In 2014, ICANN threw open the domain-name system. Anyone with deep pockets could apply to run their own top-level domain, and hundreds arrived: .guru, .ninja, .pizza, .london, .app, .shop. The pitch was that .com was full, and the future was a thousand flavors of ending. A decade on, we can grade the experiment honestly.
Mostly, it flopped
The blunt verdict: most new gTLDs underperformed badly. Registration counts came in far below the hype, secondary-market prices stayed thin, and — crucially — user trust never really moved off .com. For a huge swath of the internet, a .guru or .click still reads as “cheap,” or worse, “scam.” When people don’t recognize an ending, they hesitate, and hesitation kills conversion. The default a normal person types and trusts is still three letters: .com.
But a handful of endings genuinely won — and the reasons they won are instructive.
The winners, and why
- .ai is the runaway success of the era. Originally the country code for Anguilla, a tiny Caribbean island, it happened to spell the two most important letters in technology. The AI boom turned it into a gold rush: .ai became the default ending for AI startups, and the registration windfall now reportedly funds a substantial share of Anguilla’s national budget — tens of millions of dollars a year flowing to an island of ~15,000 people. Cultural fit plus scarcity plus a technology wave: that’s the recipe.
- .io became the developer and startup darling — “input/output” to an engineer, short and clean. For a decade it punched far above its origins.
- .co positioned itself as a global “company/commerce” alternative to .com and found steady adoption as a credible short substitute.
- .xyz got a marquee endorsement when Google restructured under abc.xyz, and leaned into being the generic, anything-goes ending for a generation that doesn’t assume .com.
- .app and .dev (run by Google) carved out trusted technical niches — helped by a clever twist: they’re HTTPS-only by default, so security is baked in.
Notice the pattern. The winners didn’t win by being available — they won by having meaning (a word, an abbreviation, a wave to ride) and enough community to normalize them. Novelty alone was never enough.
The hidden risk in the two biggest winners
Here’s the catch domain investors keep learning the hard way: .ai and .io are country codes, not generic endings. They belong to territories, and territories are subject to geopolitics.
The clearest warning shot: in 2024, the UK agreed to hand sovereignty of the Chagos Islands to Mauritius. The .io ccTLD is tied to the “British Indian Ocean Territory” — a designation that, on paper, may cease to exist. Nobody knows yet exactly what becomes of .io registrations, and reasonable people disagree about the timeline and the outcome, but the mere question is something no .com owner ever has to ask. Similarly, .ai’s fortunes are ultimately tied to a single small government’s administration of the string.
This isn’t a reason to avoid them — it’s a reason to hold them with your eyes open. A ccTLD is a lease on someone else’s sovereignty, not a freehold.
What it means for how you buy
Put it together and a simple framework falls out:
- Want durability and universal trust? .com. It’s boring, it’s expensive, and it’s still the only ending nobody has to think about.
- Want a brandable, on-trend niche name — an AI product, a dev tool, a startup — and you understand the ccTLD risk? .ai / .io / .co / .xyz can be excellent, sometimes better than a clumsy compound .com.
- Buying to hold and resell? Weigh the geopolitics. A great .com is a freehold in a stable country. A great .ai is a fantastic house on land the landlord could, in principle, re-title.
The new-TLD experiment didn’t replace .com — but it did prove that a few endings, backed by meaning and community, can become real, valuable markets of their own. The trick, as always, is telling the few from the many.