SAFE Note Dilution Is Decided Before Your Series A Price Is Set — Most Founders Model It After

A founder raises $300,000 on a SAFE with a $4 million cap in month two of the company's life. Momentum holds, so a second SAFE closes six months later: $500,000 at a $6 million cap. A well-known operator-angel wants in near the end of the year, and a third SAFE closes at $400,000 on an $8 million cap. Each of these felt like a fast way to keep the company funded between priced rounds. None of them looks large on its own. Add them up, though, and they already represent a meaningful share of the SAFE note dilution the founders will absorb. Most founders don't run that addition until a Series A term sheet forces them to.

This isn't a story about any one SAFE being a bad deal. Every instrument above is a normal seed-stage financing on its own. The problem shows up only when three or four of them sit on the same cap table at once, because the mechanism Y Combinator built into the post-money SAFE has a specific, underappreciated property: later SAFEs cannot dilute earlier ones. That protection has to come from somewhere. It comes entirely from the founders' side of the table.

Why stacking SAFEs feels costless in the moment

SAFEs exist because priced rounds are slow. There's no valuation negotiation, no board seat, usually no more than a five-page document and a wire transfer. An angel can commit in a phone call and send money the same week. For a founder trying to extend runway between a pre-seed and a proper Series A, that speed is the entire point, and it's a genuine advantage over spending six weeks on a priced round for $400,000.

The cost of that speed is that nothing about signing a SAFE feels like selling equity. No shares are issued. The cap table software doesn't usually need updating the same day. The founder's percentage ownership, as far as any dashboard shows, hasn't moved. The dilution is real, but it's deferred and invisible until a priced round forces every outstanding instrument to convert at once, which is exactly the moment founders are least equipped to renegotiate anything.

How SAFE note dilution actually adds up (not the way most guides describe it)

A lot of fundraising content describes SAFE stacking as a compounding effect, as if each new instrument recalculates the ones before it and the founder's position erodes faster with every close. Y Combinator's own post-money SAFE documentation says something more specific, and more useful: additional SAFEs do not dilute the SAFEs that came before them. Only the current stockholders, in practice the founders and any common stock, absorb the dilution from a new SAFE, right up until the priced round.

That single design choice is why each post-money SAFE has a clean formula: the investor's ownership percentage is simply the amount invested divided by the valuation cap, fixed at signing and untouched by whatever gets raised afterward. It also means the aggregate effect of stacking several SAFEs is much closer to plain addition than to some exotic compounding curve. Three SAFEs that each imply 5 to 8 percent ownership do not turn into 30 percent through some multiplier. They turn into roughly the sum of the three, taken straight out of the founders' share, one signature at a time.

Addition sounds manageable. It's the fact that nobody actually does the addition that causes the surprise. Three SAFEs signed across ten months, negotiated with three different investors, documented in three different PDFs, rarely get summed in one place until a Series A lawyer builds the real cap table.

A worked example: three SAFEs, one Series A

Take the three SAFEs from the opening example. Each implies a fixed ownership percentage the moment it's signed, calculated as investment divided by cap:

Before a single share of Series A preferred stock exists, the founders' fully diluted ownership has already dropped to roughly 79 percent, not because any one SAFE was unusually aggressive, but because three ordinary ones were signed without anyone adding the column. The Series A itself, typically priced to sell another 15 to 25 percent of the company to the new lead investor, comes on top of that, not instead of it.

What normal looks like, per Carta's 2025 data

The 20.8 percent in the worked example isn't an extreme case. Carta's State of Pre-Seed 2025 report puts the median post-money SAFE valuation cap at roughly $10 million for rounds raising $250,000 to $1 million, and around $15 million for rounds raising $1 million to $2.5 million, across the startups Carta tracks. On dilution specifically, a typical $1 million to $2.4 million SAFE round works out to a median of about 19 to 20 percent, with a meaningful share of companies landing north of 25 percent before their priced round even happens.

That range is the honest baseline to compare against, not a number to be alarmed by on its own. Twenty percent of the company sold across several SAFEs before a Series A is common, sometimes appropriate, and rarely fatal. What matters is whether a founder can say what their own number is on the day they're deciding whether to sign the next instrument — most can't, because nobody has been asked to add it up before.

The two things that actually make it worse than addition

Plain addition is a fair mental model for the base case. Two mechanisms can still push the real outcome past what a founder calculated by summing caps, and both are worth naming specifically rather than folding into a vague sense that stacking is risky.

• Most Favored Nation clauses. If a founder later signs a SAFE with better terms, typically a lower cap, for a different investor, an MFN clause lets an earlier SAFE holder swap into those same terms. The ownership percentage a founder thought was locked in on day one can move upward months later, after the founder has stopped thinking about that instrument at all.
• Uncapped or discount-only SAFEs. A SAFE with no valuation cap converts at the Series A price itself, adjusted only by whatever discount was negotiated. If the round prices lower than the founder hoped, that SAFE converts into more shares than a capped instrument would have, precisely at the moment the founder has the least leverage to push back.

Both mechanisms move the outcome in the same direction: worse for the founder than the naive sum implied. Neither is a reason to avoid SAFEs. Both are reasons to know, instrument by instrument, which of the SAFEs on the cap table carry an MFN clause and which are uncapped, rather than treating every SAFE as interchangeable paperwork.

Modelling the full stack before the next SAFE, not after

The fix here doesn't require a lawyer for every close. It requires treating the cap table as one running spreadsheet instead of a folder of signed PDFs.

• Keep a single running total of implied ownership across every outstanding SAFE, updated the day each one closes — not reconstructed from memory before a priced round.
• Before agreeing to a new SAFE's cap, ask what percentage of the company is already spoken for by the ones already signed, and treat the new instrument's percentage as additive to that number, not independent of it.
• Flag which instruments carry an MFN clause. Those are the ones that can still move after signing, and they deserve a note next to them, not a fresh read of the contract every time a new SAFE is discussed.
• Ask the Series A term sheet's option pool assumption in writing before comparing two offers on headline valuation alone. A bigger pre-money pool ask can cost more real ownership than a slightly lower price per share.
• Rerun the full conversion table, every SAFE at its own cap or discount, against the actual round price, before signing the Series A, rather than trusting the lead investor's model as the first time anyone checked the arithmetic.

When stacking isn't the problem

None of this argues against SAFEs, or against raising more than one. A company that keeps its combined SAFE dilution comfortably inside Carta's 19 to 20 percent median, raises each tranche at a cap that reflects real progress rather than just momentum, and keeps a running tally as it goes, is doing exactly what the instrument was built for. The instrument itself isn't the risk. The gap between what founders think they've sold and what the paperwork actually adds up to is.

A priced seed round avoids the stacking question entirely, at the cost of the legal overhead and valuation negotiation a SAFE is designed to skip, a trade worth naming explicitly rather than defaulting away from because SAFEs feel easier in the moment. The instrument was built to make each individual close simple. Making the aggregate visible is still the founders' job.