Nobody values the composite.

A comparable here is a narrative anchor, not an input to an arithmetic.

We are pre-revenue with no reference customer. Revenue multiples require a denominator and ours is zero. Every company in the tables below is a narrative anchor or a stage anchor. There is no calculation in this memo that produces a valuation, and any slide that implies one will be taken apart in the room.

The two anchors that do carry weight

1. Stage arithmetic. PitchBook's median VC defence-tech valuation moved $42.8M (2024) → $146M (2025), against a $22.8M all-sector average. That is a median across all stages, so it frames a ceiling for the category rather than a price for a seed.

   PitchBook.

2. Gate arithmetic. The same research identified six things that unlocked large pre-revenue rounds, ranked by how often they were the unlock. That, not the comp set, is what prices this round.

The six gates, and where we stand on each

1. A named principal. The highest-impact item in the record, and the slowest to acquire.

   Not held

2. A deposit-bearing customer commitment. Deposits received — not a letter of intent.

   Not held

3. A slot in a named ecosystem. NSIC OT, a DIU prototype, a Carahsoft/NASPO line, a UL 2755 listing.

   Not held

4. A cost or supply-chain arbitrage argument. The BOM, the cost model and the tier ladder exist.

   Held

5. Government relationship velocity.

   Not held

6. A working article.

   In build

Four businesses, four multiple regimes.

Which comp set you choose determines the answer, and each choice costs something. That is the honest content of Hugh's own closing question about how anyone puts a value on this.

Stage comps: the most honest anchor we have.

Pre-revenue or early-revenue rounds in adjacent categories.

Category comps: evidence that serious money believes this.

Companies proving the category is investable — not our peer set.

Mature comps: ceiling and floor framing only.

The grown-up version of each of our four identities. Do not apply any of these multiples to us.

M&A precedents: realised prices beat marks.

This is the tier that should change how we think about the top of the range.

Identity 2, narrowed — priced as early-stage dual-use infrastructure, not as software.

The narrowing matters more than the choice. The broad version of identity 2 — "formally verified AI" — is occupied by Harmonic at $1.45B and Axiom at $1.6B, and priced at $170M when it actually sells. The narrow version is the one the folder has already identified as having no commercial occupant anywhere in the US:

So the pitch is: the insight is the asset, and the insight is a form-factor insight.

A proof harness moves the correctness burden off the model, which makes cheap open weights sufficient, which makes owned hardware the right form factor, which is the only form factor municipal finance law is built to fund. That is precisely the shape Heron Power raised $140M pre-revenue on — a supply-chain and form-factor arbitrage, not a benchmark and not a category claim.

What this choice costs us, stated honestly

1. It gives up the software multiple. We are asking to be valued on stage and insight, not on an ARR trajectory. That caps the headline number this year.

2. It puts us in a room with Harmonic and Axiom whether we like it or not, and our answer has to be that we do something narrower than they do rather than something better. That is a harder sentence to say and a more durable one.

3. It gives up the Armada comparison as a flattering anchor. Armada's $2B pre-money came with a procurement vehicle and a factory. Invoking it invites the question of which of those we have.

4. It makes the heat-reuse leg decorative. Identity 3 is financed by project finance, not venture equity, and Deep Green is ahead of us in the US with £200M behind it. It should stay in the deck as a differentiator and stay out of the valuation argument.

5. Identity 4 has no pre-revenue comparable at all. See the limits in §10.

Where no good comparable exists.

The folder's convention is to state limits first, so:

1. There is no pre-revenue comparable for identity 4 (govtech). Every govtech comp in the record — Tyler, Civica, Veriprajna's published $150k–$400k per jurisdiction — is a revenue business or a price list. Nothing prices a pre-revenue municipal software company. This is a real hole and it should be said rather than papered over.

2. There is no comparable for the composite. No company in the record is simultaneously an appliance maker, a verification-software business, a heat-reuse play and a govtech vendor. That is the finding, not a failure of the search — and it is the actual answer to Hugh's question about how anyone values this. The answer is that nobody values the composite; they value whichever identity you make them look at.

3. No formal-verification company has a realised exit above $170M (Jasper, 2014). OneSpin and Scaleway/Qarnot went undisclosed. If the exit thesis is a strategic sale of the verification asset, that is the record.

4. Helsing's Series A valuation was never disclosed. €100M is an amount, not a price. The same is true of Heron, SiPearl, Machina and Hadrian's seed. Most of our best stage comps have no published valuation at all — which is itself the reason this memo produces a range from gates rather than from numbers.

5. Deep Green's £200M is of undisclosed instrument. See the warning in §06.

6. Code Metal's $80M is a secondary source. Confirm against a DoD announcement before any external use.

Three comps, with one sentence of framing each.

Nothing here uses a retired claim.

No comp narrative here uses a retired claim.

Specifically, nothing in this memo says hallucination-free, guarantees correct AI outcomes, insurable, fully formally verified, zero sorry, more efficient than anyone, fast, or replaces data centres. Every figure carries a date and a source; every undisclosed valuation is marked as undisclosed rather than estimated; pre-money and post-money are distinguished wherever the source distinguishes them.

Ranked by cost, not by impact.

1. An ecosystem slot. NSIC is $500K–$3M, non-dilutive, fixed-price OT, explicitly for pre-production dual-use hardware, and companies already in mass production are ineligible — the programme wants pre-revenue teams. Rolling submissions, ≤20-slide deck. This is a process, not a recruitment, and it is the cheapest credibility in the federal system.

2. One deposit-bearing pilot. Not an LOI. The word investors now look for in a press release is deposits.

3. A named principal. The highest-impact and slowest item, and the one the research says is the only thing that consistently unlocked a large pre-revenue round.

4. A published benchmark, framed as coverage × accuracy, measured monthly.

5. Settle pre- versus post-money, and settle the instrument. Costs nothing and is currently unresolved.