YC’s directory grew from a single small batch into a broad, multi-cohort program.
Across the displayed universe, the annual directory total rose from 8 companies in 2005 to 765 in 2026—about 96×.
Scale changes what an aggregate can show. An annual count can describe the breadth of a visible directory; it cannot establish how many companies raised capital, received an investment, or became successful.
From eight companies in 2005 to 765 in 2026.
Annual totals aggregate all named batches in each calendar year. The 2024–2026 years include more than two batches.
The apparent 2021 peak and subsequent annual variation are directory counts, not a measure of investment volume or company quality.
Recent cohorts are more explicitly AI-tagged and more B2B-oriented in the snapshot taxonomy.
The shift is visible in the source labels, but those labels are inputs to analysis—not a substitute for examining individual companies.
The explicit AI tag became common in recent cohorts.
Share of companies carrying exactly YC’s AI or Artificial Intelligence tag, among records with tags.
The 2005–2010 tag coverage is 48.1%; its 3.0% tag share is shown, not extrapolated. “AI” here describes the source taxonomy, not an investment thesis or a measure of technical capability.
Read the shift one era at a time.
Select a cohort to inspect aggregate size, taxonomy coverage, location, and industry mix.
- B2B61.2%
- Consumer7.2%
- Fintech7.0%
- Healthcare8.5%
- Industrials12.3%
- Other3.8%
First-listed San Francisco city share fell from 60.2% in 2005–2010 to 34.1% in 2022–2023, then measured 65.8% in 2024–2026. This is a current-location field, not a record of where companies started.
Non-US-only records expanded, then receded in the newest cohorts.
Share of US-only or non-US-only records that are non-US-only.
The country rule compares US-only and non-US-only records, excludes mixed and unknown assignments, and does not measure headquarters history or founder nationality.
As cohorts widened, the case for batch exposure became less about choosing one theme correctly.
More companies, a shifting mix of categories, and incomplete outcome evidence make a narrow selection thesis harder to infer from a directory alone.
For an investor seeking YC exposure, the public record supports a portfolio construction observation: a broad batch can hold multiple expressions of a changing platform shift without requiring a single company, sector, or geography call to carry the result. It does not identify winners or establish expected returns.
The directory’s current status mix and incomplete listing history reinforce that distinction. They are useful for framing breadth and time horizon, not for calculating an index return, exit rate, or future liquidity.
A directory label is descriptive, not a realized-outcome ledger.
The source preserves a current label for current records. It does not reveal ownership, returns, or what happened to a record that later disappeared.
Most records in the snapshot are marked active.
Current status labels across the 6,299-company named 2005–2026 universe.
- Active69.0%
- Inactive17.4%
- Acquired13.3%
- Public0.4%
These are current directory labels among current directory survivors. They are not an exit-rate calculation and should not be treated as investment outcomes.
Seven public-market events meet the primary-citation rule. That is evidence, not an estimate.
The retained ledger admits an event only when a primary source supports the operating-company identity, route, and actual first-trading date.
Do not turn this into a base rate: the evidence set is incomplete by design, and a current Public directory status is never substituted for a documented historical listing event.
Snapshots can show a directory changing. They cannot explain why.
The historical pass begins in August 2024 and samples monthly source revisions, with explicit gaps retained rather than papered over.
Every removal figure is bounded by the available record: the retained history starts on August 22, 2024, samples monthly, and has a no-snapshot interval from February 9 to May 3, 2026.
Primary materials remain accessible; the page stays aggregate-only.
Access Fund retains the reproducible research package in its operations repository. The public page exposes derived aggregates and upstream references, not company-level source records.
Data source: Y Combinator’s public company directory, pinned October 18, 2026.
- 01Pinned YC company directory snapshot
The raw upstream data used to compute these aggregates. This page renders no company-level records from it. - 02YC’s standard deal
YC’s public explanation of its standard investment terms. - 03YC open-source API repository
The upstream repository and history used for the pinned snapshot and survivorship audit. - 04SEC EDGAR primary-source example
An EDGAR filing from the primary-source evidence used in the retained listing-confirmation ledger.
Read the method and limitations
The pinned source is the yc-oss/api commit aa715ce, retrieved October 18, 2026. It contains 6,306 rows; the page scope excludes unnamed or out-of-range batches.
- The directory is a point-in-time source. Current directory labels are not historical outcomes, investment returns, or portfolio performance.
- AI share counts exact YC AI or Artificial Intelligence tags among records with tags. Tag coverage differs across cohorts, so this page shows its denominator.
- San Francisco share uses the first listed city in the current snapshot. It is not a historical headquarters measure.
- Listing evidence is incomplete: seven first-trade events meet the primary-citation rule, while the source page reported 24 historical listings.
- Survivorship history starts in August 2024, uses monthly snapshots, and contains one no-snapshot interval. Directory absence is not evidence of company failure.
Handoff v1 · generator: funds/AF-access-fund-memo/scripts/charts/generate_yc_by_the_numbers_handoff.py · source repository: access-fund/access-fund-ops
Explore the structure and purpose of Access Fund’s Demo Day Funds.