The Real Cost of a Bad Sunset

Sunsetting a live service properly is not free. Someone has to build the offline version, decide what the game's final state looks like, test it, and ship it, all for a product that has already stopped making money. That's a real cost, and it's fair to ask whether it's worth paying. If a studio skips all of it and just turns the servers off, what actually happens to them? Does it cost them anything at all?

The honest answer isn't "nothing." It's that the bill still arrives. It just arrives later, bigger, and on someone else's timeline.

The good version

Grasshopper Manufacture shut down Let It Die after nine years. Instead of just pulling the servers, they released an offline version for $19.99, a one-time purchase with no microtransactions ever again, so the people who'd played it could keep playing it. That took real, deliberate engineering effort for a game that was no longer earning anything. Nobody made them do it, and this is genuinely rare: outside a couple of small, Japan-only cases, it's hard to find another example that comes close.

The bad version

EA shut down Anthem servers in January 2026. Anthem's campaign was built to be played solo, and it went dark along with everything else: the game needed a server connection even for the part that never needed one. Former BioWare executive producer Mark Darrah later revealed, in a public post-mortem, that code for local and private server use had existed internally right up until a few months before Anthem's 2019 launch, code he said was still "there to be salvaged and recovered." Asked whether EA would actually release it, he didn't hold back: he considered it extremely unlikely.

Amazon's New World is bad in a different shape. There was no solo mode to needlessly kill, New World was always a shared multiplayer world. But Amazon runs AWS, one of the largest cloud server operations on the planet, the literal business of renting out server capacity to anyone who asks, and the community never got the option to rent servers directly from Amazon and keep the world running on their own terms. In-game purchases were cut off well before the servers themselves closed, which is more considerate than an abrupt cutoff, but a considerate process and a surviving game are two different things, and New World only delivered the first. What's actually missing isn't server capacity. It's the game server software, packaged in a form the community could run: decoupled from Amazon's internal authentication, anti-cheat, and live-ops systems, documented, supported. That's real engineering work nobody budgeted for, on a product already decided to lose money.

Link! Like! Love Live! failed differently again. The game shut down after three years, and the announcement landed so abruptly that Shunsaku Yano, the game's script director, found out only days before players did. That's not a technical failure like the others. It's a communication failure, the kind that happens when a studio treats a shutdown as an afterthought instead of a planned event with its own internal process.

Ubisoft shut down the original Crew in March 2024. Because the game required a constant internet connection, once the servers went dark, the entire game became unplayable, including its single-player campaign, the part that traditionally has no reason to need a server at all. Players had asked for an offline patch. Ubisoft had reportedly done something similar for older Assassin's Creed games before. They declined anyway.

Did The Crew actually cost Ubisoft anything?

The obvious place to check is the next game in the series. The Crew Motorfest, effectively The Crew 3, had the best first week in franchise history when it launched in September 2023. But The Crew 1's shutdown wasn't announced until three months later, so that record week can't tell us anything about the backlash.

Motorfest also launched separately on Steam, in April 2024, right after The Crew 1's servers actually went dark. That launch landed next to the live backlash instead of months before it, and it shows exactly what reversed timing does: Motorfest's Steam rating dropped to Mixed within three days, Valve disabled the reviews, and the price got cut in half almost immediately. The Crew 2 took it worse, dropping to Mostly Negative, its lowest score ever, even though its all-time rating held at Very Positive. A 50 percent price cut within days isn't subtle. That's real lost sales, not a guess.

The cost didn't stop there. The backlash reportedly helped push a class action lawsuit against Ubisoft, and contributed to a new California law protecting digital buyers. Valve started showing licensing warnings on every Steam purchase because of it, not just Ubisoft's games.

A year later, Ubisoft reversed course completely. The Crew 2 got a real offline mode in October 2025, called "Hybrid Mode," with one for Motorfest confirmed as coming. The studio built the exact thing it had refused to build the first time, just later, under pressure, after all of the above had already happened.

Planning means planning the end too

Here's something counterintuitive: planning a live service needs to include how it ends, not just how it starts. The team writing the code needs to know this from day one. Some systems, login requirements, matchmaking, live-tuned economy values, will eventually need to be flexible enough to be modified or bypassed entirely, and they should be built that way from the start. That's live development work, and it belongs in the initial scope, not a ticket filed the year the game stops earning.

Developers who've actually done the retrofit version confirm why this matters. Itchie, a Japanese programmer and producer who has worked at Square and SNK, told Automaton that mechanisms like progression tracking, inventory management, and reward calculations are normally handled by servers, and relocating them to run locally after the fact creates a fresh set of problems: save file tampering, synchronization issues, data that no longer lines up once the game resumes. Kei, another programmer interviewed in the same piece, put a number on it directly: asked to calculate the cost of taking a game offline after the decision to shut it down had already been made, the estimate came back roughly equal to building an entirely new game.

That's the retrofit cost. It isn't the cost of planning for it from the start. The developer behind Trolddom, an indie MMORPG, added offline support to his own multiplayer game after launch rather than before, and put it plainly: "You should try to plan ahead... don't be like me and glue it on later." The expensive version and the cheap version are the same feature. The only difference is when the decision gets made. It worked, too, Trolddom now lists optional offline mode and custom servers as a standard feature on its Steam page, sitting at Very Positive.

Even The Crew's own Hybrid Mode points at this. Players following the reverse engineering effort noted the team had reportedly started recording network traffic while the servers were still live, gathering the data they'd eventually need before anyone had officially decided to build an offline mode at all. That's still a retrofit. It's just a retrofit with a head start, which is exactly the kind of foresight that costs far less when it's actually part of the plan from day one.

What this actually means

Doing it right costs money. Skipping it doesn't save you that cost, it just delays it, adds legal and regulatory risk on top, and still ends with the same engineering work getting built eventually, more expensively and with none of the goodwill a proactive version would have earned.

That's the actual case for treating a shutdown as live development work from the start, not a crisis to manage later. Waiting doesn't remove the bill. It just adds interest. And it's a communication failure on top of a financial one: Ubisoft's own reversal a year later is the clearest admission that the thing players asked for the first time was buildable all along.