What is normal dilution at each stage
Median dilution, raise size and post-money valuation per round, next to what founders actually still own at each stage. Both sets are Carta's own published figures, kept apart because they measure different things.
How much of the company sells in one round
| Stage | Median dilution | Median raise | Median post-money |
|---|---|---|---|
| Pre-seed | not published | — | — |
| Seed | 18% | $4.1M | $24.3M |
| Series A | 18% | $14.4M | $80M |
| Series B | 12% | $25M | $191M |
| Series C | under 10% | $40M | $391M |
| Series D | 8% | $63M | $789M |
- Median dilution
- not published
- Median raise
- —
- Median post-money
- —
- Median dilution
- 18%
- Median raise
- $4.1M
- Median post-money
- $24.3M
- Median dilution
- 18%
- Median raise
- $14.4M
- Median post-money
- $80M
- Median dilution
- 12%
- Median raise
- $25M
- Median post-money
- $191M
- Median dilution
- under 10%
- Median raise
- $40M
- Median post-money
- $391M
- Median dilution
- 8%
- Median raise
- $63M
- Median post-money
- $789M
Pre-seed No priced-round median exists, because pre-seed is almost entirely unpriced: 93% of pre-seed rounds in Q2 2026 were SAFEs. The dilution is real but deferred, and it lands all at once when those SAFEs convert at your first priced round.
Seed Carta reports seed dilution has been ticking down for months. Option pool top-ups sit on top of this figure, and so does any SAFE converting into the round.
Series A Carta calls this the lowest Series A dilution at any point in the last few years. For comparison, the median Series A in Q1 2025 ran 17.9%, down from 20.9% a year before that.
Series B Dilution per round drops from Series A onward: the pre-money is far larger, so a bigger cheque buys a smaller slice, and pool top-ups shrink.
Series C Carta publishes Series C as selling less than 10% of the company on median, rather than as a point figure, so treat 10% as a ceiling and not an average.
Series D Late capital is the least dilutive per dollar, but by this point founder leverage over terms, board composition and structure matters more than the percentage.
Valuations are published post-money, as Carta reports them. We do not subtract the raise to show a pre-money median, because the median-valuation company is not the median-raise company and the difference of two medians is not itself a median. Carta, VC startup fundraising benchmarks, published 10 July 2026: more than 1,000 software rounds closed in the six months to July 2026, excluding bridges and extensions. About half of those companies would meet a definition of AI-native, by Carta's own estimate.
What founders actually still own
| On raising | Median founding-team stake |
|---|---|
| Seed | 56% |
| Series A | 36% |
| Series B | 21.8% to 27.3% |
| Series C | 16.1% |
Seed The median founding team still holds a majority at seed, and the employee pool is 12.1%.
Series A Majority control is typically gone by the A. Carta splits this by sector: 37.5% in digital industries, 30.5% in physical ones.
Series B Carta publishes this one as a split rather than a single median: 27.3% for AI founding teams, 21.8% for non-AI.
Series C The crossover point. At Series C the median employee option pool (16.8%) is larger than median founder ownership (16.1%).
Collective founding-team ownership, fully diluted, not per founder. Carta, Founder Ownership Report 2026, published 12 March 2026: rounds raised on Carta from 2021 through 2025.
Compounding the round medians gives 67% at Series A. Carta measures 36%.
Take the dilution column at face value and the arithmetic says a founding team keeps about 82% through seed and about 67% through Series A. Carta's own measurement of real cap tables at those same stages is 56% and 36%. Nearly half the ownership that founders expect to still hold at Series A is gone, and none of it is visible in a single round's headline dilution figure.
Four things account for the gap. Option pool top-ups are placed pre-money and charged entirely to founders. SAFEs signed before the round convert into it. More than one round often happens inside a single stage label, as bridges, extensions and second seeds. And medians do not compound: the median seed company is not the median Series A company, so chaining medians together describes no real company at all.
This is why the calculator asks for your pool top-up and your SAFEs separately rather than applying a stage average. A founder who models round dilution alone will overestimate what they keep, every single time.
Where the market sits now
Falling at every stage. Carta describes Series A dilution as lower than at any point in the last few years, and seed dilution as having ticked down for months. Terms currently favour founders more than they have since before the 2022 reset.
The down-round rate fell to 11.4% of all new rounds in Q1 2026, back in line with 2019 and 2020 levels. Just over 19% in Q1 2025, above 20% in seven of the eight quarters from Q2 2023 to Q1 2025, and about 17% by Q3 2025. If you last raised in 2021 or early 2022 the risk is still real, but it is no longer the base case.
Rising at every stage from seed through Series C between Q1 2025 and Q1 2026, and sharply so at the top. At seed the 90th-percentile valuation in Q1 2026 was nearly four times the median; at Series A, nearly five times. A median is a poor guide to what a hot round looks like.
More than 60% of all venture capital raised on Carta in Q1 2026 went to AI companies, the highest share recorded. An AI foundational-model startup at Series A was raising at a $300M median valuation against $55M for a non-AI startup at the same stage. These are not comparable markets, and a single blended median hides that.
Carta, State of Private Markets Q1 2026, published 29 May 2026, and the fundraising benchmarks above.
Frame the decision in terms of expected value, not percentage. If this round's post-money is $X and your remaining ownership is Y%, your expected paper value is X x Y. If that exceeds the expected value of your alternative path (smaller round, bootstrap, different valuation), the dilution is worth it. Use the homepage calculator to model both paths and compare.