Accept the cap you can clear. Chasing the cap you can brag about is how today’s good news becomes tomorrow’s down round. A valuation cap does not price your company today. It fixes a strike price on a priced round you have not yet raised, and if that round comes in below the strike, the number you celebrated becomes the number that reprices you downward. When the path to clearing a high cap is thin, a lower cap paired with a discount is usually the better trade, because a discount rides whatever the market decides and a cap bets against it.
Here is the mistake, stated plainly. Founders negotiate the cap as if it were a valuation. It is not. A cap is not a valuation. It is a bet on a valuation you have not yet earned. The counterparty knows this. The instrument is built around it. You should price the bet before you sign it.
Why these three models
A cap is a pricing mechanism with three moving parts, and no single lens catches all three. So run three that disagree by design.
Comparative statics handles the mechanics: move the cap up, and watch the probability of a future down round move up with it, holding your real trajectory fixed. Base rates handle the event the cap is a promise about: how often a seed company actually reaches the priced round that would clear the cap. Prediction-market price formation handles the part you do not control: the next round’s price is a verdict a crowd of investors reaches, not a number you set, and in a thin market that verdict is noisy. One model is equilibrium, one is regime-and-base-rate, one is about aggregation under uncertainty. Three different failure modes, so the errors do not stack the same way.
The ensemble method would ordinarily append a behavioral layer. It is real here, founders anchor on the headline cap and read it as validation, but it does not earn its own card. It rides inside the comparative-statics card as the reason the derivative feels like a win. I say where it sits and move on.
The framework: what the three models see together
1. The mechanism: raise the cap, raise the down-round probability
Treat the price of your next priced round as a number you cannot know yet, drawn from a range of plausible outcomes. Call the cap you agree today C. A post-money SAFE with cap C converts, roughly,7 as if your company were worth C at that next round. If the round prices above C, the cap bites in the investor’s favour and everyone is fine: it was an up round relative to the mark. If the round prices below C, you have a down round at conversion, with worse terms, deeper dilution, anti-dilution math biting earlier investors, and a signal to the market that your last number was fiction.
Now the comparative static. Hold your actual business trajectory fixed and increase C. The probability that your next round prices below C rises mechanically, because you have raised the bar the market must clear. The headline number goes up and your down-round risk goes up in the same motion. That is the sentence founders miss: the derivative of down-round probability with respect to the cap is positive. A higher cap enlarges the wager while it flatters the number.
This is also where the cap-versus-discount trade lives. A discount, say twenty percent off the next round, does not set a strike. It rides whatever price the market lands on and takes a fixed haircut off it. A discount can never manufacture a down round, because it never bets against a number. A cap can, and does, whenever the number is set above what you grow into. The Y Combinator post-money SAFE ships in four shapes for exactly this reason: cap only, discount only, cap and discount, and most-favoured-nation.1 Founders reach for cap only and treat the discount as a giveaway. It is closer to insurance.
Comparative statics. Assumes your trajectory is independent of the cap you set. Fits because conversion is mechanical: a strike above the realised price forces a down round. Breaks when the cap itself changes behaviour, a high mark that funds a real acceleration that clears it. Counteracts the instinct to maximise the headline. May reinforce the behavioral read: the higher number feels like validation, which is precisely why the rising risk stays invisible.
2. The base rate: the priced round often never arrives
The cap is a promise about a specific event, the next priced round. So ask the only question that matters before you set a strike on it: how often does that event happen at all.
The answer is sobering even in deep markets. On Carta’s data, 30.6 percent of United States companies that raised a seed round in the first quarter of 2018 reached a Series A within two years. For the first-quarter 2022 cohort, the same figure was 15.4 percent.2 Recent cohorts graduate at roughly 8.9 percent inside their first year.3 And when the priced round does not come, a bridge stands in for it: about 40 percent of seed-stage rounds in 2024 were bridges, up from 36 percent the year before.4 A bridge is the market telling you the priced round you promised a cap against did not materialise on schedule.
Now layer Africa on top. The continent’s early-stage funnel is thinner, the follow-on pool is smaller, and the exits that pull capital back through the system are rarer. Total 2024 startup funding was down about 25 percent year on year by one credible count, with equity at 1.5 billion dollars.5 The next-round investor who would clear your cap is one of a small set, and many of them pulled back. You are setting the strike on a round most companies in your reference class never raise.
Base rates and reference class. Assumes your company behaves like its cohort until you have evidence it does not. Fits because graduation is the exact event the cap is priced against, and it is measurable. Breaks when you have a genuine structural reason to sit in the top decile, real acceleration, not a good quarter. Counteracts the inside view that your round is the exception. May reinforce the mechanism model: a low graduation rate is a high prior probability the round prices below any ambitious cap.
3. The price you cannot set: the next round is a crowd’s verdict
Suppose you clear the base-rate hurdle and a priced round does arrive. Here is the part the cap negotiation hides: you still do not set the price. A priced round is a market aggregating the independent judgements of many investors. It is a wisdom-of-crowds verdict on what you are worth, and like any such verdict it lands somewhere in a distribution, not on a point.
The dispersion of that distribution is the whole game. In a thin market it is wide. Consider that two credible trackers cannot agree on how much African startups even raised in 2024: one puts the total at 2.2 billion dollars, another at 3.2 billion, a gap of more than 45 percent on the same year.6, 5 If the market cannot price the aggregate to within half, it will not price your specific round to a point either. The verdict has fat tails.
This is what makes the discount structurally different from the cap in a market like yours. A cap is a single point you fix today, blind, against a wide and noisy future distribution. A discount is a rule that reads the distribution after it resolves and takes a fixed slice. The value of riding the market rather than betting a point rises with the variance of the price. African early-stage prices are high-variance. So the instrument that rides the aggregate is worth more here than it is in a deep, liquid market where the next round’s price is nearly known in advance.
Prediction-market price formation. Assumes the priced round aggregates many independent valuations into one number. Fits because that number is a crowd verdict, not yours to set, and its spread is observable in how badly the market prices even totals. Breaks when one buyer sets the price alone, a DFI or a strategic pricing to mandate, so there is no crowd to aggregate. Counteracts the illusion that a cap fixes your value. May reinforce the case for a discount over a high cap whenever the future price is wide.
GEER: the moves, from the one that costs nothing to the one you cannot undo
Order them by cost and reversibility.
- Compute the clearing probability before you counter. Free. Take each cap on the table, place it against the graduation base rate and your own runway-funded milestone, and estimate the chance your next round prices above it. If a high cap only clears in the top decile of outcomes, it is a liability wearing the costume of a compliment.
- Ask for the discount, not just the cap. Cheap and reversible inside the negotiation. A cap-and-discount SAFE lets the instrument ride the market when the market is kind and holds a floor when it is not. Founders concede the discount to hold a higher cap. That is backwards in a thin market.
- Lower the cap to what a fundable milestone supports. Tie the number to the specific proof this money buys within the runway it buys. A cap you can grow into within the round’s own timeline is a cap that converts clean.
- Cap the stack, not the note. Post-money SAFEs accumulate. Three SAFEs at an ambitious cap imply an ambitious post-money in aggregate, and the priced round must clear the sum, not the last one. Track the implied post-money across every instrument outstanding, because that is the real strike.
- Choose the instrument last, and deliberately. Priced round, capped SAFE, uncapped SAFE with a discount, or a note each move the down-round risk differently. This is the least reversible lever, so it belongs in the dated portfolio below, not in a reflex.
RADAR: what to settle before the SAFE is signed
DO NOW (T+0 to T+3). Reprice every offer on down-round probability, not headline cap. For each cap, write the chance the next round clears it given the base rate and your funded milestone. If the highest cap has the worst clearing probability, counter with a lower cap plus a discount and say why in one line: you are buying a clean conversion, not a bragging number.
HEDGE (T+3 to T+14). Add the cheap tail insurance. Put a discount or a most-favoured-nation clause on the instrument so it tracks the market’s verdict instead of your guess. Then sum the implied post-money across all SAFEs outstanding and make sure the total is a strike you can clear, not just each note alone.
DEFER AND TRIGGER (T+14 to T+28 and beyond). The irreversible move is letting a cap stand into a round that will not clear it. Pre-commit the trigger now. If, by the milestone this money was meant to fund, your next-round pipeline has produced no term sheet priced above your cap, open the bridge or extension conversation before the priced round forces a down mark. Treat the bridge as a normal outcome. Two of every five seed rounds now are bridges.4 Deciding to raise one on your timing beats being repriced on someone else’s.
CHAIN: what usually happens after you sign a top-of-market cap
Match on structure, not sector. The reference class is companies that fixed a strike against an unrealised future round at the top of a hot market, whatever they built. The base rate for that class is unkind: most did not clear the ambitious cap, the graduation rate to a priced Series A ran in the mid-teens for the recent cohorts, and roughly two in five of the rounds that did happen were bridges rather than fresh priced rounds.2, 4 The modal outcome after a proud cap is a flat round, a bridge, or a recut set of terms.
Adjust for the present state. You are raising into a market that shrank about a quarter year on year and where the follow-on pool is smaller than the United States base these rates come from.5 Push your down-round probability above the imported number, not below it.
Then subtract the counterfactual. Some of those down rounds would have happened at any cap, because the business slowed. That harm is not the cap’s fault, so do not charge it to the cap. What the cap is responsible for is the incremental down round: the case where a lower strike would have converted clean and the ambitious one detonated instead. That residual, and only that, is what you are actually negotiating. It is smaller than the headline horror stories and larger than zero.
Matrix-break flag. If your next round will be led by a development-finance institution or a strategic buyer pricing to a mandate rather than to market comparables, the crowd-verdict model breaks. There is no aggregate to ride, so a discount loses its edge and the price is one principal’s to set. In that world the negotiation is about the covenant, not the cap, and this ensemble is looking at the wrong instrument.
Where these three models go blind
The ensemble prices the probability of a down round. It cannot tell you whether this particular investor will wield the cap adversarially or reprice with you cooperatively when the round comes in soft. The relationship, the reputation cost of a hard recap, the investor’s own reserve position, none of that is in the math, and all of it moves the real outcome. The models also cannot see the genuine break: the company that accelerates hard enough to clear any cap and make this entire analysis moot. That founder exists. The base rate says you should not assume you are her until the evidence says so, which it has not yet.
So the honest read folds the behavioral point back in. The reason founders skip this whole exercise is that the high cap arrives feeling like a verdict already delivered, a validation to bank rather than a bet to price. Treat it as the bet. Here is the action that survives the blind spots: counter to the cap you can clear on the base rate, add the discount, and pre-commit the bridge trigger, because you cannot bank a breakout you have not yet produced, and a cap you clear clean costs you nothing you were ever going to keep.
Joshua Agonya Pi’Rwot, Founder.
Sources and notes
- Y Combinator, “Post-Money SAFE” documents and user guide, which publish the instrument in four forms, valuation cap only, discount only, cap and discount, and most-favoured-nation, and specify how the cap sets the conversion price at the next priced round. ycombinator.com/documents. Verified: page returns 200 and the body carries the “Valuation Cap” and “Discount” document variants.
- Carta, VC Fund Performance data (Q1 2024), reporting 30.6 percent of United States seed companies from the Q1 2018 cohort reaching Series A within two years versus 15.4 percent for the Q1 2022 cohort. Carta’s data host bot-blocks direct fetches; figures read via the readable mirror Konvoy, “Failure to Launch: The Series A Crunch,” konvoy.vc, and corroborated at incisive.vc. Both verified to carry the 30.6 and 15.4 percent figures in body.
- Carta, State of Seed data, on recent seed cohorts graduating to Series A at roughly 8.9 percent within their first year. Read via the readable mirror SaaStr, “The Real State of Seed Today,” saastr.com. Verified: body contains the 8.9 percent figure.
- Carta, State of Private Markets, on approximately 40 percent of seed-stage rounds in 2024 being bridge rounds, up from 36 percent in 2023. Read via readable mirrors incisive.vc and Forum Ventures, forumvc.com. Both verified to carry the 40 and 36 percent figures in body.
- Africa: The Big Deal, “2024 Round-Up,” reporting total African startup funding of 2.2 billion dollars in 2024, down about 25 percent year on year, with equity at 1.5 billion dollars, down 11 percent. africathebigdeal.com (PDF, machine-readable text layer). Verified: figures present in extracted text.
- Partech, “2024 Africa Tech Venture Capital Report,” reporting total 2024 funding of 3.2 billion dollars, of which equity 2.2 billion and debt 1.0 billion (31 percent). The gap between this total and the Africa: The Big Deal total for the same year is the measurement dispersion cited in the text. partechpartners.com, readable mirror ecofinagency.com. Verified: both carry the 3.2 billion total and the equity and debt split in body.
- Note on conversion mechanics. This piece treats a post-money SAFE cap as converting, to first approximation, as if the company were valued at the cap when the next priced round exceeds it; the exact share count depends on the SAFE’s price-per-share definition and any discount. The directional claim, that a cap set above the realised round price produces a down mark at conversion, holds across the standard forms in fn1.