Twitch will not tell you how many subscribers a channel has. Not you, not a brand, not the trackers that publish the leaderboards. Every public Twitch subscriber number you have ever read, including every number in this article, is somebody's estimate of a figure only the streamer can actually see.
That one fact shapes everything below. It decides who gets called the record holder, who leads today, what a subscription really pays, and why a sponsorship priced off a sub count is priced off a guess. So we went to the record book, the live ranking, Twitch's own payout terms, and the brand campaigns that bought subscriptions by the hundred thousand.
Who has the most Twitch subscribers right now?
On the live subscriber ranking we read in September 2026, ironmouse led with 108,857 active subscriptions. Here is the whole top ten as that page had it:
- ironmouse, 108,857
- Cinna, 100,199
- PlaqueBoyMax, 98,697
- Jynxzi, 91,188
- caseoh_, 67,262
- KaiCenat, 55,401
- HasanAbi, 45,438
- Caedrel, 44,634
- stableronaldo, 37,382
- Joe_Bartolozzi, 31,731
Now the part that matters more than the order. Read the tracker's own description of what it is counting: "This list ranks channels based on the number of subscriptions gained from August 12th to the present day." That is not a standing total. It is a rolling window of roughly the last month, which happens to be about as long as a single subscription lasts. The same page adds that rankings are "updated daily" and warns that values "may have a margin of error".
We read that page twice in one day. The first read had Jynxzi first with 119,705 and a window starting 11 August. Hours later the window had rolled to 12 August, ironmouse was first, and Jynxzi had dropped nearly 30,000. Nobody gained or lost 30,000 subscribers over lunch. The window moved.
Our take: treat the phrase "most subscribed Twitch streamer" as a headline about the last 30 days, not a statement about the size of a community. A number that moved by nearly 30,000 between two reads of the same page on the same day is not a number to build a media plan on.
Who holds the all-time Twitch subscriber record?
Kai Cenat. Guinness World Records lists 1,095,265 subscribers on his channel, achieved on 30 September 2025, the last day of his Mafiathon 3 subathon, and notes that he is "the first Twitch channel to reach 1 million subscribers".
A subathon is a marathon broadcast, usually around the clock for weeks, where incoming subscriptions extend the timer. Almost every record below was set during one.
Here is the wrinkle nobody prints. For that same night, Streams Charts publishes a peak of 1,112,858. The adjudicated record and the tracked record are 17,593 subscriptions apart, and the official one is the lower of the two. A certified record is a verified moment. A tracker peak is a poll.
How the record climbed from 269,154 to a million
- 2018: Ninja peaked at 269,154 subscriptions, the first widely recognised holder.
- April 2021: Ludwig Ahgren passed him during a 30 day subathon. Ninja's reply: "Records are meant to be broken, I would be lying if I said wasn't a little sad but congrats @LudwigAhgren on holding the new sub record on twitch".
- March 2023: Kai Cenat ended a 30 day run with more than 300,000 concurrent subscriptions, passing Ludwig's mark of more than 283,000. He called it the Mafiathon.
- September 2024: ironmouse, a VTuber who streams as an animated avatar, took it with 320,478 on day 30 of a charity subathon, half the earnings pledged to the Immune Deficiency Foundation.
- November 2024: Kai Cenat reclaimed it with Mafiathon 2, peaking at 728,525 and gaining 659,300 subscriptions in a month.
- September 2025: Mafiathon 3 crossed a million. Snoop Dogg phoned in and gifted 50,000 subs, telling him "I'm finna get you 50,000 so you can get over the record".
One more distinction worth keeping straight: most subscribed is not most watched. The Guinness record for concurrent viewers on a Twitch stream is 3,846,256, set by Ibai Llanos at his La Velada boxing event in July 2024. Different metric, different record, different streamer.
What happened to Kai Cenat's subs after the record?
It collapsed by roughly 96 per cent in four months. In February 2026, after more than four months without going live, Kai Cenat held "just under 40,000 subscribers after losing 11,000 in the last month", down from a certified 1,095,265. On our September 2026 read he was back to 55,401, still sixth.
The money shrinks the same way. When Ludwig's 2021 record was totted up, the gross was about $1.6 million. After Twitch's cut it was $935,601. After $365,350 to charity and $177,000 to moderators and staff it was $393,251, and after California tax it was $206,928.68. Roughly $207,000 of a $1.6 million headline reached his bank account.
Our take: a subathon record is an event, not an audience. Price the event if you are buying the event. Never price a year of partnership off a number that was true for one night: 1,095,265 in September 2025, under 40,000 by February 2026.
Why can nobody actually see a Twitch subscriber count?
Because Twitch gates it behind the streamer's own permission. The Get Broadcaster Subscriptions endpoint needs a user access token carrying the channel:read:subscriptions scope, and the documentation is blunt about whose token: "The ID in the broadcaster_id query parameter must match the user ID in the access token." A brand cannot read it. A journalist cannot read it. Only the broadcaster, or someone the broadcaster has handed a token to.
What is public is thin. Live viewer count is public. The follower total is public, though the follower list needs a moderator scope. There is no endpoint that returns yesterday's viewers, which is why every historical chart you have seen is a tracker's own polling history rather than Twitch data.
Compare YouTube, where the Data API hands any caller a channel's subscriberCount with no permission from the creator at all, "rounded down to three significant figures". That asymmetry is the whole reason Twitch vetting works differently from every other platform, and we come back to it at the end.
One detail on that same ranking should worry anyone quoting it. The column headers still say "Tier 1 ($4.99)", "Tier 2 ($9.99)" and "Tier 3 ($24.99)". US Tier 1 on the web has been $5.99 since July 2024. Multiply that tracker's columns by that tracker's own labels and your revenue estimate is wrong before you start.
What does a Twitch subscription actually pay the streamer?
Less than list price times a split, and the gap is where most published estimates fall over. Start with what a viewer pays.
What a subscription costs
- United States, web: Tier 1 went from $4.99 to $5.99 in July 2024, the first US increase in Twitch's history.
- United States, mobile app: $7.99 since October 2024, across 40 regions, to absorb Apple's and Google's 30 per cent cut. The web price was untouched.
- Elsewhere: Twitch resets prices market by market. In 2024 it moved the UK from GBP 3.99 to GBP 4.99, Canada from CAD 6.99 to CAD 7.99, Australia from AUD 7.99 to AUD 8.99, and Turkey from TRY 9.90 to TRY 43.90, saying the changes applied "only to subscriptions purchased on desktop or mobile web".
That last point is not trivia. A Tier 1 sub costs roughly $1.25 in Russia and about $0.86 in Ukraine. The same subscription, the same line on a leaderboard, a fifth of the money.
What share the streamer keeps
- The standard share is 50 per cent of net subscription revenue.
- Hold 100 Plus Points for three consecutive months and it becomes "60/40 net revenue on paid and gift subscriptions".
- Hold 300 Plus Points and it becomes 70/30. Twitch cut that threshold from 350, opened the programme to Affiliates, and removed "the US$100K cap on net revenue at the 70/30 revenue share level for all streamers".
- Points come from subscriptions, weighted "Tier 1 Subs = 1 point, Tier 2 Subs = 2 points, and Tier 3 Subs = 6 points".
Note the words Twitch uses every single time: net revenue. When it raised prices it promised streamers "the same net revenue share on subs and gift subs". Net, not list. The share is applied after billing and platform costs, which is exactly why the mobile price is two dollars higher than the web price for the same product.
The part that breaks the maths: Prime
A Prime subscription looks identical on a leaderboard and pays on a completely different basis. Since June 2024 Twitch has paid a fixed rate per Prime sub set by the subscriber's country rather than a share of revenue, a change it said moved payouts by "less than 5% in the vast majority of countries". On our read, Jynxzi had 17,234 Prime subs against 23,968 paid. xQc had more Prime than paid, 9,303 against 7,161. Any model that treats those as paid subscriptions is inventing money.
Everything else on the invoice
- Bits: Twitch's stated rate is "1 cent per Bit used to Cheer", and broadcasters get the same amount however the viewer acquired the Bits.
- Ads: "Both Partners and Affiliates can automatically earn 55% net ad revenue by running at least 3 ad minutes per hour."
- Payout still needs Affiliate or Partner status and a $50 threshold, even though since May 2026 subs, Bits, emotes and badges switch on for eligible streamers globally.
Our take: this is why we will not publish a tidy table of what 1,000 subs pays, and why you should distrust anyone who does. Watch it go wrong in public. In February 2026 a mainstream outlet estimated Kai Cenat was earning $163,527 a month by assuming a 70/30 split and $5.99 for every single subscriber, for a streamer who had not gone live in four months. Its own piece concedes "exact figures for the different tiers of subs aren't available". Four errors in one number: the split is not public, the base is net revenue, a fifth of a typical base is Prime at a flat country rate, and prices are regional.
What does the sub count hide that the mix reveals?
Who actually paid, and whether they meant it. The ranking publishes a breakdown underneath the headline, and it is far more useful than the headline. From our September 2026 read, gifted subs as a share of each channel's paid subs:
- Cinna: 98.0 per cent gifted
- ironmouse: 95.6 per cent gifted
- PlaqueBoyMax: 92.8 per cent gifted
- HasanAbi: 67.3 per cent gifted
- Jynxzi: 32.3 per cent gifted
- caseoh_: 21.2 per cent gifted
- Caedrel: 3.6 per cent gifted
Cinna and Caedrel sit six places apart on the same list. One community bought almost all of its subscriptions for other people during an event. The other is paying for itself. Treating those two as the same signal is the most expensive mistake in Twitch sponsorship planning.
Two more things fall out of the same table. First, the tier mix marketers imagine barely exists: on every row we checked, Tier 1 accounted for at least 98.7 per cent of paid subs, so modelling a blend of Tier 2 and Tier 3 revenue is modelling a rounding error. Second, the headline does not reconcile to its own breakdown. Jynxzi's paid and Prime columns sum to 41,202 against a headline of 91,188. Kai Cenat's sum to 15,226 against 55,401. The top-line number is modelled, not added up.
The extreme case is instructive. In December 2024 one Counter-Strike co-streamer went from under 20,000 subscriptions to 351,665 during a single major tournament, and "over 97% (343,706) of the subscriptions were gifted", bought in a region where Tier 1 costs about a fifth of the US price. That is the second highest peak ever recorded, and it is almost entirely one week of gifts at a discount.
Even the trackers say so. Streams Charts warns readers that in cheaper regions "a streamer appealing to audiences in these cheaper regions may accumulate a lot of subscriptions, but not much profit".
Our take: ask what share was gifted before you ask how many there are. On the list above that share runs from 3.6 to 98.0 per cent. A gift is a fixed term that does not renew itself, so a gifted-heavy peak has a cliff built into it roughly a month later, by design.
Why is September the worst month to read a Twitch leaderboard?
Because Twitch runs a sitewide sale across it. SUBtember 2026 ran from 28 August to 1 October with 25 per cent off one month and three month subs and 30 per cent off six month subs.
Three mechanics inside that promotion distort any September reading:
- It is Twitch's money, not the streamer's. Twitch states plainly that it "covers the difference so streamers will still get the same payout as a full-priced sub".
- It only discounts new business. The promotion applies to first terms of new paid subs, and to Prime and gift subs being converted into paid ones. An existing recurring sub gets nothing, so the sale actively manufactures conversions that would otherwise have happened slowly or not at all.
- It pulls future months forward. A discounted six month sub bought in September does not come back for renewal in October. It lands again in March.
Brands are inside the promotion too. Twitch names State Farm and Minecraft Dungeons II as the 2026 sponsors, with State Farm "gifting nearly 100K subs" across participating channels on the opening day and Minecraft Dungeons II adding one free sub for every five gifted in the closing week. There is even a Bits bonus running 12 to 18 September, adding 10 per cent to Bits spent on Custom Power-ups of 100 or more.
The record calendar follows the sale. Mafiathon 3, the run that produced the record at the top of this article, ended on 30 September 2025, inside that year's window.
Our take: never compare a September leaderboard to any other month. You are reading a month Twitch discounts by 25 to 30 per cent and subsidises itself. Compare October to October, or November to November. If a creator's deck opens with a September screenshot, ask for the November number.
Which brands bought Twitch subscriptions outright?
State Farm, Rockstar Games and McDonald's Germany, and the biggest of those buys is sitting at the top of today's leaderboard. Subscriptions are not only a loyalty metric. They are a media unit you can purchase.
State Farm turned 75,000 subs into a game show prize
State Farm's Gamerhood Season 5 premiered in July 2026 and finished on 28 August 2026. PlaqueBoyMax won, and State Farm "gifted PlaqueBoyMax and his community 75,000 subs as his grand prize, marking the largest ever Twitch sub-drop by an account on a single livestream". More than 120,000 subs were gifted across the season, to a cast including IShowSpeed, Cinna, JasonTheWeen and Emiru.
Now look back at our leaderboard read. Cinna is second and PlaqueBoyMax is third, with 98.0 and 92.8 per cent of their paid subs gifted, two weeks after that finale. Two of the three most subscribed channels on Twitch in September 2026 are, in large part, the output of an insurance company's marketing campaign.
Who else bought subs, and who did not need to
- Rockstar Games gave away "more than 600,000 Subscriptions" in a single week in December 2023, issued as tokens redeemable for a month of Tier 1 or a gift sub on any channel. Twitch called it "our Biggest Sub Gifting collaboration ever".
- McDonald's Germany bought sub bombs. In March 2021, with agency Jung von Matt SPORTS and streamer PietSmiet, it funded one-off gifted-sub drops worth EUR 750 each to 16 micro-streamers across two livestreams. The agency reported recipients' average viewership rising from the mid-20s to the hundreds.
- A public company reported a subathon as a marketing result. GameSquare's release on the September 2024 FaZe subathon claims "300K+ new Twitch subscribers", "710K+ follower growth", 2.2 billion minutes watched, and that FaZe was "the #1 brand by minutes watched on Twitch during September".
- Honda became the first exclusive automotive partner of Twitch Rivals in July 2024, running "Chat's Civic Mods", where chat polls customised a 2025 Civic that was then given away at TwitchCon. No spend was disclosed.
- Wendy's is still the reference case for doing it without buying subs at all. Its "Keep Fortnite Fresh" stream destroyed in-game burger freezers and reported 1.5 million minutes watched, more than 250,000 live views and social mentions up 119 per cent, though those figures come from the agency's own award entry.
Which Twitch brand deals went badly?
Four are worth your time: a sponsor exodus inside 48 hours, a rulebook Twitch deleted after a day, a category ban that cut gambling viewership by 75 per cent, and a record charity subathon whose money went missing. The failure modes are worth more than the success stories.
In June 2024, after allegations he had inappropriately messaged a minor, Midnight Society ended its relationship, studio head Robert Bowling saying "If you inappropriately message a minor. I can not work with you. Period." Turtle Beach terminated the next day. The San Francisco 49ers said they "will not be working with him going forward". Mountain Dew and Gillette said they had not worked with him for some time.
The line that should stay with any brand is the reporting in the same piece: the rumours "were an open secret within the industry for years, but his impressive follower numbers and high engagement were enough to convince major brands to slap their logos on his content without doing enough due diligence". The metrics were real. The diligence was not.
On 6 June 2023 Twitch published branded content guidelines that would have banned burned-in video and display ads and capped on-screen sponsor logos at 3 per cent of screen size. On 7 June it deleted them: "These guidelines are bad for you and bad for Twitch, and we are removing them immediately." The creator org OTK had put it plainly: "Small streamers who rely on their once-a-month Hello Fresh sponsorship cannot pay their bills under these new guidelines."
The gambling money left a crater
Twitch announced a gambling ban in September 2022, effective 18 October 2022, naming Stake.com, Rollbit, Duelbits and Roobet, after a streamer allegedly took around $200,000 from viewers and peers to fund a gambling habit. Trainwreck said he had been making "up to a million a month from the gambling companies". Ten months later Twitch said gambling viewership had dropped 75 per cent.
A record subathon's charity money went missing
ironmouse's record run raised money for the Immune Deficiency Foundation. In July 2025 she left her agency VShojo, alleging it had failed to pay $515,000 owed to the charity. She then ran the campaign herself and raised over $850,000. If you are sponsoring a charity subathon, the flow of funds is a diligence item, not a detail.
On the best available evidence, usually not. Ilya Morozov at Northwestern's Kellogg School and Yufeng Huang at the University of Rochester matched the top 60,000 Twitch streamers from May to December 2021 against Steam player data and Comscore sales. Organic streams lifted gameplay by about 3 per cent. Sponsored streams did less. The median return on investment was minus 95 per cent.
Morozov's summary: "For most of the games in our data, [sponsored streams] are not effective at all." The exceptions were obscure games from small developers and critically acclaimed titles, which is a useful shape: streaming works when it tells people something they did not know.
Set that against what the market quotes. A typical vendor rate card lists $100 to $800 per sponsored stream at 50 to 500 concurrent viewers, rising to $15,000 to $80,000 and up above 20,000, while disclosing that "All rate estimates are generated algorithmically based on publicly available data and industry benchmarks". No sample, no sources, no timeframe.
And against what the seller says. Amazon's own Twitch Research Power Group survey, fielded in July 2025 across eleven markets with 9,960 respondents, found 54 per cent of respondents trust their favourite Twitch streamers to recommend brands and products, rising to 67 per cent among highly engaged users. Useful, but it is the platform surveying its own audience about itself.
Our take: a median of minus 95 per cent is not a channel that fails. It is a channel being used for the wrong job. Twitch sponsorship is discovery. It works for things people have not seen and it is an expensive reminder for things they have. Buy it to introduce, not to remind.
What is Twitch actually selling brands in 2026?
Not the leaderboard: Twitch is selling Drops, campaigns streamers opt into from a dashboard, and access to Affiliates rather than stars. In August 2026 Twitch opened sponsorships to Affiliates, not just Partners, with a requirement that they be "Creator Sponsorship Certified", plus deal flow through Wehype and a live EA Battlefield campaign. Its evidence for the move: in its Minecraft campaigns, "over 50% of participating, paid streamers" were Affiliates.
Those Minecraft campaigns are worth seeing up close. Tiny Takeover, in April 2026, paid up to $1,000 per eligible streamer, first come first served, for streaming at least one hour of Minecraft in the right category and opting in through the Sponsorship Dashboard. That is a brand campaign with no negotiation, no agency and no sub count involved.
Meanwhile the metric Twitch itself reports growth on is Drops, not subscriptions. Its State of Gaming 2026 counts more than 7,500 Drops campaigns, up 20 per cent, and 40 million users claiming them, up 15 per cent, against 8.6 billion hours of gaming watched so far this year. At TwitchCon it said Drops that require a subscription to unlock "put an additional $11M in streamers pockets".
Four 2026 mechanics quietly change how sub numbers behave:
- Gift 'Em All lets one person gift up to 1,000 subs at once.
- Twitch says streamers who set up custom badges for an event saw "up to 50% more gift sub revenue" on day one, and that channels running gift sub discounts see "up to 80% more gift subs".
- Streamers can run their own 35 per cent discount on gift sub bundles of five or more, using a quarterly allowance of promo hours and tokens that caps at 96 hours and 12 tokens a year.
- The bar to become an Affiliate dropped in 2026 to four hours across four days, three average concurrent viewers and 25 followers.
How do you vet and price a Twitch streamer?
Start from the position that you cannot verify the headline number, and build the check around what you can see. Eight steps, in order.
- Ask for the Earnings page, screen-shared, not a slide. You want the sub split: paid Tier 1 against gifted against Prime. This is the only way to learn whether a sub count is recurring money or expiring gifts, and it is the single highest-value ask in the whole process.
- Ask for the last ten stream summaries, not the best one. Average concurrent, unique viewers and minutes watched together. Average concurrent alone hides a single spike.
- Compare unique chatters to average concurrent viewers. The published red flag is roughly "one chatter for every 500 viewers", alongside "suspiciously linear or perfectly stable" viewership patterns.
- Discount any viewership history from before September 2025. Twitch's viewbot detection change in August 2025 cut measured platform viewership by roughly 18 per cent a day. A rate card built on older averages is priced against viewers that no longer count.
- Ask how the channel is doing on viewbotting enforcement. Twitch now caps concurrent viewers on persistent offenders: "we will apply a cap to the streamer's CCV for a fixed period of time, on all of the Twitch surfaces." An industry whitepaper flagged more than 41,000 channels with 50 or more average viewers had at least one suspicious stream in a single quarter.
- Ask where the audience arrives from. The creator's dashboard splits internal Twitch discovery from external referrers such as YouTube, TikTok and Discord. External referrers point you at channels you can verify without asking anyone's permission.
- Put disclosure cadence in the contract. The FTC's position on live streams is that a single disclosure is not enough because "viewers can start watching at any time", and it recommends "multiple, periodic disclosures throughout the stream". It also reminds advertisers that "Your company is ultimately responsible for what others do on your behalf".
- Price the deliverable, not the sub count. Hours on air, segments, clip rights, a code or a link you can measure. If the only justification for the fee is a leaderboard position, you are paying for a rolling 30 day window that Twitch discounted.
What can Click Analytic tell you about a Twitch streamer?
Nothing, and we would rather say that plainly than imply otherwise. Click Analytic covers Instagram, TikTok and YouTube. Our database holds 400M+ creator profiles across those three platforms. We do not hold Twitch data, we do not score Twitch channels, and no tool we sell can tell you how many subscribers a Twitch channel has, because as this article has shown, nobody outside the channel can.
What we can tell you is the half of the homework that is actually verifiable. Almost every streamer in this article runs clips on YouTube, TikTok and Instagram, takes brand deals there, and sends their Twitch audience there. On those platforms you can read an audience without asking the creator: engagement, growth, audience geography and authenticity, and past brand collaborations.
Our take: Twitch hands you one number you cannot verify. Instagram, TikTok and YouTube hand you several you can. Do the Twitch checks with the streamer in the room and their dashboard on screen, and do the off-Twitch checks yourself before that meeting. If a creator's Instagram and TikTok audiences look bought, the Twitch numbers you cannot audit deserve no benefit of the doubt either. If you want the same exercise on a platform where the subscriber count is public, our rundown of the most subscribed YouTube channels is the easy version of this article.
