Opinions are like noses (polite version).  Everyone has one.  Consider this:

- A huge chunk of new gTLD applications were from brands.  When/if they use them, it will not affect Internet users, since they’re not selling names. 

- Another huge chunk of new gTLDs are from portfolio applicants, none of whom appear to be in danger of disappearing.

- Yet another large group are fronting ethnolinguistic communities (.eus, .kurd, etc.).  As these TLDs are often manifestations of nationalistic feelings can’t be played out on the real-world stage (if they were, they’d have a ccTLD), they are important to these impassioned communities.  Hence, unlikely to go dark.

- Quite a few of the single-applicant new gTLDs are quite well-heeled.  I see Russian billionaires, very successful business people, etc., who see this at least partially as a vanity project.  Sure they might shut down, but are likely to do so in an orderly manner, protecting domain name owners.

- What about others, those that do fail?  Are they likely to go dark?  I don’t think so.  .HIV was picked up by Uniregistry, for instance, and .HIV registrants continue to use their names. The fact is that for those with infrastructure already in place, simply keeping the TLD going is a very small cost.  The incremental cost is all around the marketing / promotion of the TLD.  There are a number of registry operators who are not only good souls, but also have a vested interest in seeing the new gTLD program succeed.

- Finally there is ICANN’s EBERO (Emergency Back End Registry Operator) program, which exists to prevent the consequences of a failure of a TLD from affecting the people who own the names.  The biggest names in the business are part of that program, and they’re contractually bound to step in and make sure the second-level names keep resolving.

- How are gTLDs of any age worse off than ccTLDs?  None of these have gone dark.  It’s a mistake to assume that all ccTLDs are run by governments—many (most?) are not.  They too have to keep the lights on.

- As to the report saying that new gTLDs are riddled with junk registrations and fraud—surely that is just as applicable to old gTLDs, as well as ccTLDs.  In my estimation (measured by proportion of UDRPs to registration count) there is more fraud / phishing / spam in legacy TLDs than there are in new gTLDs.  There are bad guys out there, to be sure, but the Internet hasn’t fallen over and neither have any gTLDs.

So really, how many new gTLDs are likely to go dark?  My guess is a flat zero.  If they were going to go dark, we would have seen some do so by now.  As it stands, we haven’t even seen ICANN invoke their EBERO program. 

Maybe this kind of prediction makes good press and wins friends in some places, but it’s really not a subject that security experts have a better view on than other people. I have a nose too.