Common Crawl revenue: $1,467,641 in FY2024, and the 13 donors nobody has named

Quick answer (October 2026). The Common Crawl Foundation reported total revenue of $1,467,641 in FY2024, $1,297,813 in FY2023, $451,447 in FY2022, $330,010 in FY2021 and $75,014 in FY2020. FY2024 is the most recent year on file: the return was signed and filed on November 7, 2025, and no later return exists. Expenses were $1,326,638, leaving a surplus of $141,003 and net assets of $1,472,532 against total liabilities of $0. Those totals are already published elsewhere and they are correct. The part that is not published anywhere is who paid. Thirteen donors are named in the foundation's own filing for FY2024 and eight for FY2023, and the two largest corporate givers in both years were Anthropic at $250,000 and OpenAI at $250,000, a combined $500,000 that matched the founder's family foundation to the dollar in each year.

Why the donor names are public at all

Almost every write-up calls Common Crawl a nonprofit, which is true, and stops there. The precise classification matters more than it sounds.

Common Crawl Foundation, EIN 26-1635908, does not file a Form 990. It files a Form 990-PF, the return of a private foundation, and it has done so for every year on record. More specifically it is a private operating foundation under section 4942(j)(3), a status the IRS granted it on September 23, 2009 and which the foundation re-asserts on every return. An operating foundation runs its own programme instead of making grants, which is exactly right for a body whose entire output is a monthly web crawl. Contributions, gifts and grants paid out were $0 in both FY2023 and FY2024.

The consequence that matters here is procedural. A public charity attaches a Schedule B listing its substantial donors, and the IRS redacts the names before the return reaches the public. A private foundation gets no such redaction. Its Schedule B is part of the public return, names and all.

So the donor list has been sitting in a public document since November 13, 2024 for FY2023 and November 7, 2025 for FY2024. The figures below are read directly from the filed XML that the IRS publishes in its Form 990 series bulk downloads, cross-checked against the extracted summary on ProPublica's Nonprofit Explorer page for the foundation.

Who paid for Common Crawl in FY2024

Thirteen contributors, $1,457,500 between them. Names are reproduced exactly as the return spells them, including its typographical slips.

Line 1 of the return reports $1,467,543 of contributions received, so $10,043 came from givers below the $5,000 Schedule B threshold. The remaining $98 of revenue is bank interest.

FY2023 had eight contributors and $1,287,500 listed, with $10,251 unlisted.

Four names appear in both years: the Elbaz Family Foundation, Anthropic, OpenAI and Civic Ventures. Nine are new in FY2024 and four did not return, among them DuckDuckGo and the Allen Institute for AI.

The arithmetic that is hard to unsee

In FY2023, Anthropic gave $250,000 and OpenAI gave $250,000. Together: $500,000. The Elbaz Family Foundation gave $500,000.

In FY2024, Anthropic gave $250,000 and OpenAI gave $250,000. Together: $500,000. The Elbaz Family Foundation gave $500,000.

Two years running, the two frontier model developers matched the founder's family foundation exactly, to the dollar. That is a coincidence of round-number giving rather than anything sinister, but it is the single clearest way to size the relationship, and it is not a number anybody has published.

Against revenue, the two labs are 38.53% of FY2023 and 34.07% of FY2024. Widen the definition to every named donor that builds or ships models and FY2023 reaches $525,000 or 40.45% of revenue, and FY2024 reaches $545,000 or 37.13% once Hugging Face and Microsoft are counted.

What the foundation says, and what the return says

On November 4, 2025, three days before it filed the FY2024 return, Common Crawl published a response to an article in The Atlantic. Executive Director Rich Skrenta wrote, under a heading on independence and funding:

Common Crawl, November 4, 2025: For over fifteen years, we were supported almost entirely by the Elbaz Family Foundation Trust. In recent years, as public and research interest in large-scale text analysis has grown, several organizations - including some AI companies - have contributed donations to support the cost of running and maintaining our public archive. These donations represent a small fraction of our overall operating needs and are disclosed publicly in our financial statements.

Every clause of that is defensible except one, and the testable clause is the phrase a small fraction of our overall operating needs. The phrase has no fixed denominator, so here it is against all four reasonable ones. The most recent return available when the statement was published was FY2023.

Two gifts covered 83.31% of what the foundation actually spent in FY2023. On no reading available in the filings is that a small fraction, and the FY2024 return filed three days later added Microsoft and Hugging Face rather than reducing the exposure.

Three things should be said in the foundation's favour, because they are all true and all checkable. The fifteen-year claim is not something this piece can test: the IRS publishes machine-readable returns only for recent years, and the FY2022 and earlier Schedule B donor lists are not in the bulk XML, so who funded Common Crawl before 2023 remains unestablished here. The statement that the donations are disclosed publicly in our financial statements is literally true, since the Schedule B is a public document. And the accompanying claim that No donor, corporate or otherwise, has any control over what we collect, publish, or remove is contradicted by nothing in either return: every self-dealing and control question in Part VI-B is answered false in both years.

Collaborators and contributors are two different lists

Common Crawl's about page carries a Collaborators list of 28 organisations. Six of the names on it are, per the filing, also donors: the Allen Institute for AI, Andreessen Horowitz and Hugging Face appear verbatim, Constellation appears as the Constellation Network that the foundation separately profiled in an April 9, 2025 blog post about blockchain provenance, and Computer and Communications Industry Association and Kearney match the filing's abbreviated Computer Communications of Washington DC and AT Kearney of Chicago. Those last two are a name-and-city match rather than a certainty, and the return gives no further identifiers.

What the page does not do is say that any of them gave money. Its entire funding disclosure is one sentence:

commoncrawl.org, about page: Gil Elbaz serves as Chairman of the Board. Primary funding comes from the Elbaz Family Foundation.

That sentence is accurate. The Elbaz Family Foundation is the largest single donor in both years. It is also, in FY2024, 34.07% of revenue, which leaves 65.93% described nowhere on the site.

And the two biggest corporate donors are not on the Collaborators list at all. Across the homepage, the about page, the FAQ, the team page and the collaborators page, Anthropic appears zero times. OpenAI appears twice, both times on the about page as the developer of GPT-3, a model trained on the archive, never as a funder. Microsoft appears exactly once in the visible text of all five, in an FAQ question asking whether Google or Microsoft could simply do what Common Crawl does. The answer given there is that the goal is to democratise the data so that everyone, not just big companies, can do high-quality research. Microsoft gave $20,000 in FY2024.

FY2024 is the year Common Crawl got a payroll

The two returns describe what look like two different organisations.

In FY2023 the return lists three unpaid directors at two hours a week each, Nova Spivack, Gilad Elbaz and Carl Malamud, and answers the five-highest-paid-employees question with the single word NONE. The organisation spent $600,149 that year and 77.26% of it went through one line called Consulting.

In FY2024 the same section lists Richard Skrenta as Executive Director at 40 hours a week on $179,211, Greg Lindahl as CTO at 40 hours a week on $173,283, Michael Birnbach as CFO at 10 hours a week on $34,075, and Gilad Elbaz and Eva Ho at $0. Jennifer English appears separately as Program Manager on $89,490. The three paid officer figures sum to exactly $386,569, which is the Part I line 13 total, and the one employee figure is exactly the line 14 total, so the compensation block foots without residue.

One detail worth not over-reading. Common Crawl's own history says Rich Skrenta became Executive Director in 2023, and the FY2023 return shows no officer compensation and no employee above $50,000. The return does not establish that he was unpaid, only that he received no officer compensation and was not among the employees paid more than $50,000 in calendar 2023. The highest-paid contractor that year was Sebastian Nagel of Konstanz, Germany, at $130,517 for software development; in FY2024 that slot is Thom Vaughan of Beverly Hills at $84,689 for the same service. Both men are on the current team page, Nagel as Distinguished Engineer and Vaughan as Principal Engineer, so neither departed.

The one-dollar tax bill

Common Crawl holds its money in cash. Total liabilities were $0 at the end of both years, and it files on the cash basis, so there are no receivables, no payables and no deferred revenue to net off a reserve figure. Its entire investment income is bank interest: $98 in FY2024 and $62 in FY2023.

A private foundation pays an excise tax of 1.39% on net investment income. On $98 that is $1.36. The foundation's reported tax for FY2024 is $1, and for FY2023 it is also $1. The body that supplies the raw material for most of the generative AI industry has a federal tax bill of one dollar, and the arithmetic is exactly right.

The distribution requirement is similarly unstressed. Minimum investment return for FY2024 is $47,859, being 5% of $957,174 of non-charitable-use assets after the 1.5% cash allowance. Qualifying distributions were $939,130, or 19.62 times the minimum. As an operating foundation it is tested instead on an income test that requires 85% of the lesser of adjusted net income and minimum investment return, which here is 85% of $98, or $83. It is not close to failing anything.

Reserves look comfortable and are shrinking. Net assets of $1,472,532 against FY2024 spending of $1,326,638 is 405 days of cover, down from 810 days a year earlier, because expenses grew 121.05% while revenue grew 13.09%. No restricted and unrestricted split is available: the foundation checks the box for bodies that do not follow FASB ASC 958 and reports the alternative equity lines instead.

A cash-flow statement the form does not ask for

Form 990-PF requires no statement of cash flows, but two balance sheets plus a cash-basis surplus are enough to derive one, and it is a useful check on whether anything unexplained is moving.

FY2024. Surplus $141,003, plus depreciation of $939 which is not a cash cost, less $5,636 spent on computer equipment in February 2024, gives $136,306. Cash went from $1,331,529 to $1,467,835, a rise of $136,306. Exact.

FY2023. Surplus $697,664 plus $4,692 of amortisation gives $702,356. Cash went from $629,173 to $1,331,529, a rise of $702,356. Exact.

That $4,692 has a nice ending to it. It is the last tranche of start-up costs the foundation capitalised on July 1, 2008 and amortised over fifteen years. FY2023, the year the AI boom made Common Crawl famous, is the year its founding expenses finished running off the books.

One place the return disagrees with itself

Part I of a 990-PF has four columns. Column (a) is revenue and expenses per books; column (d) is Disbursements for charitable purposes, and Part XI takes its total straight from there to compute qualifying distributions, under a heading reading Amounts paid (including administrative expenses) to accomplish charitable, etc., purposes.

In FY2024, column (d) is blank on line 13. The $386,569 of officer compensation is reported per books and is not counted as a charitable disbursement, so the column (d) total comes to $939,130 rather than $1,325,699.

The same return then fills in Part VIII-A, Summary of Direct Charitable Activities, and puts the foundation's largest direct charitable activity at $1,325,698, describing it in the foundation's own words as working towards an open web that allows open access to information and enables greater innovation in research, business and education. That figure is column (d) plus the officer compensation, less one dollar, the dollar being the excise tax, which is not a charitable activity. FY2023 shows the same one-dollar relationship: column (d) is $595,457 and Part VIII-A is $595,456.

So one part of the return says the foundation spent $939,130 on its charitable purpose and another says $1,325,698, and the gap is the salaries of a full-time Executive Director and a full-time CTO at an operating foundation whose programme is the crawl they run. The form's own instructions anticipate columns (b), (c) and (d) not tying to column (a), so this is not a filing error, and nothing turns on it for tax: the foundation clears its distribution test either way. It does affect what a reader sees. ProPublica's summary prints Charitable Disbursements of $939,130 at 70.8% of expenses and Compensation of Officers at 29.1% beside it, with a footnote warning that officer pay may be included in charitable disbursements. In this case it is not, and a reader taking the printed 70.8% at face value would conclude that the other 29.2% of Common Crawl's spending went somewhere other than the programme.

What the existing answers get right

The best page currently answering this question is a long ABAB News piece from April 18, 2026. It is accurate. It carries the full revenue series through FY2024 including expenses and net assets, it correctly describes the governance handover from Spivack, Elbaz and Malamud to Elbaz, Eva Ho, Birnbach and Skrenta, and it quotes both the about page and the November 2025 response. Its only limitation is the one this piece exists to fill: it sources the funding question to what Common Crawl says about itself, and never opens the Schedule B.

Wikipedia's Common Crawl article carries no revenue figure at all, so it is silent rather than stale, and its one funding sentence is right in substance: it says the foundation was funded almost exclusively by the Elbaz Family Foundation Trust until 2023, when it started receiving donations from the AI industry. Like everyone else, it names no donor and no amount.

For a smaller body where the money arrives by a completely different route, the Bytecode Alliance Foundation is the instructive contrast: a 501(c)(6) whose revenue is 100% membership dues, with zero employees and eight unpaid directors.

What this piece could not establish

Three things, stated so nobody assumes otherwise. The pre-2023 donor mix is unknown here, because the IRS bulk XML does not reach back far enough and the FY2022 index entry carries no retrievable archive. Whether the Schmidt Foundation of San Francisco is one of the better-known Schmidt philanthropies is not established; the return gives a name and a ZIP code and nothing else, and the same caution applies to Skye Global AI and to Peter Zhang. And the identification of Computer Communications with the Computer and Communications Industry Association rests on a name and a city, not on anything in the filing.

Nothing in either return suggests wrongdoing, and nothing here is an allegation of any. The foundation's figures foot, its tests pass, its disclosures are filed on time and its donor list is, in the strict sense, public. The finding is narrower and duller than a scandal: the organisation at the centre of the argument about who owns the text that trains language models is funded, to roughly a third of its income, by the companies training those models, and the only place that is written down is a tax form.