Sources
Every figure on this site traces to a filing or a public-records disclosure you can open yourself. This is the list — 783 documents from 63 publishers — with the period each covers and the most recent one we have read. 83% of the figures we publish open the filing they came from, 0.3% at the exact page or row; 17% still point at a publisher’s reports page while we find the document. Where a source has a gap or a caveat, the page that uses it says so.
How far a click goes
Every figure on a fund or sponsor page is a link to its source. How precise that link can be depends on what the publisher gave us.
| A click opens | Figures | Share |
|---|---|---|
| At the page or row | 1,505 | 0.3% |
| The document | 462,468 | 82% |
| A publisher’s reports page | 96,960 | 17% |
| Not linked — leaked or unverified, excluded | 102 | — |
Shares are over the 560,969 figures that link somewhere (including 36 from a manager’s own publication). The figure is recomputed after every load.
The universe — SEC datasets
These define how many private funds exist and how big they are. They are the denominator behind every “share we track” figure.
Form ADV Schedule D 7.B.(1) — every private fund, every filing
Fund name, private-fund id, type, gross asset value, investors, master/feeder structure, and the adviser behind each.
Filings from 2011-11-08 to 2026-07-31. Historical bulk CSVs through 2024-12-31, then one file per month from adviserinfo.sec.gov.
Form D — every private offering
What each fund sold and when, dated at first sale, collapsed to the latest amendment of each offering.
4,324 filing dates across the quarterly datasets.
Sponsors — managers who file their own numbers
Publicly traded managers report fund-level performance in their SEC filings. These are the fund’s own figures, typically gross of fees.
| Manager | Documents | Covering | Latest |
|---|---|---|---|
| KKR · 10-K, 10-Q | 62 | 2010-12-31 | 2026-06-30 |
| Blackstone · 10-K, 10-Q | 51 | 2012-06-30 | 2026-06-30 |
| Ares · 10-K, 10-Q | 50 | 2014-03-31 | 2026-06-30 |
| EQT · half-year report | 27 | 2018-12-31 | 2026-06-30 |
| Blue Owl · 10-K, 10-Q | 21 | 2021-06-30 | 2026-06-30 |
| TPG · 10-K, 10-Q | 19 | 2021-12-31 | 2026-06-30 |
| CVC Capital Partners · half-year/annual report | 5 | 2024-06-30 | 2026-06-30 |
| Thrive Capital · manager_letter_leaked | 1 | 2026-06-30 | 2026-06-30 |
| Apollo · 10-K, 10-Q | 61 | 2011-03-31 | 2026-03-31 |
| Carlyle · 10-K, 10-Q | 57 | 2012-03-31 | 2026-03-31 |
| Andreessen Horowitz · manager_publication | 1 | 2025-09-30 | 2025-09-30 |
| Founders Fund · reported_unverified | 1 | 2024-12-31 | 2024-12-31 |
| Oaktree Capital Management · 10-K, 10-Q | 13 | 2016-09-30 | 2019-09-30 |
Public pensions — investors who disclose their positions
Public-records disclosures of what each plan holds. These are one investor’s position, net of fees — the dollars are that pension’s stake rather than the size of the fund, though the multiples are ratios and compare directly.
How to read particular sources
Terms a source sets on its own numbers that you could not infer from the numbers. We follow them, and say so here rather than quietly.
| Source | What to know |
|---|---|
| The Regents of the University of California | Two things this table does not say about itself. (1) It covers ACTIVE commitments only — “it doesn’t include any exited partnership investments” — so fully realised funds drop out and the sample is survivorship-filtered. (2) In the reports prepared by Cambridge Associates (through March 2017), the figures for Accel, Kleiner Perkins Caufield & Byers and Sequoia Capital were FROZEN at 31 March 2003 and reprinted unchanged for a decade, because those managers stopped providing fund-level data. We date those rows 2003-03-31 rather than by the report that carried them, so one stale figure stays one observation instead of drawing a flat line across ten years. UC’s post-2017 figures are its own and current. |
Manager-published figures
Neither a filing nor a public record: a manager publishing its own numbers. Published, attributed and dated, but unaudited and sometimes a subset the manager chose — so weigh them differently. We include them where a major firm appears in no disclosure we can otherwise reach.
| Source | What it is |
|---|---|
| Andreessen Horowitz — “a16z: The Power Brokers” | The manager’s own published table of nine funds, as of 30 September 2025. Its disclosure states the figures are unaudited, and the table is headed “Select” and excludes a16z Bio funds — so it is a subset the manager chose, not a complete record. Included because a16z is among the largest venture managers in the country and appears in no public pension disclosure we have found. |
Leaked documents
A manager’s own figures that the manager did not publish. Not a filing you can open, and not the manager choosing to speak — so they are named here rather than mixed in with either. We think fresh numbers on a major firm are worth having; you may weigh them differently, which is the point of saying so.
| Source | What it is |
|---|---|
| Thrive Capital — Q2 2026 investor letter | Fund-by-fund gross and net multiples, DPI and IRRs for twelve Thrive funds as of 30 June 2026. Thrive sent this letter to its limited partners and published nothing; TechCrunch reports it was “leaked to Bloomberg”, and the chart we transcribed circulated via the Pari Passu newsletter. Newcomer had already run Thrive’s fund-level figures as a scoop in February 2026. So these are the manager’s own numbers, unaudited, that you cannot go and check at the source — which is why they sit here rather than beside the filings. A thirteenth fund, Thrive VI Growth, is NOT loaded: the chart’s labels read 4.3x gross against 4.4x net, which cannot be true, so the row was held back rather than guessed. |
| Founders Fund — returns as of 2024 | Ten funds with net IRR, net multiple and DPI, from Founders Fund I (2005, 7.8x) through the 2023 growth fund. THE WEAKEST-SOURCED TABLE HERE, and labelled so. It circulated as an image posted to X in March 2025, widely attributed to TechCrunch — and we could not find the TechCrunch article containing it. TechCrunch’s known piece on the firm (4 March 2022) was checked directly and publishes no fund-level performance at all. The manager published nothing. With no cash flows to reconcile, the only check possible is that DPI cannot exceed the net multiple, which catches a transcription slip and not a wrong figure. Included because historic returns for the top venture franchises are worth having; weigh it accordingly. |
Checked and rejected
Sources we opened and did not use. Publishing this is deliberate: an absence should read as a decision rather than an oversight, and it stops the same dead ends being probed twice.
| Source | What it actually is |
|---|---|
| Virginia Retirement System | 340-page annual report with asset-class returns only — no fund-level table. |
| New York State Common Retirement Fund | Monthly reports announce NEW COMMITMENTS; they carry no marks. |
| Texas Teacher Retirement System | 94-page board book of asset-class and benchmark performance, no fund table. |
| Massachusetts PRIM | 119-page annual report. Vintage years and programme-level returns, no fund-level table. |
| New Mexico SIC | Posts an Investment Holdings Report that is worth another look; not yet a marks source. |
| San Francisco Employees' Retirement System | 68-page board deck reporting performance by STRATEGY only — venture, growth, private equity total — with no fund-level rows anywhere. |
| Alaska Permanent Fund | Publishes quarterly private equity reports with fund commitments and vintage years, but Alaska statute AS 37.13.200 forbids disclosing individual investment RESULTS — only asset-class aggregates. So it can tell us which funds it holds and never how they performed. The bar is a law, not a formatting problem, so this will not change. |
| Norway (NBIM) | The most transparent fund in the world and structurally empty for us — it holds no private equity at all, only listed equity, fixed income, real estate and renewable infrastructure. |
What we will not use
| Kind | Why not |
|---|---|
| Preqin, PitchBook, Burgiss and other commercial datasets | We never seed, backfill or republish a commercial dataset. Where one of them PUBLISHES an industry aggregate we cite it beside ours and explain the difference — that is comparison, not sourcing. |
Missing a source you think we should read? There is a reporter at the bottom of every page.