Fourteen Years On, the Root Is Expanding Again. The 2012 Round Is the Handbook.
New gTLD applications close 12 Aug 2026. Building on 2012 lessons, new rules tighten controls and support entrants. October’s published list will reveal the industry’s future direction.
The top of the internet is taking applications again.
Fourteen years since the last window. This one closes on 12 August.
For a 227,000-dollar evaluation fee, any organization can propose a new domain extension — a private slice of the root zone, the top level of the internet’s names.
Last time, 1,930 applications arrived, aimed at a namespace that until then held 22 generic extensions.
Most of what they bought, you have never typed.
The 2012 round put more than 1,200 new extensions into the root zone, close to 100 of them in non-Latin scripts — the first of the round, delegated in October 2013, was Arabic, seventeen months after the application deadline.
Fourteen years is long enough to measure what happened next.
At the end of 2025, .com alone carried 161 million domain names — more than three times the 47.8 million held by every extension from that round combined.
The brand-owned extensions did worse.
By one industry tally from the end of 2024, roughly 130 of about 500 delegated brand extensions had been handed back or terminated, and around 126 of the survivors held a single domain name: the placeholder ICANN’s contract requires.
The story of these near-empty brand extensions would be worth a piece of its own.
What matters here is the shape of the whole dataset: the outcomes sorted themselves by category, and they sorted early.
A brand extension with no migration plan stalled the same way, whatever industry the brand came from.
A generic word with no sales channel behind it stayed empty no matter how good the word was.
The failures repeated by type, and the types were visible in the applications themselves.
The 2026 round opens with its outcomes already legible — and the proof is its own rulebook.
ICANN calls the round’s rulebook the Applicant Guidebook.
Against 2012, the title reads literally: a guide compiled from the last round’s case files.
The final version was published in December 2025.
Three of its rules sit directly over 2012 wounds.
Closed generics.
In 2012, companies applied to hold dictionary words as private space — .book, .search — because nothing in that year’s rules said they could not.
The ban arrived retroactively, after the applications were in, and the strings went to the shelf: when the ICANN Board confirmed in January 2024 that the model stays banned in the new round, .search and .book had still not launched.
A dictionary word, bought a decade earlier, frozen by a rule change that came after the purchase.
This time the prohibition is in the book before the window: closed generic applications are off the table unless a public-interest test for them exists — and none does.
The interesting part is the admission inside the rule: the sharpest risk of 2012 was rule risk.
The terms changed after the purchase.
Backend operators.
In 2012, technical capacity was a promise inside the application — every applicant described the registry it intended to run, and evaluators read the description.
In 2026 the promise has been moved outside the application and tested first: ICANN published its initial list of evaluated Registry Service Providers on 30 January 2026, three months before a single application could be filed.
Every applicant must name, from that list, the provider that will run its extension’s machinery — the registry core, the DNS, the DNSSEC signing — before the extension can reach the root.
Anyone who has run registry-side infrastructure reads the shift instantly: the extension is the shop window, the backend is the plant.
The window display is about to gain hundreds of new names.
The plants will belong to a short list of operators — the visible market broadens while the invisible one consolidates.
The entry fee.
The bluntest mechanism of 2012 was the last-resort auction: when several applicants wanted the same string and refused to settle, ICANN auctioned it among them.
Sixteen strings went that way, and the bidding raised roughly 225 million dollars, which sat in a segregated fund with no assigned purpose.
That fund is now financing this round’s access program.
The Board routed 5 million dollars of it to the Applicant Support Program in September 2024 and another 4.9 million in March 2026 — enough to carry up to 75 supported applicants at a 75 to 85 percent discount on the evaluation fee, with bid credits and reduced operator fees behind it.
Seventy-five such applications were in the pipeline as of 19 March.
The money companies once spent outbidding each other for the same word now pays the entry fee of applicants the old price had excluded.
All three rules lean in the same direction: the round moved its risks forward.
The provider exam sits before the application, the ban before the window, the support decision before the round — the Applicant Support Program took applications from November 2024 to December 2025 and closed before the main window even opened.
The 2026 guidebook is the 2012 autopsy, reissued as rules.
Which turns this October into something unusual: a forecast published as a list.
About nine weeks after the window closes, ICANN makes every applied-for string public — Reveal Day, mid-October on the current schedule.
Every string and every applicant, visible at once.
The weeks in between are an administrative check: ICANN verifies the applications for completeness and accuracy.
If your company’s name appears on that list under someone else’s application, the clock is already running: the string list is confirmed two weeks after Reveal Day, and government early warnings and objection windows follow from there.
Checking the list takes one afternoon.
It belongs on a calendar now, because the deadlines will not wait for a quarterly legal review.
The grounds for objection are few and named, and the deadlines are short; anyone else gets the public comment period.
Contention has a pressure valve this time: an applicant whose string collides with another’s can switch to a pre-submitted replacement string within fourteen days of Reveal Day.
And an applicant that reads its own odds honestly can still leave — the round runs a structured refund schedule, in three windows.
For every other entry, the handbook supplies the reading.
The string itself tells you almost nothing — 2012 proved a perfect extension moves no one on its own.
A pre-evaluated backend is table stakes now.
The question that separates applicants is the one the last round actually asked, and the ghost towns failed: does anyone outside the applicant have a reason to type this extension?
That is also the one question the new rulebook cannot regulate.
Every rule in it repairs a supply-side failure — who may apply, on what infrastructure, at what price.
Demand is not in the book, because demand cannot be written into one.
Which is why the ghost towns remain possible, and why the reading skill matters.
The categories to sort by are the 2012 ones.
Behind a brand application, look for a migration plan — a placeholder-only future is visible from the outside.
Behind a generic word, look for the sales channel: the registrars who will actually offer it.
Behind a community application, look for whether the community exists anywhere outside the application.
And the calendar keeps the excitement honest.
Applications without contention or objections reach the root 14.5 to 19.5 months after Reveal Day — late 2027 at the earliest, with launches stretching into 2028.
A string on the October list is a claim, not a product.
Fourteen years ago, the industry read the reveal list as a treasure map, and priced it accordingly.
It was a forecast, and most of it has since come true — by category, if not by every string.
This October the list will be a forecast again.
The difference is that the method for reading it is now public — written into the round’s own rules, with a price list attached.
The first expansion was an experiment. This one is an open-book exam — and the book is 2012.
Magyar változat: [zona.hu/tizennegy-ev-utan-ujra-bovul-a-gyokerzona-a-2012-es-kor-a-kezikonyv/]