The Median Seed Valuation Is $19.7 Million. It Describes 426 Rounds, 27% Fewer Than Last Year.
Carta's Q3 2026 data shows record seed prices and a shrinking count of priced seed rounds. The second number changes how to read the first.
The median seed valuation in 2026 is $19.7 million pre-money, according to Carta's State of Private Markets report for Q3 2026, written by Hamza Shad and Kevin Dowd and published in early November. That is 84% above the 2021 median, a year most people remember as the top of the last cycle. The same report says investors closed 426 priced seed rounds on Carta in the quarter, 27% fewer than in Q3 2025.
Most coverage leads with the first number and mentions the second in passing. For a founder planning a 2027 raise, the order should be reversed. A median is computed over the rounds that happened. When the set of rounds that happen gets smaller, the median can move a long way without any single company being priced differently. This piece works through how far, using the report's own figures and some arithmetic.
What the Q3 2026 report says
We could not open the full 16-page report, so the figures here come from two places: Carta's own key takeaways as republished by The Idea Farm, which cover the valuation, seed and down-round numbers, and Doug Dyer's summary in the TheFundCFO newsletter, dated 10 November, which reproduces more of the stage-level detail. Where a number below comes only from that summary, the table says so.
| Measure | Figure | Change |
|---|---|---|
| Capital raised, all stages, Q3 | $49.3B across 1,179 rounds | Round count down 21% year on year |
| Seed capital deployed, Q3 | $4.1B | Up 21% year on year |
| Priced seed rounds, Q3 | 426 | Down 27% year on year |
| Series A rounds, Q3 (via summary) | 365 | Down 22% year on year |
| Median seed pre-money, 2026 | $19.7M | Up 84% on 2021 |
| Median Series A pre-money, 2026 | $63.3M | Up 56% on 2021 (via summary) |
| Down rounds, share of Q3 priced rounds | 12.8% | Lowest in four years; peak was 23.4% in Q1 2024 |
| Series A rounds that were bridges, 2026 (via summary) | About one in three | Up from 29% in 2025 |
| Rounds with participating preferred, Q3 (via summary) | Over 10% | Highest in six years |
| Median days from seed to Series A (via summary) | 664 | Down from a peak of 869 |
Read as a list, this is a healthy market: prices up, down rounds rare, founders moving faster between rounds. Read as a population, it is a market in which fewer companies get through each gate and the ones that do are priced higher. Both readings are true. Only the second tells you what to expect when you are the one raising.
How a shrinking round count lifts the median seed valuation
Suppose nothing about prices changed between last year and this one. Every company that could raise a seed round would be offered the same valuation as before. The only difference is that investors pass on the cheapest 27% of rounds, the ones with the least traction or the least fashionable sector.
The surviving rounds are the top 73% of last year's distribution. Their median sits halfway through that 73%, which is 27 + 36.5 = the 63.5th percentile of the old distribution. So the reported median jumps from last year's 50th percentile to last year's 63.5th, and no term sheet was repriced to get there.
How big that jump is depends on how spread out seed valuations are. Carta does not publish the spread in the material we could read, so the sketch below assumes one. Treat the output as an illustration of the mechanism, not an estimate of the real effect.
import random, statistics
random.seed(7)
SIGMA = 0.6 # assumed spread of log-valuations, not a Carta figure
# Last year's market: 10,000 would-be seed rounds, median pre-money 10.0
last_year = [10.0 * random.lognormvariate(0, SIGMA) for _ in range(10_000)]
# This year: identical companies, identical prices.
# The only change is that the cheapest 27% never get a priced round.
cutoff = sorted(last_year)[int(0.27 * len(last_year))]
this_year = [v for v in last_year if v >= cutoff]
print(round(statistics.median(last_year), 1)) # ~10.0
print(round(statistics.median(this_year), 1)) # ~12.2With that assumed spread, cutting the bottom 27% lifts the median by about 22%. A tighter spread (0.5) gives roughly 18%, a wider one (0.8) about 30%. The 84% rise since 2021 is far larger than any of these, so selection is not the whole story. Prices have risen as well, most visibly for AI companies. The point is narrower: some unknown share of the headline increase reflects which companies got a round, and that share does not transfer to a company on the wrong side of the cut.
“A median over the rounds that closed says nothing about the rounds that did not.”
The average seed round is about twice the median
A second piece of arithmetic points the same way. Divide the $4.1 billion of seed capital by 426 rounds and the average Q3 seed round comes to about $9.6 million. Work backwards from the year-on-year changes and Q3 2025 had roughly $3.4 billion across about 584 rounds, an average near $5.8 million. The average seed round grew by about two-thirds in twelve months.
The median seed round is much smaller. Secondary summaries of Carta's mid-2026 benchmarks put it between $4.1 million and $4.5 million. An average more than double the median means a small number of large rounds account for much of the capital. A handful of AI infrastructure seeds at nine-figure valuations can add 21% to seed dollars while the count of ordinary rounds falls by a quarter.
So "seed capital is up 21%" and "the median is at a record" describe the top of the distribution well. Neither says much about a B2B software company with $300,000 of annual recurring revenue and no AI in the pitch.
"Seed" is a smaller category than it used to be
Part of the fall in priced seed rounds is relabelling. The earliest money now mostly arrives on SAFEs and convertible notes, which Carta tracks in a separate pre-seed series.
The scale gap between the two series is large. Value Add VC's comparison of venture trackers, published in October, notes that Carta's pre-seed series recorded $3.19 billion across 11,546 SAFEs and convertible notes in Q2 2026. A single round usually contains several instruments, so that is not 11,546 companies. It is still an order of magnitude more activity than 426 priced seed rounds.
When small first cheques move out of the "seed" label and into "pre-seed", the rounds left carrying the seed label are later, larger and better evidenced than a 2021 seed was. Comparing a 2026 seed median with a 2021 seed median compares two different kinds of company. Some of the 84% is price inflation, some is selection, and some is that the word now means something closer to what a small Series A meant five years ago.
A valuation is a price plus terms
The summary of the report flags one more detail that the headline median hides. More than 10% of Q3 rounds included participating preferred stock, the highest rate in six years. Participating preferred lets an investor take their money back first and then share in what remains, so two rounds at the same pre-money can pay founders different amounts in a modest exit.
Liquidation preferences above 1x and cumulative dividends remain rare, according to the same summary, so this is not a return to 2002. It is a reminder that when prices are high and investors are selective, some of the negotiation moves out of the number and into the structure. A benchmark that reports only the number will miss it.
What to benchmark against instead
The median is useful for knowing what the market paid. It is a poor anchor for what you should ask. Four comparisons do more work.
- Your own cohort. Ask investors and other founders what companies in your sector, city and revenue band have closed at in the last two quarters. Five real data points from comparable companies beat a platform-wide median.
- The next-round bar. A seed at $19.7 million pre-money sits 3.2x below the $63.3 million Series A median. Investors at Series A will expect the business to have grown into that step. A higher seed price raises the bar you must clear and does nothing to raise your odds of clearing it.
- The bridge rate. About one in three Series A rounds in 2026 was a bridge, according to the summary of Carta's data. A seed price you cannot grow past in two years is how a company ends up in that third.
- The clock. The median gap between seed and Series A is 664 days. If you allow four to six months for the raise itself, which is our assumption and not a Carta figure, a seed round has to fund something like 26 to 28 months for the median company.
Which number to carry into 2027
Every figure in a venture report is conditional on a deal having closed on that platform, at that stage label, in that period. The report cannot count the rounds that did not happen, and those are the ones a founder most needs to know about.
The practical habit is to pair each price with its count. $19.7 million across 426 rounds, down 27%, is a different fact from $19.7 million alone. If Q4 shows the seed count recovering while the median holds, prices have moved for a broad set of companies. If the count keeps falling and the median keeps rising, the market is narrowing, and the median is describing fewer and fewer of the founders who read it.
Frequently asked questions
Related reading
Annual Prepay Is a Retention Tool, Not a Growth Tool. ChartMogul's Data on 2,500 SaaS Companies Shows Why
ChartMogul's billing report found that sub-$1M SaaS companies with mostly monthly revenue grew far faster. Annual plans protect retention, but they are the wrong first lever.
Per-Resolution Pricing Sounds Fair. The Bill Depends on How the Vendor Defines "Resolved"
Per-resolution and per-action pricing for AI agents shifts the real negotiation to one word. A worked cost model and five contract clauses to settle before you sign.
SOFTEX to EDF on 1 October: The Real Change for SaaS Exporters Is Monthly Reconciliation
From 1 October 2026 the SOFTEX form gives way to a monthly Export Declaration Form. The new 15-month window helps, but matching invoices to remittances is now your job every month.