Robust funding analysis · controlling for power law + age

Beyond power law — what YC funding really looks like

The headline 'YC raised $8.83B' is mostly six unicorns. Strip the top 1% and the picture collapses: median raise is $500K (the YC SAFE itself), and post-launch RFS cohorts do not raise more per company than pre-launch cohorts once you control for age. This is the version of the analysis that survives in venture — where one outcome makes the fund but the median is the population.

39%

of capital in top 1%

$500K

median raise

$15.0M

p90 raise

themes with age-matched growth >1.0×

1. The power law — 1% of companies hold 39% of capital

12 companies (1% of those with funding data) account for 38.6% of the $8.83B total. Any metric that uses 'sum raised' is reporting on those 12.

When you order all 1,215 funded companies by total raised and walk down the rank, the cumulative curve climbs almost vertically at the start. The top 1% holds 39% of all capital, the top 10% holds about 86%. That is a textbook power law — exactly what venture portfolios are supposed to look like, but it means group-level totals are dominated by handful of outliers.

Takeaway

Stop comparing sums. Two cohorts can differ by orders of magnitude in sum just because one has a Legora ($860M) and the other does not. Sum is a measure of how many unicorns are in your bucket, not how the typical company in your bucket is doing.

2. Strip the top 1% — the distribution barely moves

Without the top 1%, the median is still $500K, p75 still $3.4M, p90 drops modestly from $15M to $13M. The unicorns own the right tail; they don't shift the middle.

On a log-scaled histogram you see a tall column at the $500K bucket (the standard YC SAFE), a fat middle in the $1-10M range (seed and bridge rounds), and a thin right tail tapering past $100M. Removing the top 1% chops the tip off the tail but doesn't reshape the body — because the body is dominated by recent batches still sitting at the SAFE floor.

$500K

median (full)

$500K

median (-top 1%)

$15.0M

p90 (full)

$13.0M

p90 (-top 1%)

Takeaway

The 'average YC company' is a $500K-$3M seed operator. The unicorn narrative applies to <2% of the dataset. If you're using YC as a signal for 'these companies will raise big', the base rate is 11% reaching $5M+ by their first year.

3. Age, not RFS, drives the funding climb

Median raise jumps from $500K at <1 year post-batch to $5-11M at 2.5-3+ years post-batch. The single strongest predictor of 'how much has this company raised' is 'how long has it existed'.

Bucket every funded YC company by years since their batch start. Companies in their first year almost universally sit at the $500K SAFE. By year 2-2.5, the median is still $500K but the trimmed mean rises (more companies break through). By year 2.5-3, the median jumps to $11M — that's when survivors close Series A. Beyond 4 years, the survivors-only sample looks similar to year 3 (mortality + reporting bias).

Age	n	Median	Trimmed mean	p25 → p75	% ≥ $5M
<0.5y	100	$500K	$807K	$500K → $500K	11%
0.5-1y	263	$500K	$780K	$500K → $500K	10%
1-1.5y	300	$500K	$521K	$500K → $500K	5%
1.5-2y	14	$2.0M	$8.4M	$500K → $18.6M	36%
2-2.5y	242	$500K	$1.5M	$500K → $2.5M	17%
2.5-3y	43	$11.0M	$16.5M	$3.4M → $31.3M	63%
3-4y	77	$5.0M	$8.1M	$2.3M → $14.1M	53%
4y+	174	$5.1M	$8.8M	$2.1M → $15.2M	51%

Takeaway

Pre-launch RFS cohorts are by definition older. Comparing their cumulative raises against post-launch cohorts (mostly fresh YC outputs) is the textbook confounder. Any 'pre dominates post' finding in funding is mostly age, not theme quality.

4. Survival funnel: % of cohort clearing each tier

By age 2.5-3 years, 91% of YC survivors clear $1M, 63% clear $5M, 47% clear $15M, 16% clear $50M. These are the conversion rates that matter.

Plotting the share of each age cohort that has crossed $1M / $5M / $15M / $50M raised makes the funding ladder explicit. The 2.5-3 year window is the peak conversion zone — old enough to have closed Series A, young enough that the survivors haven't yet been censored by quiet failure. Beyond 4 years the survival-bias kicks in differently: only the still-active ones show up in the data, and the late-stage Series B/C numbers settle.

Age	n	≥ $1M	≥ $5M	≥ $15M	≥ $50M
<0.5y	100	20%	11%	5%	0%
0.5-1y	263	18%	10%	2%	0%
1-1.5y	300	11%	5%	2%	0%
1.5-2y	14	50%	36%	29%	7%
2-2.5y	242	34%	17%	8%	1%
2.5-3y	43	91%	63%	47%	16%
3-4y	77	82%	53%	25%	9%
4y+	174	80%	51%	29%	9%

Takeaway

If you're building base rates for 'will a YC company hit X by year Y', this is the table. The drop-off between $5M and $15M (63% → 47%) is the Series A bar. Between $15M and $50M (47% → 16%) is the Series B bar. Everything beyond is noise.

5. RFS amplification — 3 views of the same data

Among 9 RFS themes with adequate cohorts on both sides of launch, ZERO show clean per-company growth post-launch after age-matching. The 'wave' you see in sum totals is the volume effect — more matched companies post-launch, not bigger raises per company.

For each RFS, we computed three pre-vs-post ratios. Sum growth is the naive sum_post/sum_pre — biased by unicorns and by cohort size. Median growth strips power law but not age. Age-matched median growth restricts both groups to companies whose batch was 6 months-2.5 years from launch — apples to apples. Every line that looks like a wave under sum collapses under age-matching. Defense, Robotics, US Manufacturing, AI Builds Enterprise Software — all show age-matched ratios ≤ 1.0.

Theme	n pre	n post	Sum×	Median×	Age-matched×
LLMs for Legacy Back Office	104	136	0.19×	0.14×	1.00×
AI Builds Enterprise Software	60	135	0.87×	0.28×	1.00×
ML for Robotics	11	39	0.95×	1.00×	1.00×
US Manufacturing	8	32	0.49×	1.00×	1.00×
Commercial Open Source	34	32	0.33×	0.16×	0.58×
Foundation Models for Biology	15	23	0.66×	0.13×	0.78×
Devtools (Internal-inspired)	12	10	0.11×	0.08×	1.00×
Enterprise Glue	8	8	0.54×	0.16×	1.00×
New Space Companies	8	7	0.09×	0.02×	0.79×

Takeaway

RFS amplification, if it exists, isn't doing it by making post-launch founders raise bigger checks. It's recruiting more founders into the theme (volume) and routing them through the same age-driven funding ladder as every other YC company. The announcement isn't an upgrade; it's a tag.

6. Per-batch curve — younger batches sit at the SAFE floor

W22 → S23 batches show medians of $4-11M; W24 onward median collapses to $500K. The 'low median for new batches' is mechanical, not a quality drop.

Plotting median raise by batch reveals a clean staircase: older batches (W22-S23) have closed Series A/B and show medians in the $4-11M range. Anything from W24 onward is still in YC SAFE territory because there hasn't been enough time. The p25-p75 band collapses to a single line at $500K for recent batches.

Takeaway

Anyone benchmarking recent batches against older ones is comparing 'companies that have had time to close 2 rounds' against 'companies that have closed 1'. Use the survival funnel (s4) instead — same age cohort comparison, no batch-age artifact.

7. The top 15 — who actually made it

These are the 15 companies carrying the weight of the entire portfolio's sum-raised. Six unicorns, vertical AI plays in legal/fintech/insurance/healthcare/banking. Notice the batches — Legora is W24, Phind/Magic S22, Tennr W23. The biggest outliers are not from the youngest batches (no time) nor the oldest (those companies have either crossed into late-stage with bigger valuations or quietly died).

Bottom line

The naive funding view says 'YC raised $8.83B, RFS announcements correlate with waves'. The robust view says: 12 companies hold 39% of that capital; the typical YC company sits at the $500K SAFE floor for 1-2 years; whether you raise more after that is driven by your age and by being in the survivor cohort, not by which RFS tag you carry. If you're investing in YC follow-on, this means: (1) base rate to $5M is 11% in year one, climbing to 63% by year 2.5-3; (2) RFS tag is not a signal of bigger checks per company; (3) sums comparing themes are reporting on whether you got lucky and caught a unicorn in your cohort. Plan accordingly.