Stacked SAFEs: The Dilution Founders Don’t See Until the Series A

The SAFE is one of the best things to happen to early-stage fundraising. It is short, it is standard, it closes in a day, and it lets a company take money from the investor who is ready now without waiting for the three who are not. I have signed them as an angel, helped founders negotiate them as an advisor, and I would still recommend one to most pre-seed companies I meet. The problem is not the instrument. It is what happens when a company signs five of them over eighteen months, at four different caps, and never adds them up.

Each SAFE, taken on its own, looks small. A few hundred thousand here, a strategic angel there, a bridge from an existing investor when a quarter ran long. None of them changes anything on the day it is signed — no shares are issued, the cap table in the company's records still shows the founders and the option pool, and nobody feels diluted. The dilution is real from the moment the money lands, but it is invisible until the priced round converts everything at once. By then it is too late to decide whether it was worth it.

How the arithmetic actually works

Most SAFEs signed today are post-money SAFEs, the version Y Combinator introduced in 2018. The property that matters is simple: each holder's ownership at conversion is, to a first approximation, their investment divided by the post-money valuation cap. Put in $500,000 at a $5 million cap and you own 10% of the company as it stands immediately before the priced round. That percentage is fixed. It does not shrink when the company signs the next SAFE.

That is a genuine improvement for investors, and it makes each SAFE easy to reason about in isolation. But it has a consequence founders underweight: because SAFE holders do not dilute one another, every additional SAFE comes entirely out of the people who were already there — the founders and the option pool. The percentages simply add. Ten, plus twelve and a half, plus seven and a half, is thirty, and the founders absorb all thirty.

The older pre-money SAFEs behave differently — they dilute each other, and the total is harder to see in advance — and uncapped or discount-only SAFEs cannot be totalled at all until a price exists. Many companies have a mix. That alone is a reason to have someone actually read every document in the stack rather than rely on the summary in a spreadsheet tab.

A stack, converted

Here is a composite of a pattern I have seen more than once. Two founders hold 95% of a company, with a 5% option pool. Over eighteen months they raise three SAFEs: $500,000 at a $5 million cap from the first angels, $1 million at an $8 million cap after some traction, and $750,000 at a $10 million cap as a bridge into the Series A. $2.25 million in total.

Ask the founders what they have sold and the answer is usually anchored on the most recent cap. "We raised about two and a quarter at ten, so a bit over twenty per cent." The real figure is 30%, because the earlier money went in at lower caps and bought proportionally more. Before a single Series A dollar arrives, the founders are already at 66.5%, not the 73% or so they were carrying in their heads.

Then the Series A: $4 million at a $16 million pre-money valuation, so the new investor takes 20%. The term sheet also asks for the option pool to be topped up to 10% of the post-money capitalisation, with the top-up counted inside the pre-money — the standard ask, and a reasonable one if the company intends to hire. The new investor is not diluted by that top-up; everyone already on the cap table is, the converting SAFE holders and the founders alike.

Run the conversion properly and the founders come out holding about 48%. The SAFE holders hold about 22%, the pool 10%, the Series A investor 20%. The founders' own back-of-envelope version — 95%, less the twenty-odd per cent they thought the SAFEs had cost, less the 20% sold in the round — had them at roughly 59%. The gap is eleven points of the company, and it is made of two things nobody modelled: the early SAFEs at low caps, and the pool top-up, which on its own cost the founders about five points.

A SAFE is cheap to sign and expensive to add up. The price is paid in full on the day of the priced round, all at once, by whoever was not paying attention.

Why eleven points matters more than it sounds

Eleven points of a company that eventually fails is worth nothing, and it is tempting to wave the whole thing away on that basis. Three reasons not to.

The first is the next round, not the exit. Series A investors look closely at how much of the company the founders will hold after the round, because it predicts how much room is left for a Series B and C before the founders own too little to stay motivated through a hard stretch. There is no fixed threshold, but a founding team that enters its Series A below roughly half invites a conversation it would rather not have — sometimes about refreshing founder equity, which is itself dilutive to everyone else, sometimes about whether the cap table is already too crowded to finance. Companies that stacked SAFEs heavily find themselves defending their own history in the first meeting.

The second is negotiating position. A founder who learns the conversion maths from the lead investor's model, in the week the term sheet arrives, negotiates the pool size and the valuation with less information than the other side of the table. The pool top-up in particular is negotiable — its size should come from the hiring plan for the next eighteen to twenty-four months, not a round number — but only if the founder arrives with that plan and has already seen what each point of pool costs them. I have written separately about why the hiring plan is the budget; this is one of the places where that pays off directly.

The third is the quiet one: trust inside the founding team and with early employees. Option grants made during the SAFE period were usually explained as a percentage of the company. If nobody modelled the stack, those percentages were overstated, and an early engineer who was told they held half a per cent learns at the Series A that it has become about a quarter of one per cent — well below what anyone had led them to expect. That conversation is harder than any of the investor ones.

What to do instead

None of this argues against SAFEs. It argues for treating them as what they are — equity sold at a price, with the paperwork deferred — and keeping the books accordingly.

Keep a pro forma cap table and update it the day each SAFE is signed, not the month the priced round starts. The formal cap table only records issued shares, so it will show the founders at 95% right up until the round closes. The pro forma one shows every SAFE converted at its cap on a fully diluted basis, and it is the only version that tells the truth about ownership in the meantime. It takes an afternoon to build and five minutes to maintain. Most cap-table software can model SAFE conversion, but it only works if every instrument, side letter and MFN clause has actually been entered, and in my experience that is where the gaps are.

Set a dilution budget for the whole pre-priced period and track against it. A common working ceiling is somewhere around 20–25% of the company sold across all SAFEs before the Series A, though the right figure depends on how much capital the business genuinely needs to reach the milestones that justify a priced round. The point is less the number than the act of choosing one: every new SAFE then becomes a decision about spending from a known budget rather than an isolated yes to an investor who happens to be ready.

Model the priced round before anyone offers one. Three scenarios are enough — a disappointing valuation, a reasonable one, a strong one — each with a realistic pool top-up and the full SAFE stack converting. It is the most useful single page a founder can have in the six months before a Series A, and it turns the term-sheet negotiation from a surprise into a comparison against numbers already understood.

That last question deserves more attention than it gets. Once a company has raised a couple of million on SAFEs and is still some way from a Series A, a priced seed round — with a real price, a real board structure and a real share issuance — is often simpler for everyone than a fifth SAFE at yet another cap. The legal cost is higher, and the process is slower, but the cap table stops being a forecast and becomes a record. For a company that has already lost track of what it has sold, that clarity is worth more than the few weeks saved.

Where this sits in the finance work

This is the most common cap-table problem I see in companies approaching their first priced round, and it is almost never caused by carelessness. It is caused by the instrument working exactly as designed — fast, simple, deferred — in a company where nobody owns the question of what has been sold in total. Founders are busy raising, investors each see only their own document, and the company lawyer drafts what is asked for. The pro forma cap table falls between all of them.

In the startup advisory work I do, reconstructing the SAFE stack is usually the first exercise before any Series A preparation, alongside the checks I described in what investors actually check in a data room. It is rarely a pleasant afternoon, but it is much better spent six months before the term sheet than six days after it. The founders who come out of their Series A with the ownership they expected are not the ones who raised less on SAFEs. They are the ones who knew, at every point along the way, exactly what they had already sold.