> ## Content Index
> Fetch the complete content index at: https://www.provenance-vc.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# The Fault-Line Premium
- URL: https://www.provenance-vc.com/the-fault-line-premium/
- Published: 2026-11-21T11:11:02.000Z
- Updated: 2026-11-22T11:03:11.000Z
- Author: Clement Stoffel
- Tags: DEFENSE, AI

Farsight Vision’s Series A will put a price on Europe’s defence growth-capital gap. Here’s how the round should be priced.

*How do you price a company whose value moves with the strategic ground beneath it — and do it at Series A, the round that sets the price of the next five years?*

![](https://storage.ghost.io/c/53/d9/53d98187-e75b-47e1-8a6d-fc7217a580ee/content/images/2026/11/ground-truth-fault-line-premium-drawn-1.svg)

---

## **Earned Up Front** 

Champagne is honest about its lifespan — one pop, then two hours of drinking it flat because the bottle’s open. Venture capital runs on the same schedule — a quick hit at the signing, then a long stretch of momentum nobody wants to call momentum. I’m aware a champagne metaphor at 2am is not the product of a rested mind. But I’m told the aftertaste is worth it. So on we go.

Let me begin with a confession, an awkward one for someone about to write about a company valuation - the last thing I noticed about Farsight Vision was the technology. The first was the hype, all of it about the defence market and none of it about the company.

Not that the technology is easy to overlook. The Farsight Vision team started work in June 2023\. By February 2024, eight months after the first request, Farsight Vision’s software was on the front line in Ukraine. It uses AI to turn drone footage into 2D and 3D terrain twins, giving near real-time intelligence even where GPS is jammed, and soldiers were using it to produce high-resolution maps in as little as seven minutes. I find that faintly insulting, given that it takes most of us longer to choose something to watch on Netflix. Today, by the company’s own count, it is in use across more than 100 Ukrainian military and security units. That is not a pilot. That is a dependency. And in February the company closed a €7.2m seed round led by Axon and SmartCap, Estonia’s state-backed defence fund.

So far, so heroic. Now the awkward part, which is my part.

The war has broken something structural in European defence, and it will not be put back. For a generation, Europe’s defence capability came from a few giant defence contractors on decade-long cycles, and Ukraine was a country the continent supported from a distance. That arrangement is gone. Ukraine is still the proving ground, but more and more it’s also the supplier, and Europe has been quick to back the companies coming out of it. What it has not learned is how to keep backing them once the story stops doing the heavy lifting and the business has to stand on its own. Farsight’s Vision Series A is that moment. And whatever number lands on it won’t stay put - every founder and every fund that follows will read it as the going rate for believing a seed or Series A company has a future. No pressure.

That’s really what this piece is about: what should that belief cost? My short answer is that the premium you pay for a company like Farsight Vision buys three things, not one, and each runs down on its own clock. Most of the mispricing I’ve seen comes from squeezing all three into one number. So let me walk you through them, using Farsight Vision as the worked example, before a term sheet does the walking for me.

---

## **The Quake Is Already in the Price**

At this point, against my own expectations and at least one promise I made in school, I have to bring in geology. The rocks are relevant. I'm as surprised as you are. Here's the thing. Geologists have a habit of mind worth stealing. When the ground moves, they do not ask what made the fault. The fault has been there for millions of years. They ask what, on that particular morning, finally let go.

Venture capital rarely extends itself the same courtesy. When a sector cracks open and money rushes in, the explanation is reflexive – a gap appeared, founders appeared, cheques appeared. I'm sceptical about this. Two things arriving together tells you about the clock, not the cause. A fault does not manufacture pressure, it releases pressure that built up out of sight, and what the rupture changes is not the problem but its visibility. A difficulty that once took forty slides to explain can suddenly be put in a sentence. So capital does not flow to problems. It flows to problems that are easy to describe, on the headline’s timetable rather than the solution’s. That’s how a sector goes from under-capitalised to over-crowded in eighteen months - a dozen companies building the same product, and the problem that drew them in. 

Take Farsight Vision, it closed a €7.2 million seed round in February. However, it is swimming in a sector where venture money nearly doubled in a single year, to $49 billion in 2025, and the number of firms writing cheques rose by two-fifths. Farsight’s own chief executive, Viktoriia Yaremchuk, has said the industry needs to get over its start-up hype.

None of this is an argument against moving quickly when the ground shifts. It is an argument about where to look. The quake is the least informative moment in the cycle. Everyone can see it. That is precisely why it is already in the price.

---

## **Follow the Pressure**

So where do you look instead? Not at the break. When a fault slips, the stress it was carrying doesn’t vanish. It moves. It’s handed down the line to the segments beside it, the ones that stayed quiet through all of this, and each now holds more than it did before the shaking started. The ground that broke has let its strain go. The ground beside it has just been handed one. The rupture is the loud part, it is also the part with the least left to give.

European defence tech has had its rupture, and the numbers are loud. By Dealroom and the NATO Innovation Fund’s count, European defence, security and resilience startups raised $8.7 billion in 2025: up 55% on the year, 13% of all European venture, more than three times the sector’s share in 2020\. That is the quake. It is on every slide and in every memo, and it is already in the price. The question worth asking is where the ground actually moved, and the answer is: in one place.

Break the total apart by stage and it stops being one market. Late-stage funding tripled, to $4.7 billion. Everything beneath it raised roughly $4 billion in 2024 and roughly $4 billion again in 2025\. Flat. Every dollar of the increase went to companies that had already won. For a founder trying to turn a seed round into a Series A, the most celebrated year in the sector’s history changed nothing.

This year has sharpened the split rather than closed it. In July, Helsing’s $1.8 billion Series E and Quantum Systems’ $1.2 billion Series D together raised twice what the entire defence-only sector raised in all of 2025\. At the other end of the market, Fox & Lion, a deal tracker, has seed investment in the sector at $36 million so far this year, down from $50 million and on course to finish a fifth below 2025\. The top keeps breaking its own records. The bottom keeps shrinking. That is the growth-capital gap in this piece’s subtitle, and it is why Farsight’s Series A matters to people who have never heard of Farsight.

![](https://storage.ghost.io/c/53/d9/53d98187-e75b-47e1-8a6d-fc7217a580ee/content/images/2026/11/funding-drawn-transparent.gif)

Europe has now proved two things. It can produce companies with a working prototype and a user under fire, and it can produce billion-dollar rounds for the handful that break out. What it has not proved is that there is money for the stage in between: the company that has already been used in the field and has not yet been paid for it, looking for the round that carries it from one to the other.

Farsight Vision is that company, point for point.

---

## **Naming the Premium** 

Time for some theory. I know, venture capital would rather not. Fair enough, except that a rupture is the one moment the old patterns stop working. So here is mine, on how to value a start-up when they do.

The **fault-line premium** refers to the additional value that investors assign to a company by virtue of its proximity to a structural rupture. A structural rupture, in the sense used here, is a sudden and large-scale discontinuity in the economic or social order that durably alters the conditions under which companies operate and, in doing so, redistributes value among them. Properly understood, the premium is not a unitary quantity but the sum of three distinct components, which differ materially along two dimensions: persistence, that is, how long each endures; and transferability, that is, how far each can be carried beyond the rupture that produced it. A substantial proportion of the mispricing observed across sectors is attributable to the failure to distinguish among these components—namely, to the practice of pricing a heterogeneous premium as though it were homogeneous.

![](https://storage.ghost.io/c/53/d9/53d98187-e75b-47e1-8a6d-fc7217a580ee/content/images/2026/11/premium-drawn-transparent.gif)

### **Ground truth**

The first component of the fault-line premium is ground truth, and the term has to be defined with care. In its narrow sense, ground truth is the empirical record a firm accumulated while its technology was in operational use - what that technology registered, what it generated and what it misjudged, during deployment. On that reading the sceptic is right. A record is bound to the circumstances of its collection and says nothing about conditions elsewhere, and if ground truth meant only the record, the premium attached to it would be difficult to justify.

The record, however, is a by-product. What the firm acquired was the discipline through which that record was assembled and continuously corrected, and it is the discipline, not its residue, that retains value beyond the theatre. The distinction is between a stock of observations, which loses relevance as conditions change, and a demonstrated method for restoring a model’s correspondence with its environment when they do.

Three features make that competence rare. The first is the cost of error - every output was judged by users whose survival depended on its accuracy, and their verdicts fed directly into the technology’s development. No trial can impose that standard, because in a trial a mistake has no victim. The second is the instability of the environment: the model was refined against an adversary actively working to defeat it, so each deployment presented conditions not previously encountered. Success under those circumstances demonstrates adaptability to novelty, and adaptability, unlike performance on a fixed test, is the quality that transfers. The third is the documentation of correction: because every failure and repair was logged, the firm can show how its capability improved rather than merely assert it. For a buyer, that is the difference between a claim of reliability, which cannot be verified, and a history of reliability, which can.

The economic logic is that of learning-by-doing, extended to an adversarial setting. A system trained on an instrumented range learns the regularities of that range, and it can, because the range is stationary and cooperates with measurement. A system trained under adversarially induced distribution shift learns something else - its own failure modes. It is this robustness, not any dataset, that generalises across theatres. In the vocabulary of the resource-based view of the firm, the resulting asset is non-replicable. A peacetime entrant can buy data, compute and personnel; it cannot buy an adversary, and it cannot compress the time over which iteration occurred. The logged corrections perform a second and distinct function. Defence procurement is marked by severe information asymmetry, since a vendor’s performance claims are costly to verify before contract. A firm whose corrections are recorded offers the buyer a history to inspect rather than a representation to trust. That lowers the cost of verification, and it supports a premium independent of the robustness itself.

Ground truth is therefore the most durable of the three components. It is a function of time elapsed under adversarial pressure, which no later entrant can accelerate, and, unlike the other two, it does not depreciate when the conflict that produced it ends.

### **Legibility**

Where ground truth concerns what a firm learned, legibility concerns whether anyone else can read it. A competence has to exist, but that is not enough; it also has to be visible to those who would pay for it. Defence technology is, in the economist’s sense, a credence good: its quality cannot be assessed before purchase, and often not after, because the conditions under which it would fail are precisely the ones a buyer hopes never to meet. In such a market, the firm that can be seen to work commands a premium over the firm that merely works.

A war on a fault line produces legibility in a form no marketing can replicate. Allied governments, their intelligence services and their procurement agencies watch the front continuously, not out of curiosity but because their own planning depends on what they see. A technology deployed there is therefore evaluated by parties who are neither vendor nor customer, who have no commercial stake in the verdict, and whose assessments carry institutional weight. That is third-party validation of a kind the ordinary procurement process cannot manufacture, because that process rests on tests the vendor helped to design. The legibility premium is thus a premium for information. What it buys is not a superior capability as such but the resolution of uncertainty about capability, and where that uncertainty is otherwise irreducible, resolution is worth a great deal.

The premium is real but non-proprietary. Its inputs are not scarce. An active front, a count of users, a recognisable investor: every firm operating in the same region has access to all three, so the signal describes a category position rather than a firm-specific capability. In signalling terms, a signal that every member of a category can emit at low cost cannot separate; it conveys membership of the category and nothing about distinction within it. The rent attaches to the narrative, and in an expanding sector narratives are competed away.

Two consequences follow for durability. First, the component’s half-life is bounded by the time a competitor needs to formulate an equivalent claim: not the years of adversarial iteration that ground truth requires, but the weeks it takes to assemble a comparable pitch. Second, legibility is demand-driven rather than supply-constrained. It inflates when capital is abundant and contracts when capital withdraws, so its value tracks the investment cycle rather than anything the firm has built. It is therefore the component for which investors most frequently overpay. The error is a category confusion: a sentence that is easy to understand is taken as evidence of a capability that is hard to replicate, when the ease of understanding is precisely what makes the sentence easy to replicate. A premium for legibility is warranted, but it should be priced as a transient advantage in access to capital, not as a durable claim on the firm’s value.

### **Urgency**

The term needs careful definition, because it is easily mistaken for haste, and the two could hardly differ more in what they answer to. Urgency, as used here, is the compression of the interval between the appearance of a problem and the deployment of its remedy, under conditions in which the length of that interval carries a measurable cost. Haste shortens the interval without reference to that cost, and often at the expense of the remedy itself; it is speed for its own sake. Urgency shortens the interval because of the cost, and in proportion to it; it is speed as a response to a price. To a peacetime organisation, delay is an inconvenience: a slipped date, a line in a status report. To the fault-line firm, it was measured in lives.

Urgency also differs from the other two components in how it is lost. Ground truth and legibility decay by degrees, and the decay can be arrested; a firm that sees either slipping can act. Urgency expires. It exists as a window, a stretch of time in which a problem is acute, a remedy is possible and someone will pay to close the gap between them, and the window has a term. Inside that term, urgency is either converted into contracted revenue or it is gone. There is no partial conversion, and there is no reopening. A closed window cannot be forced back open by harder work, because the window, not the work, was the source of the value.

---

## **Farsight Vision - Standing Firm Through a Series A Quake** 

Now for the part that matters. I confessed at the outset that I noticed the calendar before I noticed the technology. Time to show my working. What follows runs Farsight Vision through the fault-line premium one component at a time and lets the three clocks do the talking. They do, and with rather more conviction than I expected.

Start with what won’t work. This round cannot be priced the way a Series A is normally priced. The standard method is familiar to the point of reflex - gather a handful of revenue-based comparables, extract a multiple, apply it. The method isn’t wrong - it’s inapplicable. It rests on the assumption that the comparables and the target create value by the same mechanism, and for Farsight Vision that assumption collapses at the first step. The comparables measure revenue. Farsight’s value, today, does not live in its revenue. Apply a multiple to the wrong quantity and you get a number that looks rigorous and means nothing.

What the situation calls for instead is decomposition. The fault-line premium is not one quantity but three, and each ages differently. There is a durable core: the part that would still be standing if every ounce of surrounding enthusiasm evaporated tomorrow. There is a decaying signal: the part that is accurate today and a little less accurate with every quarter the market is allowed to catch up. And there is a time-limited option: the part worth a great deal inside a defined window and almost nothing outside it. Price the three separately, each on its own logic and on present evidence, and only then put them back together.

The chart below draws that exercise on a single calendar, and it repays a minute’s study before we go on. Each line is one component of the premium, plotted by what it is worth to an investor. The vertical rule marks today. Everything to its left has been proved by use. The shaded ground to its right is still to be proved, and that shaded ground is what the Series A is buying. The numbered circles are reported facts. The curves are my reading of what those facts add up to, and the rest of this piece is the case for drawing them that way.

![](https://storage.ghost.io/c/53/d9/53d98187-e75b-47e1-8a6d-fc7217a580ee/content/images/2026/11/calendar-drawn-transparent-1.gif)

**Ground truth, or what you are actually buying**

Start with the navy line, the only one that never turns down.

It is tempting to say that you are buying Farsight’s data. You are not. Data from a front line goes stale as fast as the line moves. What transfers is rarer: a demonstrated ability to keep a model faithful to an environment that is hostile and never stops changing. Teece, Pisano and Shuen (1997) would call it a dynamic capability, something quite different from the stock of resources Barney (1991) describes.

The evidence runs along that navy line. Marker 1 is where it began: a first device at the front in November 2023, for units with no GPS and data links days out of date. What followed was a run of corrections. A design that depended on knowing the drone’s coordinates was rebuilt to work without them, and by Marker 2, in January 2025, the team had its own models and its own maths in place of GPS and ground control points. An edge device was retired when 6K imagery drove it past its thermal limits. Processing migrated to servers and handsets because the field demanded it, not because a roadmap did. None of this was strategy in the boardroom sense. Every change was a correction, and every correction was forced.

Every correction was also logged, and the log matters more than it looks. It does precisely what Akerlof’s (1970) work on asymmetric information says a market needs - it turns a vendor’s unverifiable claim into a history a buyer can inspect, and so lowers the cost of checking. A peacetime rival can hire the same engineers and rent the same compute. It cannot hire the adversary, and it cannot hurry the years in which the learning by doing took place (Arrow, 1962).

So ground truth belongs at the centre of the valuation, not in a footnote to it. The chart’s margin puts it in a line: proved, compounds with every flight, pay at signing. Even the dashed branch, which shows what a ceasefire would do, only levels off. Fresh data slows and the capability stops compounding, but nothing already learned is unlearned. Of the three clocks, this is the only one running forwards.

Its one weakness is that it commands a premium only from a buyer who can read it. That is where the grey line comes in.

**Legibility, or the sentence anyone can write**

The grey line deserves suspicion. It climbs steeply, peaks in the summer, and has been sliding ever since.

The pitch that will carry this round can be said in one breath and checked in an afternoon: a product at the front within six months of founding, more than 6,000 active users, a suite spanning the ISR stack from feed ingestion to GPS-denied navigation, and a seed round co-led by the maker of the Taser and a sovereign defence fund. It is a very good sentence. The problem is how many companies can write one just like it.

In Spence’s (1973) terms, most of it is a pooling signal. Brave1 counts around 2,500 Ukrainian defence-technology firms, and a fair share of them could produce something close at almost no cost. A signal like that admits you to the category; it does not distinguish you within it. The chart marks the two moments that matter. At marker 3, in February 2026, Axon and SmartCap co-led a €7.2m seed and, in effect, validated the sentence. At marker 4, five months later, Helsing and Quantum Systems closed their mega-rounds and the sentence became the sector’s. The grey line peaks at exactly that point. A signal is worth most just before everyone else learns to send it.

Three clauses do more work, and they deserve real weight. Axon’s co-lead is a costly signal - a manufacturer with its own product line spends real diligence, and stakes real reputation, when it backs a seed-stage company. SmartCap puts sovereign capital behind a Ukrainian-founded firm. And the Delta integration means allied planners have seen Farsight’s Vision output at work, not merely on a slide.

The remainder should be priced for what it is - a temporary fall in the cost of capital that another year of defence-tech fundraising will compete away. The classic error, one Farsight’s Vision own chief executive has warned against in urging the industry past its start-up hype, is to mistake how easily a story is told for how hard a business is to copy. By the right-hand edge of the chart the grey line has all but reached the floor, and the margin gives the verdict: cost of entry, not return. Legibility may decide whether an investor gets an allocation, and on what terms, but it should not move the headline number.

**Urgency, or the part with a clock on it**

Now the red line - the highest on the chart, and the one an investor can least afford to take at face value.

Its own caption explains both its height and the trouble with it - used daily, cannot be invoiced in Ukraine. Six months from founding to the front line was urgency already converted into use. The Series A is the attempt to convert use into contracted revenue while the window stays open. While the problem is acute, the remedy exists and somebody is willing to pay. That is a real option with an expiry date (Dixit and Pindyck, 1994), and it carries an uncomfortable dependency. The window is held open by the war.

A lasting ceasefire would close it from both sides. It would remove the acute demand that gives the product its value, and it would end the live conditions under which people who are neither vendor nor customer can watch it perform. The absence of a lasting ceasefire is therefore, however uneasily, one of the conditions on which this valuation rests. It is not a comfortable sentence to write, but I would rather write it than leave the reader to find it.

That is also why the money must be spent inside the window: on contracts beyond the war zone, where use can be invoiced, and on turning the daily use of five Ukrainian services into a documented record, so that a buyer in Warsaw or Copenhagen can inspect a history rather than take a vendor’s word for it. Copenhagen is not hypothetical. Marker 5, in August, is the Danish trial, Farsight’s first mark outside the war, which arrived in the same month as an EIC grant for its geolocation work.

Then follow the red line past the dotted rule at mid-2027, where the chart turns dramatic. That rule is the test: Farsight’s first contract priced as software. If it lands, the line carries on along the top as recurring revenue. If it does not, the line drops off a cliff and finishes in the bottom corner beside the faded grey one: no contract, worth nothing. The failed option and the spent signal end up in the same place. Hence the margin’s instruction: milestone it. The deal should be staged in the way Kaplan and Strömberg (2003) find throughout venture contracts. Urgency belongs in milestones, not on the balance sheet.

**Putting it back together**

Read the chart’s right-hand margin from top to bottom and you have the term sheet in miniature: milestone it, pay at signing, cost of entry. Ground truth is the foundation, priced for what transfers and paid for on the day. Legibility is a premium the investor should expect to lose, and should therefore concede in terms rather than in price. Urgency is an option, realised only if conversion happens within the term, and released against milestones.

Priced this way, and on present conditions, the figure comes out markedly above what a conventional Series A model would produce. The reason is the one I began with. The conventional model reads the absence of peacetime revenue as a deficiency. In a company built under fire, it is simply the one piece of evidence that could not yet have been collected. The whole left half of the chart says as much - proved by use, never by invoice.

The conditions are not small print on that result. They are the result. Recurring revenue, contracts beyond the war zone, field performance that remains verifiable, a record that grows easier for outsiders to read: each is a test of portability, and portability is what the shaded half of the chart is asking the round to buy.

---

## **The Last Word**

It’s convenient how everything hangs together by the time I’ve finished explaining it, and how that’s also roughly when the champagne goes flat. The audience nods, and the absence of follow-up questions passes for agreement. So let me bring a little mess back in, before the conclusion snaps shut like a trap.

But who cares about any of that? What matters is catching the market’s growth and cashing in, right? Or, the three components don’t hold up. There’s nothing measurable underneath them, no multiple to point to — it’s subjective all the way down.

Fair. It’s not crazy to say I’ve just made valuing a start-up messier, not easier. But a messy argument that holds together will get you closer to a real answer than a tidy one that lines up neatly and tells you nothing.

Farsight Vision’s Series A, when it comes, will look ambitious, and that is consistent with a market organised around structural rupture rather than evidence against it. The case for it is conditional, but the conditions are the argument, not hedges around it. Think of them as a bingo card with four squares: recurring revenue, contracts beyond the war zone, field results that can still be verified, legibility that keeps improving. I’ll be watching that round more closely than I’d care to admit, pen in hand, to see how many get crossed off.

Written at 4:37am, on something that tastes like champagne but is almost certainly just exhaustion. My eyes are filing a formal complaint, so I’ll stop here.

Thanks for staying to the end of a long and, let’s be honest, fairly dense piece. If your head aches a little, mine does too. Call it a shared memory. The next one will be shorter, livelier and easier on the brain.

[Farsight Vision | Decision Intelligence & Situational Awareness for Humans and RobotsFSV Platform delivers AI-powered decision intelligence for humans and robots. Enhance situational awareness, simulate operations, and accelerate data-driven decisions.![](https://storage.ghost.io/c/53/d9/53d98187-e75b-47e1-8a6d-fc7217a580ee/content/images/icon/favicon-7efa78cb-80a0-4322-9d58-ed2ba8fbde25.png)Farsight Vision![](https://storage.ghost.io/c/53/d9/53d98187-e75b-47e1-8a6d-fc7217a580ee/content/images/thumbnail/og-4cb9d15f-32e9-4f2f-916f-b179084136cf.png)](https://fsv.global/en?ref=provenance-vc.com)

Farsight Vision Website