Every discounted cash flow has two halves. The first is the explicit forecast: five or ten years of projected cash flows, argued line by line. The second is the terminal value, a single figure standing in for everything that happens afterwards, usually produced by an exit multiple or a perpetuity growth formula and usually arrived at in the last ten minutes before the deck is printed.
In a mature business the split might be sixty-forty. In a frontier business it is closer to five-ninety-five, and in a pre-revenue one it is a rounding error against everything. The strange thing is how little the process changes to reflect that. Founders and investors spend their meetings arguing about the five per cent.
I. A number nobody defends
Ask most people what a launch company or a frontier model lab is worth and you will get a number. Ask what it is a function of, and the honest answer is: a belief about the world in 2038, held with more confidence than anyone would admit under questioning.
This is not a criticism. It is the correct answer for this asset class. If you are underwriting a company whose product does not exist at scale yet, whose market is partly constructed by its own success, and whose cost curve is the entire investment thesis, then the near-term numbers are a progress report and nothing more.
The problem is that the machinery of a raise — the deck, the model, the diligence list, the committee memo — is inherited from businesses where the explicit period genuinely mattered. So the conversation goes where the machinery points it, which is at the least informative part of the picture.
A founder gets six weeks of questions about next year's pipeline, and roughly eleven minutes on the assumption that actually determines whether the investment works.
II. Where the value actually sits
Take a company building inference capacity. The explicit period is four years of deploying hardware, signing contracts, and burning capital. Almost every dollar of value in the investment sits past that: in what utilisation settles at, what the hardware refresh cycle costs, whether the contracted demand renews at a price that clears the depreciation, and whether the position is defensible once three competitors have built the same thing.
Or a launch provider. The explicit period is development, qualification and early flights — a period in which the correct financial forecast is a large negative number with error bars. All the value is in cadence at maturity, price per kilogram at that cadence, and the share of a market that mostly does not exist yet.
In both cases the terminal value is not a residual. It is the investment. And it resolves, always, into a small number of physical or economic assumptions that can be named and argued about directly:
- The useful life of a depreciating asset.
- The cost curve, and how much of it is learning versus scale.
- Whether demand at maturity is contracted, procured, or hoped for.
- What the position looks like when the third competitor has the same capability.
Four sentences. That is usually the whole thing. Everything else in the model is the arithmetic that follows.
III. Two failures of translation
Given that, there are exactly two ways to fail at the raise, and we see both constantly.
The first is refusing to state the terminal at all. The deck describes the technology in loving detail, presents a market-size chart, and stops. The investor is left to construct the terminal argument themselves, in their own head, using their own assumptions — which will be more conservative than yours and which you will never get to see or rebut. You have outsourced the most important argument in your company to a stranger with a scheduling problem.
The second is stating it as an assertion. A slide claiming a twenty-billion-dollar market and a plausible share of it, with no mechanism connecting today's company to that outcome. This reads as unserious, and it is worse than silence, because it tells a disciplined investor that you have not done the work they were hoping you had done.
The correct move is neither. It is to state the terminal explicitly, decompose it into the handful of assumptions that carry it, show what each is worth, and then show your evidence for each — including where the evidence is thin. A frontier investor is not looking for certainty. They are looking for a founder who knows precisely which uncertainty they are being paid to take.
IV. What replaces the forecast
If the explicit period is not where the value is, the model should not be built as though it is. In practice this means three changes.
Build the mature-state economics first. Start at the end. What does one unit of this business look like when it works — one satellite, one deployed cluster, one manufacturing line — at steady state? Get that unit credible, defensible and boring, and then work backwards to how many of them exist and when.
Replace projections with milestones. The explicit period is not a revenue forecast; it is a sequence of things that must be proved, each of which collapses a specific uncertainty in the terminal. Present it that way. The next round is priced on which uncertainties you have retired, not on which quarter you beat.
Show the downside without flinching. Give the case where the cost curve stalls, where the anchor customer does not renew, where the depreciation schedule is two years shorter than you assumed. Founders avoid this, believing it weakens the pitch. It does the opposite: a downside case that has clearly been constructed by someone who understands their own business is the single strongest credibility signal in a raise. The investor is going to build it anyway, and worse.
V. Consequences for the raise
Once you accept that the terminal is the company, several practical things follow.
The investor universe is smaller than you think, and better than you think. Underwriting a terminal value requires a mandate that permits it, a hold period that survives it, and a partner who has done it before. That is not most funds. It is a specific and identifiable set of investors, and reaching thirty of the right ones beats reaching four hundred of the wrong ones by an enormous margin.
Sequence matters more than volume. The order in which investors hear the story determines the story. The first three meetings are where you find out which of your four assumptions is weakest, and you want that information before the investor who might actually lead the round has formed a view.
Structure is part of the argument. If a large share of the capital is buying depreciating assets, equity is often the wrong instrument for that share. Splitting the ask — equity for the thesis, debt or vendor finance for the steel — frequently changes the round from difficult to obvious. This is ordinary in project finance and strangely rare in venture.
And valuation is a consequence, not an input. A number you assert is a negotiating position. A number that falls out of a terminal argument the investor has now internalised is a conclusion they reached themselves. The second one holds under pressure; the first one does not survive the first committee.
VI. Why this is an intermediation problem
None of the above requires a firm like ours. A founder with the time, the training and the relationships can do all of it. The trouble is that these are the three things a founder building something physically hard has least of, and the raise competes directly with the work that generates the evidence the raise depends on.
There is also something a company structurally cannot do for itself: run a competitive process. A founder can talk to investors. A founder cannot easily put thirty of them on the same timetable, hold information back, manufacture tension, or tell a fund that terms are due Friday. Those are the mechanics that determine price, and they are the mechanics that public-market and M&A processes have used for decades precisely because they work.
Frontier companies have inherited the hardest valuation problem in finance and been handed almost none of the tooling built to handle it. That gap is the reason this firm exists.
This is a draft essay written during the build of this site. Rewrite in your own voice, and replace any illustrative example with something you can stand behind before publishing.