In July 2025, Substack raised $100 million in a Series C at a reported valuation of $1.1 billion, with The Chernin Group among the backers. A large share of the Substack statistics roundups published since then still list the company’s valuation as $650 million.
That figure is from the March 2021 Series B. It was correct for roughly four years and it has been wrong for more than one. It still circulates because Substack statistics posts cite each other rather than primary sources, and nobody checks the date on a number that looks plausible.
So this is a list of numbers with their dates attached, and a short method for not getting caught out by the next one.
The $650 million number is five years stale
Here is the funding history as it actually stands. Substack raised approximately $90.2 million across five rounds through early 2021, including a $65 million Series B in March 2021 that set the $650 million valuation. Then in July 2025 it raised $100 million more at a reported $1.1 billion, a round covered at the time by Bloomberg and Variety and backed by investors including The Chernin Group and backers of Skims.
The gap between those two valuations is the interesting part, and it is where a stale citation does real damage to an argument. Someone building a case that the paid-newsletter model plateaued can point at $650 million and a flat line. Someone building the opposite case can point at a 69% valuation increase over four years during a period when most consumer media companies were marked down heavily. The same company, two stories, and the one you tell depends entirely on which year you accidentally stopped reading.
This is not a small error in a trivia list. It is the number that people cite when they are deciding whether to build a business on the platform.
The mechanism that keeps a dead number alive is worth understanding, because it applies to every statistic you will ever borrow. A roundup published in 2023 cites a 2021 figure without a date. A roundup published in 2025 cites the 2023 roundup. By 2026 the number has four apparently independent sources, none of which checked the original, and the volume of agreement reads as confirmation. Nothing was fabricated at any step. The date was simply dropped at the first one.
On the usage side, the figures with the firmest sourcing are these. Roughly 35 million total active subscriptions as of September 2025. More than 5 million of those paid, same period. More than 20 million monthly active subscribers as of May 2025. And 47.6 million unique website visitors in September 2025.
Read the paid ratio rather than the headline. Five million paid against thirty-five million active subscriptions puts conversion at roughly one in seven. For a platform where every publication is free to read by default and payment is entirely voluntary, that is a strong number, and it is the single most useful figure on this list for anyone modelling their own newsletter.
Be careful with the overlap in those usage figures, though. Total active subscriptions counts subscription relationships, not people, and one reader following twelve publications contributes twelve. The 20 million monthly active subscriber figure counts humans. Treating the 35 million as an audience size overstates reach by a wide and unknowable margin, and the two numbers are regularly presented side by side as though they measured the same thing.
The creator economics are where the data thins out and the spin thickens. More than 17,000 writers earn money on the platform. Fifty-two newsletters earned $500,000 or more annually as of January 2025, generating at least $40.2 million between them. The top ten authors collectively earn around $40 million a year as of October 2025, with Heather Cox Richardson’s Letters from an American reported at roughly $5 million annually. Substack takes 10% of subscription revenue, and Stripe takes 2.9% plus $0.30 per transaction with about 0.5% on recurring payments, putting the effective total take near 13.5%.
Note the asymmetry in how those creator figures are sourced. The top-end numbers come from company announcements and press interviews, because success stories are promotable. The distribution underneath comes from nothing, because no platform publishes its median. Any time you see a creator-economy statistic, ask whether the party that released it had a reason to want it known. That question alone explains most of what is missing from the public record.
What is the Citation Half-Life?
The Citation Half-Life is the period after which a published statistic is more likely to be wrong than right, and it varies enormously by what the number measures.
Funding and valuation figures have a short half-life, roughly the interval between rounds. For a growth-stage company that is eighteen months to three years. Any valuation older than the company’s most recent raise is simply incorrect, and there is no interpretation that rescues it.
User and subscriber counts have a medium half-life, maybe twelve to eighteen months for a growing platform. They do not become false so much as they become uninformative, and the direction of the error is predictable: a growing platform’s old numbers understate it.
Ratios have the longest half-life of all. The one-in-seven paid conversion figure, the 13.5% effective platform take, the shape of the earnings distribution: these change slowly because they reflect structural features rather than momentary scale. If you are going to cite one number from a post like this in two years, cite a ratio.
Policy and pricing figures sit in their own category, with a half-life that is unpredictable rather than long or short. A platform fee can stay fixed for six years and then change in a week. These are the numbers to re-check before every use rather than on a schedule, because unlike a subscriber count they move in jumps and the jump is never pre-announced in the roundups.
The method follows directly. For every statistic you plan to use, find the date it describes, not the date it was published. Those are routinely years apart, and the publication date is the one that makes a stale number look fresh. Then ask which category it falls into. A valuation with no date beside it should be treated as unusable rather than approximately right.
Build the date into how you store numbers and the discipline takes care of itself. A research note that reads “5M paid subs (Substack, Sept 2025, co. announcement)” survives being pasted into a deck two years later, because the staleness is visible to whoever opens it. A note that reads “5M paid subs” is a trap you set for your future self, and you will walk into it.
Two specific habits catch most problems. Trace the number to a primary source: a company announcement, a regulatory filing, or original reporting by an outlet that did the work. And when a figure appears in five roundups with identical phrasing, treat that as evidence of copying rather than corroboration, because it usually is.
There is a third habit that matters more as research moves to AI assistants. Ask the assistant for the date and the primary source alongside the figure, and treat an answer that cannot produce either as unverified. These systems are trained on the same roundups that dropped the dates, so they reproduce stale numbers fluently and confidently. The confident tone is not evidence. It is the same copying problem with better grammar.
Cite these numbers, and cite their dates
Pulled together, the fourteen Substack statistics worth having at hand, each with what it describes and when.
The $1.1 billion valuation and the $100 million Series C, both July 2025. The $650 million valuation and $65 million Series B, both March 2021, useful only as historical comparison. The approximately $90.2 million raised across five rounds before the Series C.
Thirty-five million total active subscriptions and more than 5 million paid, both September 2025. More than 20 million monthly active subscribers, May 2025. The 47.6 million unique monthly website visitors, September 2025.
More than 17,000 writers earning money. Fifty-two publications above $500,000 a year and at least $40.2 million in combined revenue among them, January 2025. Around $40 million annually for the top ten authors combined, October 2025. Roughly $5 million a year for the single largest publication. And the 10% platform fee, which with payment processing lands near 13.5% effective.
Write the date into the sentence rather than into a footnote. “Substack passed 5 million paid subscriptions in September 2025” ages gracefully into a true historical statement. “Substack has 5 million paid subscriptions” becomes false the moment the number moves, and it will.
This habit also protects you in a way that is easy to miss. A dated claim that later turns out to have been misreported is a sourcing error you can correct cleanly. An undated claim that goes stale looks like carelessness, because the reader cannot tell whether you were wrong or merely late. One extra clause buys you that distinction permanently.
Which Substack figures actually tell you something?
Most of the list above is context. Three of the numbers are genuinely decision-relevant, and they are not the ones that get quoted in headlines.
The paid conversion ratio, around one in seven active subscriptions, is the realistic ceiling for a publication that does what the platform is designed for. If your model assumes you will convert a quarter of your readers, you are assuming you will outperform the aggregate of every publication on the platform by a wide margin. Possible, but it should be an explicit assumption rather than an accident.
The earnings distribution is the number that should change behavior most and almost never does. Fifty-two publications above $500,000 against more than 17,000 writers earning money means roughly three tenths of one percent reach that level. The top ten taking $40 million between them makes the concentration sharper still. This is a power-law market, and the median outcome is nothing like the examples used to sell the model. Note also what the public data does not say: how many of the 17,000 earn enough to replace a salary. That figure is not published, which is itself informative.
The 13.5% effective take is the one to run against your own arithmetic before you commit. On a $10 monthly subscription it is about $1.35, and whether that is cheap depends entirely on what you would otherwise pay for hosting, payments, email delivery and discovery. For a writer starting from zero it is usually a bargain. For an established publication with its own audience and infrastructure, it is often the most expensive line item in the business.
So the tactical next step is small and worth doing today. Take whichever Substack statistic you have been repeating in decks or pitches, find the date it actually describes, and check it against a primary source. If it turns out to be the $650 million valuation, you now know what to replace it with, and you have learned something useful about where the rest of your numbers came from.