On October 7, ICANN published the applications moving forward in the 2026 new gTLD round: 1,615 applications, 333 of them from brands. The list also showed 263 contention sets, meaning groups of applicants who want the same string. In my earlier CircleID piece, I warned that a crowded fight is not the same as value. Here is the follow-up question. What does a contention set tell you, and when does it turn into a price?

Why This Round Is Different

In 2012, applicants in a contention set could settle things privately, and many did, usually through a private auction. Those prices mostly stayed quiet, though a few leaked.

That route is closed. Section 5.2.3 of the 2026 Applicant Guidebook bans private resolution, including private auctions and paid withdrawals. What is left is a replacement string, Community Priority Evaluation for qualifying community applicants, and an ICANN auction as the last resort.

The October 7 list is still a draft. The replacement period runs from October 8 to 21, and String Confirmation Day on November 17 brings the final strings and updated contention sets. The objection window then runs to March 16, 2027. Expect some of the 263 sets to shrink or vanish before any auction.

What the 2012 Auctions Showed

Take three ICANN auctions from the last round. .app had 13 applicants and sold to Google for $25 million in February 2015. .shop started with nine applicants, seven qualified for the auction, and GMO Registry paid $41.5 million in January 2016. .web had eight applicants in the auction, and Nu Dot Co, widely reported to be backed by Verisign, paid $135 million in July 2016.

The string with the most bidders, .app, sold for less than a fifth of what .web fetched. The count did not predict the price. Who was bidding, and what the winner could do with the string, mattered more.

That matters now. .agent has 13 applicants, the same number .app had, and Google, Meta and OpenAI are reported to be among them.

What a Contention Set Does Tell You

It tells you something, just not a price. The standard evaluation fee is $227,000 per application, up from $185,000 in 2012, so applicants committed serious money before knowing whether they would win the string. When two or more of them independently decide a string is worth that risk, it is a real, if rough, sign of belief in it.

It also shows where the interest is. Traverse Legal counts 461 three-letter applications, and 318 of them (69%) are in contention. AI, agent and crypto strings are crowded too.

And each set that goes to auction becomes a dated, public price event. The last round often kept those prices private.

What It Does Not Tell You

It does not price a second-level domain. A registry’s willingness to pay for a whole extension is a different thing from what a buyer pays for one name under .com. Both can follow the same pull, generic words with commercial demand, but one does not convert into the other.

It does not prove demand either. Sophia Bekele argues in “Governance Theater” that the program works more like a toll booth than a market, since ICANN collects its fees whatever the outcome. I won’t take a side on that. It is still a useful warning: applicant counts partly reflect a willingness to pay a fixed fee, not proven demand from end users.

How I would read the list before November 17

Start by comparing the October 7 list with the November 17 one. A set that disappears, because of a replacement string or a withdrawal, was never going to produce a price.

Then look at who is in each set, not just how many. If 2012 is any guide, a few strategic bidders with a clear plan for a string can behave very differently from a crowd of portfolio applicants.

Finally, wait for actual auction results before drawing any line to .com. In my own work, I look at closed sales, not asking prices, and the same discipline applies here.

This is analysis, not valuation or investment advice.

Which contested string do you expect to fetch the most at auction, and what would change your mind?