The Rise of Live Service Games and Why Some Fail

Somewhere around the halfway point of this year the count of games shut down in 2026 passed fifty. Not delisted, not quietly abandoned. Servers off, unplayable, gone.

That number is only shocking if you think of games as products. Once a game is a service, closure isn’t a failure state, it’s the ordinary end of a lifecycle, and the industry spent the last decade building an enormous number of things that end this way.

Why publishers chased it

The appeal was never subtle. A premium game sells once. A live service game sells continuously, to a smaller number of people, for years. The most successful examples generate more revenue annually than a hit premium release does in total, and they do it with predictable quarterly figures that public companies find much easier to plan around.

The problem is that the model doesn’t scale sideways. It concentrates.

The maths nobody wanted to do

A live service game doesn’t compete for a purchase. It competes for the same evenings, week after week, for years. Players have a finite number of those, and the games already occupying them have five to ten years of content, established friend groups and enormous production budgets behind them.

So a new entrant isn’t asking someone to spend €60. It’s asking them to leave something they’ve already invested hundreds of hours in. That’s a far harder sell, and it explains why the genre’s failures are so total. A mediocre premium game still sells some copies. A live service game that doesn’t reach critical population is unplayable, which accelerates the decline into a spiral.

What the notable failures had in common

Anthem is the clearest example of the structural problem. BioWare shipped it in 2019 with the framework of a live service game and not enough content to sustain one, attempted a full redesign, cancelled that redesign, and left it running for years in a state nobody was updating. The servers went off on 12 January 2026, and because the game was online-only it became permanently unplayable on that date. Seven years of a product existing without a plan.

XDefiant failed differently. Ubisoft built a competent arena shooter and put it directly against the most entrenched incumbents in the genre. Reviews were reasonable. It couldn’t hold a population against Call of Duty, and servers closed in June 2025.

Concord is the extreme case, pulled within about two weeks of launch. The postmortems focused on character design and marketing, but the more useful reading is that it entered the most saturated part of the market with nothing that distinguished it from free alternatives, at a premium price. There’s a coda worth noting: modders brought it back on custom servers more than a year later, which says something about preservation that the official version never managed.

What the survivors did

The ones that worked share a few traits, and none of them are secret.

• A hook that isn’t available elsewhere. Not better than a competitor, actually different from one.
• Content velocity from day one. The first six months determine whether a population forms at all, and a game that arrives with a promise of content instead of content usually doesn’t get a second chance.
• Monetisation that doesn’t read as hostile early. Aggressive monetisation works once players are invested. Deployed at launch it reads as extraction and kills word of mouth.
• Willingness to change direction fast. The successful ones responded to their first month of data with structural changes, not roadmap reassurances.

Where it leaves buyers

The practical implication is that buying into a live service game at launch is a bet on the population rather than the product. The game you’re paying for might not exist in eighteen months, and increasingly that isn’t a hypothetical.

It has also started to attract regulatory attention, with campaigns pushing publishers to leave games in a playable state after support ends. Whether that goes anywhere is unclear. What is clear is that the industry has now produced enough shutdowns for players to have noticed the pattern, and enthusiasm for the model has cooled accordingly.