17 to 50x multiple, and the whole bull case rests on rearmament driven by Ukraine and Iran. A major US or European drawdown would remove the contracted demand those prices are built on, and the frothiest names, battlefield AI and drone makers priced for flawless execution, would fall hardest. Slower growth dents valuations; a real cut breaks them.
Has any venture-backed defence company successfully gone public?
Yes, and it is the proof the bull case leans on. Palantir listed in 2020 and has at times carried a higher market capitalisation than Lockheed Martin, showing a technology company can scale inside the defence establishment. But it is still the exception rather than the rule: Anduril, Helsing, Shield AI and Saronic remain private, and SpaceX has never listed either. The path to public markets is proven, but narrow, and it runs through the Program of Record.
