Feasibility is not viability: three tests before your cap table closes

Two kinds of company, and the one nobody can fund

The paper's framing is uncomfortable and useful. There are two kinds of aerospace startup, and the difference is not technical sophistication, team strength or capital raised — it is the order in which the founder does the work.

Demand-first companies start with a customer who will pay for something, then engineer the minimum technical achievement that captures that spend. Feasibility-first companies start with a capability they want to build, then go looking for someone willing to pay for it. The second route, in the authors' words, produces "an extended valley of death, a series of bridge rounds, and a long tail of technically impressive ventures that never reach commercial deployment."

Both postures can be legitimate. The paper's two illustrations make the point better than a rule would. Varda is demand-first: it did not start with a reentry capsule and look for things to make in it. It started from a commercial and regulatory insight — that a microgravity-grown polymorph of an existing pharmaceutical may qualify as a new composition of matter, restarting the patent clock on an asset the drug company would otherwise lose to generic erosion — and designed the capsule around a specific molecule for a specific customer, with early revenue from government reentry-data work funding the path. AstroForge is openly the other thing: on its founder's own terms, "a research program funded as a company", and its investors know that is what they bought.

The failure mode is neither of those. It is the third case: a feasibility-first company presenting itself as demand-first. "Be one or the other," the paper says, because "the cap table will discover the truth eighteen months in and the company will not survive the discovery."

Which raises the practical question this article is about: how do you tell, from inside, which one you are running?

Test one: count your miracles

The miracle count is the number of independent breakthroughs your terminal economics depend on — not your next milestone, but the version of the company that makes money.

• One miracle is a venture bet.
• Two is a hard bet that needs aligned, patient capital on a 15–20 year horizon.
• Three or more is a research programme, and it should be funded as one — competitive research grants, agency programmes, university partnerships, sovereign deep-tech vehicles — "not as venture-backed companies on a fund-cycle clock."

Counting honestly is harder than it sounds, because every founder under-weights the miracles in their own plan: each one looks tractable when you examine it alone. The discipline is a single question. If the engineering works, does the company work? If yes — and the only open question is whether the team can build the thing — the count is one. If the technology working still leaves the customer to be invented, the regulation to be written and the supply chain to be built, each of those is its own miracle. Reusable launch was one miracle when SpaceX was raising in the early 2010s, because the customer and the regulatory pathway already existed. Asteroid mining is at least three: the mission, the refining process, and the existence of a market for the recovered material at the projected unit economics.

The count is not a judgement on ambition. It is a prescription for which money to raise. Here is that prescription translated into instruments a European team can actually apply to — the paper's own mapping is to US vehicles, so this table is ours:

One European wrinkle the paper cannot give you: our instruments are more forgiving about high miracle counts than US venture is, and that is a genuine structural advantage. A three-miracle European team has a real, well-funded home in collaborative research. What it does not have is a route to a venture round that survives diligence — so the mistake to avoid is not ambition, it is mislabelling.

Test two: name the person who signs the purchase order

The second diagnostic is the one most pitches fail. The paper's version of the deep-tech valley of death — the gap between TRL 6 and TRL 9 — is that it is a customer problem, not a capital problem. Capital would arrive in most categories if a customer existed at the relevant scale, willing to pay under a defined contract structure. It does not arrive because the customer-formation work has not been done, "and no amount of additional capital can substitute for that work."

That reframe changes what an instrument is for. The US mechanisms the paper credits — SBIR Phase III's non-competitive follow-on authority, Other Transaction Agreements, and the April 2026 Strategic Breakthrough Awards of up to $30M matched 1:1 conditional on a formal Program of Record commitment — all address customer formation rather than capital. Winners "are exiting the valley of death by acquiring an anchor customer with the institutional authority to actually field the product." The commercial analogue is the structured anchor-customer agreement, where a large buyer commits to a defined volume conditional on technology milestones: NASA's COTS programme for cargo was the canonical case, CLPS for lunar payloads a more recent one, and the xEVAS spacesuit programme — where NASA buys the service, not the hardware, under agreements with a combined $3.5B ceiling — the clearest current example. "These are not subsidies."

Europe has this category of instrument, and European founders systematically under-use it because it looks like procurement rather than funding:

• ESA procurement is the closest thing we have to an anchor-customer machine — a contract with a buyer that has the authority and the budget to field what you build, and, unlike EU grants, one where the contractor keeps the IP it generates.
• EUSPA and EU framework contracts, and pre-commercial procurement under Horizon Europe, which exist precisely to create a first institutional buyer.
• National agency contracts — the fastest route to a named customer for most member-state teams, and the one with the shortest decision chain.
• Defence-adjacent lines under EDF and EUDIS, where the buyer is structurally motivated. We covered that shift in Europe's defence-driven space security shift.

What Europe does not have is an equivalent of SBIR Phase III's non-competitive follow-on authority — the mechanism that lets a government customer buy again without re-competing. Anyone who tells you the EIC Accelerator plays that role is wrong: it supplies capital, not a committed customer, and the paper's whole argument is that these are different problems.

Then the uncomfortable question, which is the demand-first test in practice: which actual person, at which actual organisation, has the willingness to write an actual purchase order, on what specific timeline, for what specific deliverable? The paper's observation is that most in-space manufacturing, on-orbit servicing and frontier-aerospace pitches cannot answer it for a single named customer. Founders who do that work before the seed round are doing demand-first development. Founders who plan to do it after are "funding feasibility-first development with capital raised against a demand-first story."

The corollary is the sharpest line in the chapter, and it applies just as well to a Horizon Europe impact section as to a pitch deck: "A founder who does not interact with the demand side in any operationally meaningful way should not construct TAM slides."

Test three: three clocks, and the one nobody prices

The conventional story about aerospace startup failure is that venture capital imposed a timeline the engineering could not support. The paper argues this is misleading. Most aerospace startups that failed in the last decade did not fail on speed — "they ran out of one of three things before the engineering converged: capital, customer demand, or insurance capacity."

Their examples: Astra reached orbit and then could not raise at a valuation that let it keep operating. Virgin Orbit had a working vehicle and a customer pipeline and ran out of capital twelve months before the next contract cycle. Boeing's Starliner is a procurement mismatch — a fixed-price contract against a development arc that needed the cost-plus pacing the customer had abandoned. None is a speed-versus-safety failure.

So: three clocks, each with its own failure mode.

1. The capital clock — how long funding lasts and how long the next round takes to raise.
2. The customer clock — from a buyer's decision to buy to the point they can actually pay under a contract, gated by budget cycles, certification gates and their own procurement process.
3. The insurance clock — from first flight to the existence of a market that will insure the system in service. Without that market the customer cannot field the system at scale, even when the engineering works.

The clocks are only loosely linked, which is the entire point: "A company can have eighteen months of cash, a buyer who commits in twenty-four months, and an insurer who won't touch the category for forty-eight." Solving only the capital clock moves the problem rather than fixing it.

The third clock is the one that never appears in a pitch deck and still ends companies. The deepest market for new space risk carries total annual capacity in the low single-digit billions of dollars — "about the size of one major constellation's launch programme for a year" — and genuinely new categories, including in-space servicing, often have no insurance market at all in their first flight years, because there is no track record to price against. The failure pattern is specific and quiet: the engineering works, the customer is willing in principle, and deployment stalls because underwriting is two or three years behind.

Part of the fix is engineering — design choices that produce data an underwriter can price, backup designs that price well, telemetry an underwriter can watch in real time. Part is policy: state indemnification standing in while the market builds confidence, which in Europe means the national space-activity acts and their liability caps rather than a single federal regime. And the operational recommendation is worth copying verbatim: bring underwriters in as design partners from the first architecture review, not as vendors after the vehicle exists.

For a European team the customer clock deserves one specific warning. ESA and EU procurement cadences are slower than the US mechanisms the paper compares against, and slower again if you are assembling a consortium. Our own read of realistic timelines — six months to two years from tender preparation to contract revenue for a large call — is in ESA vs. EU procurement. If your capital clock is shorter than that, you do not have an engineering problem or a sales problem. You have a capital-structure problem, and the paper is right that no amount of engineering speed fixes it.

What the survivors did differently

The pattern the authors draw across all of this is the paper's central thesis, and it is not a generic startup platitude: in aerospace, the decisions that determine whether a company survives are made early, in choices that do not feel architectural at the time. Capital structure is architectural. Regulatory exposure is architectural. Customer-cycle alignment is architectural. Supplier and workforce strategy is architectural.

The reason it is sector-specific is timing constants. Procurement cycles, certification gates, standards-body output cadences and insurance-market formation are all long enough that the gap between an architectural decision and a reactive one is the difference between a company that ships and one that does not. "The conventional founder discourse treats year one as operational and year three as strategic. The successful aerospace founders we interviewed treat year one as strategic and year three as operational, and the inversion is consequential."

They are also careful about what the companies that failed got wrong: in most cases they "made the right decisions on the right problems but put them at the wrong time." That is a more forgiving and more actionable diagnosis than "they were badly run".

The diagnostic, in the order to run it

1. Count the miracles, out loud, with someone who will argue. If the engineering works, does the company work? Write the number down before you write the deck.
2. Check the cap table against the count. One miracle can carry generalist venture money. Two needs patient capital plus an institutional anchor. Three or more needs a research instrument, and pretending otherwise is the failure the paper describes.
3. Name the buyer. A person, an organisation, a timeline, a deliverable. If you cannot, you are feasibility-first regardless of how the pitch is framed — and the honest move is to spend the next quarter fixing that rather than raising against it.
4. Find the anchor-customer route. ESA procurement, EUSPA or EU framework contracts, pre-commercial procurement, a national agency contract, a defence-adjacent line. Grants fund TRL progression; an anchor customer is what closes TRL 6–9.
5. Time all three clocks on one sheet of paper. Months of runway, months to the customer's first payment, months to an insurance market. If the first is shorter than the second, fix the capital structure. If the third has no answer, name the underwriter you will talk to this quarter.
6. Then design. The point of running these before the cap table closes is that afterwards, most of the answers stop being yours to choose.

Two of the three tests are about demand rather than technology, which is the paper's whole argument compressed. If you want the wider context — the four drivers, the servicing regimes, and the five predictions with the authors' own counter-arguments — that is in our European read of the full paper. And if the regulatory side is where your architecture is least decided, start with regulation as a design constraint.

FAQ

Sources