We reverse-engineered Telegram's channel IDs and now we can tell when each channel was created
Since late 2017 Telegram has drawn each new channel ID at random from a block of 100 to 250 million and moved to a fresh block every few months. The blocks date any channel to within a season, show the IDs passing 2³¹ in May 2024, and turn our index into a register of births
Every Telegram channel has a number. The app does not show it, but every bot, client library and archiving tool sees it, and it stays the same when a channel is renamed or changes hands. We wanted to know whether that number says when a channel was created. It does, though not in the way we expected.
Until late 2017 Telegram handed out channel IDs in order. Since then each new channel gets a random number from a block of 100 to 250 million, and every few months Telegram moves on to a fresh block. We mapped all 15 blocks used between September 2015 and September 2025 from 247,850 channels whose posts we could date. That turns a channel's ID into the season it was born in, and our index of 1.6M channels into a register of births.
- Telegram issued channel IDs in order until late 2017. Since November 2017 every block has been filled at random: among channels that went silent while their block was in use, the rank correlation between ID and date is between −0.06 and 0.10, against 0.78 in 2015–16.
- Blocks are used one at a time. The rate at which a block's channels fall silent halves in the month the next block opens.
- A block starting at exactly 2,147,483,648, or 2³¹, opened in May 2024. A raw channel ID issued since then no longer fits in a signed 32-bit integer.
- Blocks are being used up faster. 200 million IDs lasted 355 days in 2021–22 and 52 days in August 2025.
- Among channels born in 2015–16 that are still public and above 10,000 subscribers, 45% are in Persian. Among those born since May 2024, Persian is 7–8%, English 32–37%, Chinese 9–13% and Burmese 4–7%.
- 16 of the 25 largest channels in our index were born in the 11 months from September 2023 to August 2024, the tap-to-earn boom.
A channel cannot post before it exists
Telegram does not publish when a channel was created, but a channel's posts give it away. Whatever the date of its last post, the channel existed before it. For 247,850 channels we know the date of the last post, or of the earliest post our crawler has seen. Each of those dates is an upper limit on when its channel was born.
One such limit says little: a channel created in 2016 may have posted yesterday. But many channels are abandoned within days of being created. Put tens of thousands of channels with neighbouring IDs together, and the earliest last posts among them sit close to the moment those IDs were first issued. Below that line there should be nothing. The earliest post date anywhere in our data is September 22, 2015, the day Telegram launched channels.
Until late 2017 the numbers ran in order. Then they stopped
The steps are flat. Inside a block, IDs at the bottom and at the top start showing up on the same day, which is what you would see if Telegram picked each new ID at random from the block. To test it we took the channels in each block that went silent while the block was still in use, and compared the order of their IDs with the order of their last posts.
In the first block, used from September 2015 to January 2017, the two orders agree: the Spearman rank correlation is 0.78, and the lower half of the IDs went silent a median 8 months earlier than the upper half. In 2017 the link weakens to 0.23. In every block since November 2017 it is gone. The correlation sits between −0.06 and 0.10, and the median dates of the lower and upper halves are days apart.
| Block | In use | Channels | Rank correlation | Median, lower half | Median, upper half |
|---|---|---|---|---|---|
| 1.00B–1.10B | Sep 2015 – Jan 2017 | 107 | 0.78 | Feb 2016 | Oct 2016 |
| 1.10B–1.15B | Jan 2017 – Nov 2017 | 75 | 0.23 | Jul 2017 | Aug 2017 |
| 1.15B–1.40B | Nov 2017 – Jun 2021 | 1,517 | −0.01 | Aug 2020 | Aug 2020 |
| 8 more | |||||
| 1.40B–1.50B | Oct 2018 – Jun 2021 | 610 | 0.01 | Sep 2020 | Oct 2020 |
| 1.50B–1.60B | Jun 2021 – Oct 2021 | 249 | −0.06 | Aug 2021 | Aug 2021 |
| 1.60B–1.80B | Oct 2021 – Sep 2022 | 1,291 | 0.01 | Apr 2022 | May 2022 |
| 1.80B–1.90B | Sep 2022 – Mar 2023 | 473 | 0.04 | Dec 2022 | Dec 2022 |
| 1.90B–1.99B | Mar 2023 – Sep 2023 | 685 | 0.01 | Jun 2023 | Jul 2023 |
| 1.99B–2³¹ | Sep 2023 – May 2024 | 1,583 | 0.00 | Feb 2024 | Feb 2024 |
| 2³¹–2.25B | May 2024 – Aug 2024 | 819 | −0.01 | Jul 2024 | Jul 2024 |
| 2.25B–2.50B | Aug 2024 – Feb 2025 | 1,344 | −0.02 | Dec 2024 | Dec 2024 |
| 2.50B–2.70B | Feb 2025 – May 2025 | 438 | 0.02 | Apr 2025 | Apr 2025 |
| 2.70B–2.90B | May 2025 – Aug 2025 | 239 | 0.10 | Jul 2025 | Jul 2025 |
| 2.90B–3.10B | Aug 2025 – Sep 2025 | 130 | 0.06 | Sep 2025 | Sep 2025 |
One block at a time
Random draws inside a block could still mean several blocks in use at once, one per data centre for instance. The rate at which channels go silent says otherwise. Many channels stop within weeks of being created, so a block keeps losing channels quickly for as long as new ones are being born into it. When Telegram moves on to the next block, that supply of short-lived channels stops.
In the block that opened in May 2024, 24 of every 1,000 dated channels went silent in August 2024. In September, after the next block had opened, the figure was 12, and in October 8. Every block from 2021 onward shows the same drop in the month its successor opens.
There is one exception. From October 2018 to June 2021 Telegram drew from 1.15B–1.40B and 1.40B–1.50B together: channels in both ranges went silent at the same rate, month after month. Channels born in those years are the hardest to date. An ID in 1.15B–1.40B only says “between November 2017 and June 2021”.
In May 2024 the IDs passed 2³¹
Most blocks begin at a round number. One does not. The block that replaced 1.99B–2³¹ on or shortly before May 16, 2024 begins at 2,147,483,648, which is 2³¹. Of the 430 dated channels with IDs in the 3 million just below that number, 22 went silent before May 2024. Of the 353 just above it, 1 did.
2³¹ is one more than the largest number a signed 32-bit integer can hold. The Bot API never exposed the raw value: it shows channels as −100 followed by the ID, a number that has always needed 64 bits, so most bots were unaffected. Code that reads raw IDs from Telegram's client API and keeps them in a signed 32-bit integer overflows on every channel and supergroup created since. The next block, 2.25B–2.50B, opened in August 2024.
Telegram is using up blocks faster
Blocks do not last a fixed time. The 200-million block that opened in October 2021 lasted 355 days. The 200-million blocks of 2025 lasted 89, 79 and 52 days.
| Block | IDs | Opened | Replaced | Days in use | Order | |
|---|---|---|---|---|---|---|
| 1.00B–1.10B | 100M | September 22, 2015 | January 29, 2017 | 495 | Sequential | |
| 1.10B–1.15B | 50M | January 29, 2017 | November 5, 2017 | 279 | Sequential | |
| 1.15B–1.40B | 250M | November 5, 2017 | June 24, 2021 | 1,326 | Random | |
| 1.40B–1.50B | 100M | October 19, 2018 | June 24, 2021 | 978 | Random | |
| 1.50B–1.60B | 100M | June 24, 2021 | October 7, 2021 | 104 | Random | |
| 1.60B–1.80B | 200M | October 7, 2021 | September 27, 2022 | 355 | Random | |
| 1.80B–1.90B | 100M | September 27, 2022 | March 9, 2023 | 163 | Random | |
| 1.90B–1.99B | 90M | March 9, 2023 | September 26, 2023 | 200 | Random | |
| 1.99B–2³¹ | 157M | September 26, 2023 | May 16, 2024 | 232 | Random | |
| 2³¹–2.25B | 103M | May 16, 2024 | August 24, 2024 | 100 | Random | |
| 2.25B–2.50B | 250M | August 24, 2024 | February 16, 2025 | 175 | Random | |
| 2.50B–2.70B | 200M | February 16, 2025 | May 17, 2025 | 89 | Random | |
| 2.70B–2.90B | 200M | May 17, 2025 | August 4, 2025 | 79 | Random | |
| 2.90B–3.10B | 200M | August 4, 2025 | September 25, 2025 | 52 | Random | |
| 3.10B–3.20B | 100M | September 25, 2025 | — | — | Random |
If Telegram retires a block once a fixed share of its numbers are taken, channels and supergroups were being created about 7 times faster in August 2025 than in 2022. We cannot check that assumption from outside. A block could also be retired early or filled more thinly, and either would shorten its life just as well.
After September 2025 our data thins out. The 32 channels in our index with IDs above 3.20B must be newer still, but they are spread from 3.26B to 5.39B, too few and too far apart to map.
Checking the clock
Channels whose birth we can place independently land where they should. Telegram News (1,005,640,892) and Pavel Durov (1,006,503,122) sit in the first block, which opened on the day Telegram launched channels. The announcement channel of Hamster Kombat, the tap-to-earn game that took off in early 2024, sits in the block used from September 2023 to May 2024. DOGS, a meme token launched by Telegram's community in the summer of 2024, sits in the block opened in May 2024.
The ID also outlives a channel's name. Besides the official channel, 79 channels in our index carry “Hamster Kombat” in their name or username. 23 of them have IDs from blocks older than 1.99B–2³¹, so they were created before the game's own channel and renamed later. A name and a picture can be copied in a minute. An ID cannot.
Who arrived when
Read as birth dates, the blocks show which communities arrived on Telegram and when. The figures below cover the 121,805 channels above 10,000 subscribers whose language we know, 91% of all channels of that size. They describe channels that were born in each period and are still public today, not every channel born then. The oldest cohort has had ten years to die off.
The first block belonged to Iran. Of the channels born between September 2015 and January 2017 that survive above 10,000 subscribers, 45% are in Persian and 18% in Russian. Iran blocked Telegram in 2018, and the Persian share of new channels did not come back: since mid-2021 it has stayed between 7% and 10%.
Russian took over next. It is the largest language among channels born from 2017 to 2022 and peaks at 42% of those born between October 2021 and September 2022, the year Russia invaded Ukraine and blocked Western social networks at home. Since 2023 the newcomers are most often English-speaking: 37% of channels born between May 2024 and August 2024. Two smaller languages grow fastest. Chinese rises from 1% of the first cohort to 13% in early 2025, and Burmese from nothing to 7%.
What they were about
Topics shift as much as languages. Among the oldest surviving channels, 15% carry religious tags and 17% political ones. Among channels born since May 2025, 2% and 6%. Humour and entertainment rises from 10% to 32%, and adult content from under 1% to 12%.
Crypto has a single peak. It is 2% of the oldest cohort and between 6% and 9% of those born from late 2017 to mid-2023. It reaches 15% for channels born between September 2023 and May 2024 and 22% for May 2024 to August 2024, the months of the tap-to-earn airdrops. Since then it has been 13–15%.
| Born | News | Politics | Religion | Education | Motivation | Humour | Film & TV | Crypto | Gaming | Shopping | Adult | Gambling |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Sep 2015 – Jan 2017 | 24 | 17 | 15 | 20 | 16 | 10 | 2 | 2 | 2 | 5 | 1 | 1 |
| Jan 2017 – Nov 2017 | 21 | 15 | 9 | 18 | 14 | 14 | 3 | 5 | 3 | 7 | 1 | 2 |
| Nov 2017 – Jun 2021 | 20 | 14 | 8 | 18 | 14 | 15 | 4 | 6 | 4 | 7 | 2 | 2 |
| Jun 2021 – Oct 2021 | 17 | 10 | 5 | 14 | 11 | 20 | 6 | 8 | 6 | 8 | 4 | 3 |
| Oct 2021 – Sep 2022 | 17 | 10 | 5 | 13 | 11 | 19 | 6 | 7 | 7 | 9 | 4 | 3 |
| Sep 2022 – Mar 2023 | 16 | 8 | 4 | 14 | 10 | 21 | 7 | 8 | 8 | 8 | 6 | 4 |
| Mar 2023 – Sep 2023 | 15 | 7 | 4 | 13 | 10 | 22 | 9 | 9 | 9 | 9 | 6 | 5 |
| Sep 2023 – May 2024 | 12 | 6 | 3 | 12 | 9 | 23 | 9 | 15 | 11 | 10 | 7 | 5 |
| May 2024 – Aug 2024 | 12 | 6 | 4 | 9 | 7 | 25 | 10 | 22 | 15 | 11 | 7 | 5 |
| Aug 2024 – Feb 2025 | 13 | 5 | 2 | 8 | 7 | 26 | 11 | 15 | 14 | 11 | 9 | 7 |
| Feb 2025 – May 2025 | 15 | 5 | 1 | 9 | 7 | 28 | 10 | 13 | 14 | 12 | 12 | 8 |
| May 2025 onward | 15 | 6 | 2 | 8 | 7 | 32 | 11 | 15 | 13 | 12 | 11 | 6 |
16 of the 25 largest channels were born in 11 months
The biggest channels on Telegram are not its oldest. Of the 25 largest channels in our index, 16 were born between September 2023 and August 2024: the announcement channels of Hamster Kombat, Blum, TapSwap, Cats, Tomarket, DOGS and their peers. Our study of silent channels follows what became of that audience.
Age still counts for something. Channels born before mid-2021 make up 30% of channels above 1,000 subscribers and 38% of their subscriptions. The median channel born in 2015–16 has 3,879 subscribers; the median one born since May 2025 has 2,026.
| Born | Channels | % of channels | % of subscriptions | Median subscribers | Posted in 30 days | Silent over a year |
|---|---|---|---|---|---|---|
| Sep 2015 – Jan 2017 | 29,043 | 3.1% | 5.2% | 3,879 | 53% | 19% |
| Jan 2017 – Nov 2017 | 22,008 | 2.4% | 3.5% | 3,462 | 54% | 19% |
| Nov 2017 – Jun 2021 | 227,313 | 24.5% | 29.4% | 3,100 | 51% | 21% |
| Jun 2021 – Oct 2021 | 70,908 | 7.6% | 7.1% | 2,762 | 49% | 20% |
| Oct 2021 – Sep 2022 | 155,618 | 16.8% | 14.7% | 2,655 | 51% | 19% |
| Sep 2022 – Mar 2023 | 63,990 | 6.9% | 6.2% | 2,672 | 48% | 20% |
| Mar 2023 – Sep 2023 | 55,683 | 6.0% | 5.5% | 2,647 | 47% | 20% |
| Sep 2023 – May 2024 | 94,156 | 10.2% | 9.9% | 2,511 | 44% | 21% |
| May 2024 – Aug 2024 | 56,693 | 6.1% | 7.2% | 2,421 | 45% | 20% |
| Aug 2024 – Feb 2025 | 90,082 | 9.7% | 7.3% | 2,242 | 45% | 17% |
| Feb 2025 – May 2025 | 35,673 | 3.8% | 2.5% | 2,178 | 46% | 13% |
| May 2025 onward | 26,030 | 2.8% | 1.5% | 2,026 | 45% | 6% |
| Born | Largest | Second | Third |
|---|---|---|---|
| Sep 2015 – Jan 2017 | Pavel Durov11.0M | Telegram News10.4M | نيمار ابن الانبار4.8M |
| Jan 2017 – Nov 2017 | Леонардо Дайвинчик5.4M | خبرفوری4.1M | Mash3.1M |
| Nov 2017 – Jun 2021 | Proxy MTProto11.8M | Telegram Tips11.6M | Roxman10.0M |
| Jun 2021 – Oct 2021 | خبرفوری اخبارفوری جنگ فوری امریکا4.4M | GETMODPC4.3M | hitvpn_blog4.2M |
| Oct 2021 – Sep 2022 | Прямой Эфир • Новости4.1M | Wallet News3.9M | Factory Community3.5M |
| Sep 2022 – Mar 2023 | Bitget Wallet Announcements5.0M | Telegram Premium2.7M | Proxy MTProto پروکسی ایرانی2.7M |
| Mar 2023 – Sep 2023 | Dotcoin Community4.8M | Trending Apps4.4M | Builders3.8M |
| Sep 2023 – May 2024 | Hamster Kombat Announcement24.3M | Blum: All Crypto – One App17.9M | tapswap community12.0M |
| May 2024 – Aug 2024 | CATS Community10.0M | Tomarket Announcement8.4M | DOGS Community8.2M |
| Aug 2024 – Feb 2025 | Bum’s Corner6.0M | Tofan Trade | سیگنال کریپتو5.1M | Zoo Family5.1M |
| Feb 2025 – May 2025 | Телега7.1M | TikTok Prime3.5M | Чё по трендам?2.6M |
| May 2025 onward | Тренды моей ленты2.3M | melomix | ملو میکس1.3M | Сёрч1.3M |
Date a channel
Paste a channel's username, t.me link or numeric ID. Bots and client libraries return the ID either as a plain number or, in the Bot API, with a −100 prefix; both work. A username is looked up in our index; a numeric ID is dated in your browser against the block table above. The same lookup lives on its own page as the Telegram channel creation date checker.
Method
Population. Our index holds 1,573,304 public channels with a known numeric ID, observed between April 24, 2026 and August 13, 2026. That covers 94% of the channels above 1,000 subscribers in the index. Groups are left out. The index is not a random sample of Telegram, and a channel that is hard to find is less likely to be in it.
Dates. For 247,850 channels we have the date of the most recent post, or of the earliest post our crawler collected. Either is an upper limit on the channel's creation date. Deleting posts cannot move it earlier than creation.
Blocks. We split the ID range into 2-million bins and looked for places where the share of channels that went silent before a given date drops to zero from one bin to the next, then refined the edges at 0.5-million resolution. A block's opening date is the start of the first dense run of last posts in it: the first date from which at least three channels (blocks opened before 2021) or 0.15% of the block's dated channels (later blocks) went silent within 30 days. A block closes when its successor opens. The dates are accurate to days for blocks opened since 2021 and only roughly before that, when our index caught few channels that died young.
Order test. For each closed block we took the channels whose last post falls between the block's opening and closing dates and computed the Spearman rank correlation between ID and that date.
Exceptions. 140 channels (0.06%) have a last post dated before their block opened. We could not explain them, and they are too few to move any figure.
Language and topics. As in our census: a channel's language is the language detected in its own description, and topics are free-form tags from a language model grouped into families. Language and topic shares are for channels above 10,000 subscribers, where coverage is above 90%.
Survivors. A channel that was deleted or made private before we observed it is not in the index. Every cohort figure describes channels born in a period that were still public in 2026, and the older the cohort, the stronger that filter.
Data and citation
The block table, the cohort aggregates and the sample behind the first chart are published as a single file: telegram-channel-id-clock.json. Sample points carry an ID rounded to 100,000 and a date only. The only channels named in the file are the largest of each cohort and the four used to check the clock.
To cite this article: