A game the world wrote off in 2020 just became one of the 23 best-selling ever

Gaming 4 min 80 sources

Cyberpunk 2077 passed 40 million copies this week — five years after a launch so broken Sony pulled it from its own store. In the same week, Nintendo and Square Enix announced two big mobile games are shutting down for good. Two opposite fates that turn on one quiet fact about how a game makes money.

Key takeaways

Cyberpunk 2077 passed 40 million copies this week, five years after a launch so broken Sony pulled it from the PlayStation Store — most of those sales came after the disaster.
A game you own can be patched and rediscovered for years; a live-service game like Mario Kart Tour, which Nintendo is shutting down in September, simply vanishes when the servers close.
Whether a bad launch is fatal depends on the business model — durable goods get second acts, service games rarely do.

The comeback nobody scored at launch

On July 3, CD Projekt announced Cyberpunk 2077 has now sold more than 40 million copies — up 5 million since November, when it hit 35 million [1][2]. That puts it among just 23 games in history known to have sold 40 million or more [3].

It’s a strange sentence to write, because in December 2020 this was the most infamous launch in years. The game shipped so broken on older consoles that Sony pulled it from the PlayStation Store and offered refunds — an almost unheard-of move [1]. CD Projekt was accused of misleading buyers with its pre-release marketing [1].

Then the studio spent years rebuilding it. Patches, a ground-up 2.0 overhaul, a paid expansion, and a Netflix anime that sent players flooding back. Co-CEO Michał Nowakowski still won’t call it finished, saying recently the studio has yet to complete a “full redemption arc” [1]. But the number is the number. Most of those 40 million copies were sold after the disaster — not during the launch that defined the game’s reputation.

Why a game can outlive its worst day

Here’s the mechanism under the comeback: a game like Cyberpunk 2077 is a durable good. You buy it once, you own the copy, and — crucially — the maker can keep fixing it after the sale. That gives it a long tail: it keeps earning for years, through patches, discounts, expansions, and cultural moments.

The cultural moment mattered enormously. When the Edgerunners anime landed on Netflix in 2022, sales spiked and players who’d written the game off came back. A sequel series, Edgerunners 2, arrives this fall — and CD Projekt openly frames the 40-million milestone as “a great foundation” for new projects in that universe [1].

None of that is possible if the launch is the end of the story. For a durable, ownable game, the launch is just the first frame of a movie that runs for years.

The opposite fate, announced the same week

Now hold that next to two other games in this week’s news. Nintendo confirmed Mario Kart Tour will shut down in September, and Square Enix is closing Final Fantasy 7 Ever Crisis — the latest in a steady retreat from mobile as publishers pull back spending [4][5].

These are live-service games: they don’t live on your device, they live on a company’s servers, and they make money by selling things to the same players over months and years. When Mario Kart Tour goes offline, it doesn’t get patched or rediscovered. It’s simply gone — and so is everything players spent inside it [4][6].

That’s the difference. A live-service game can’t stage a Cyberpunk-style comeback, because it doesn’t persist. There’s no shelf to sit on, no copy to rediscover, and every day it stays open costs money to run. When the spending stops, the game ends. For this model, the launch window really is close to the whole story — get it wrong and there’s rarely a second act.

The middle path: ship it unfinished on purpose

Between the two sits a third model, and it happens to hit its milestone today. Palworld — the creature-collection survival game — passed 40 million players and releases its “1.0” full version on July 10, two and a half years after launching in “early access” [7].

Early access means selling a game while it’s still openly unfinished, then improving it in public with players’ money and feedback funding the work. Palworld broke sales records in early 2024; its studio said the profits were almost too large for a team its size to handle [7]. (One caution on the number: 40 million players isn’t 40 million sales — the game is also on Microsoft’s Game Pass subscription, so free-to-the-player installs count too [7].)

Early access is the durable-good idea taken to its logical end: the game admits the launch isn’t the finish line. It’s a bet that a patient audience will judge the destination, not the starting point.

What players actually see

Line the three up and a pattern appears. The question that decides a game’s fate isn’t “was the launch good?” It’s “can this game keep changing after you buy it — and does anyone still own it?”

Cyberpunk could recover because it’s a thing you own that its maker kept mending. Mario Kart Tour can’t, because it’s a service that stops when the servers do. Palworld built the mending into the plan from day one.

For players, this is why “wait for reviews” and “wait a year” have become such common advice — and why a game pulled from a store can still, five years later, end up on the shortlist of the best-selling games ever made.

02 · Lesson · why it matters

The verdict gets fixed on the worst day — and the thing keeps changing

We judge things at their most fragile moment and treat that snapshot as the whole story, but whether the snapshot is fair depends on whether the thing can still move.

The version everyone remembers no longer exists

In December 2020, Cyberpunk 2077 was the punchline of the year. It launched so broken that Sony pulled it from its own store and handed out refunds. That was the moment the world made up its mind.

Five years later it has sold more than 40 million copies — a place among the two dozen best-selling games ever made. Most of those sales came after the disaster. The code was patched, rebuilt, expanded. The version that earned the reputation is not the version that earned the sales. It stopped existing years ago.

The reputation stayed anyway. This is the thing worth seeing: a verdict formed at a thing’s worst moment tends to freeze there, even as the thing keeps moving out from under it.

A snapshot posing as a summary

Think about what a launch-day judgment actually is. It captures a game on the single day it is most fragile — untested at scale, unpatched, seen for the first time. Then that one frame gets treated as the summary of everything the game is and will become.

There’s an arrangement underneath this, and it isn’t anyone’s villainy. Reviews land at launch because that’s when buyers need the buy-or-skip signal, and that’s when attention — and traffic — is highest. The whole machine is built to render a fast verdict at the most fragile moment. It serves the buyer who wants to decide today. It quietly punishes anything that mends slowly and improves in the dark.

A snapshot is honest about one instant. It just isn’t a summary. We keep mistaking the two.

The same habit, fair for one thing and cruel for another

Here’s where it gets sharper. That same launch-day verdict is roughly fair for some games and deeply unfair for others — and what decides which is a fact most players never think about.

A game you buy and own is a durable thing. The maker can fix it after the sale, and it can be rediscovered years later — a discount, an expansion, an anime that sends people flooding back. Its worst day is the worst it will ever be. Judging it then is judging it at the bottom.

A live-service game — one that lives on a company’s servers and sells to the same players for years — is different. It can’t quietly recover, because it doesn’t persist. When the spending stops, the servers close and it’s simply gone. For that kind of game, the launch verdict is close to the truth, because there is no long second act.

So the honest question is never just “how good was the launch?” It’s “can this thing still change — and will it still be here?” The same snapshot is a fair reading of one and a cruel one of the other. The judgment didn’t change. What could move underneath it did.

Who is holding a frozen verdict

This isn’t really about one game. Almost everyone is carrying a version of it.

The developers spent five years fixing something the world had already filed under “disaster,” working against a verdict that had stopped listening. The players who wrote it off in week one still hold an opinion about a game that no longer exists — confident, and years out of date. And the pattern runs far past games: a person judged by their worst month, a plan dismissed after its roughest week, a place remembered for the year it was at its lowest. The snapshot gets filed, and the filing stops updating even as the thing walks on.

We are all somewhere in this — both the ones frozen by an old verdict and the ones still holding one. It’s not a flaw in bad people. It’s just how cheaply a first impression hardens into a permanent record.

What the seat can’t see

A launch-day verdict is one seat’s view of one moment. From that seat, on that day, it may even be right. What the seat can’t see is the years the thing keeps moving after the judgment is filed and the attention has gone.

Seeing that doesn’t mean every written-off thing is secretly a triumph in waiting — most broken things stay broken, and a live-service game that dies at launch really is over. It means the verdict and the thing are two separate objects that drift apart with time, and the confidence we feel about the old snapshot is worth about as much as the day it was taken.

03 · Lab · your turn

Lock the Verdict

Place your judgment on a game's five-year arc and feel the gap between the frozen verdict and the version that kept changing — and why the model decides if the snapshot is fair.

04 · Hope · carry this

A game the world wrote off spent five years quietly becoming one of the best-selling ever. A verdict is rarely the last word, and patient hands can still rewrite the record long after the attention has moved on.