CD Projekt Red: The Studio That Overpromised and Rebuilt

CD Projekt didn’t start as a game studio. Founded in Warsaw in 1994 by Marcin Iwiński and Michał Kiciński, it began as a distributor and localiser, importing Western PC games into Poland and doing the unglamorous work of subtitling and dubbing them for a market international publishers mostly ignored. The studio arm, CD Projekt RED, wasn’t formed until 2002, and it built its first game around a licence nobody outside Poland particularly wanted: Andrzej Sapkowski’s Witcher novels, a saga beloved at home and unknown almost everywhere else. That founding decision — bet on a story with a devoted regional audience rather than chase a trend — set the pattern for everything the studio did well for the next decade, and it’s worth holding onto, because the same instinct that built the Witcher games nearly sank the one that followed them.

The Witcher years

The Witcher (2007) ran on a modified version of BioWare’s Aurora engine, wore its budget openly in clunky combat and rough voice work, and still landed because Geralt of Rivia’s world refused the fantasy genre’s comfort: morally clean quests, tidy endings, evil that announced itself. Sapkowski’s monster-hunter-for-hire, contracted to solve problems nobody wants solved, gave the studio a voice — sardonic, class-conscious, suspicious of anyone claiming the high ground — before it had the budget to match its ambitions.

The Witcher 2: Assassins of Kings (2011) is the game where the budget caught up. Built on the studio’s own REDengine, it was the first CD Projekt title to look genuinely competitive on the PC hardware of its day, and its structural trick — a branching Act 2 that plays entirely differently depending on a choice made at the end of Act 1 — was a rare instance of a story actually forking rather than reconverging on the same beats with different scenery. It’s an expensive way to build a game (you’re paying full production cost for content half your players will never see), and it’s also the clearest evidence that this studio understood choice as a design problem that had to be built, tested and paid for like any other system.

The Witcher 3: Wild Hunt (2015) is the culmination and the one people still cite as a genre high-water mark: an open world where the side quests carry as much writing density as the main story, two expansions — Hearts of Stone and Blood and Wine — regularly named among the best downloadable content ever shipped, and a version of Geralt’s world that felt lived-in rather than populated with icons. It won game-of-the-year awards across the industry and it built CD Projekt Red’s reputation as the studio that respected its players: DRM-free through its own storefront GOG.com, free major patches, expansions priced like actual expansions. That reputation is the thing Cyberpunk 2077 spent.

The promise the launch couldn’t keep

Cyberpunk 2077, adapted from Mike Pondsmith’s Cyberpunk 2020 tabletop system, was announced in 2012 and spent eight years as the most-anticipated game nobody outside the studio had properly seen. It slipped through 2020 in stages — April to September, September to November, November to a final December 10 date — each delay accompanied by reassurance that the extra time was buying polish; what it was actually buying, reporting later showed, was a production running well past its depth.

The trouble was real, and reporting during and after development — most visibly Jason Schreier’s coverage for Kotaku and later Bloomberg — described sustained mandatory crunch on a title that had spent years marketing itself as the studio that didn’t work its people that way. That’s the part of the story that isn’t a design conversation: a public promise about labour practices, made repeatedly, that the studio’s own senior staff later acknowledged it hadn’t kept. It doesn’t need embellishing to land; the gap between the stated policy and the reported reality is the whole indictment, and CD Projekt’s leadership owned it publicly rather than disputing the reporting.

The December 2020 launch then delivered the second half of the disaster: a game that ran acceptably on high-end PC and buckled on base PS4 and Xbox One, the platforms the marketing had never stopped targeting. Sony pulled the game from the PlayStation Store entirely eight days after release — a first for a major first-party storefront and a AAA title of this size — and didn’t reinstate it for six months. Refunds went out at a scale the industry hadn’t seen for a single-player game, shareholder lawsuits followed over the marketing claims, and the studio’s stock lost roughly half its value inside weeks. None of this required a critic to invent anything; the public record did the damage on its own.

The design work under the bugs was frequently strong — Night City’s verticality, the character-build flexibility, Judy and Panam’s questlines — and almost nobody could see it clearly at launch, because the game kept crashing before the writing could land. I’ve covered the launch and the years of patches that followed it in more detail elsewhere on this desk; the short version is that a great game shipped inside a broken one, and CD Projekt spent three years digging the good one out.

The patch as apology, the update as argument

What makes this a genuinely unusual studio story, rather than just another troubled launch, is what happened next. CD Projekt didn’t quietly patch bugs and move on — it kept shipping structural revisions for years, culminating in the 2.0 update in September 2023, which rewrote the skill trees, overhauled the police and traffic systems, and rebuilt the cyberware and perk economy from close to scratch. That’s not a bugfix; that’s a studio publicly re-litigating its own systems design in front of the people it disappointed, on a three-year timeline, for free.

Two things helped the rehabilitation land. Netflix’s Cyberpunk: Edgerunners anime, released in September 2022, introduced Night City to an audience that had never touched the game and sent players back to it in numbers the original marketing campaign never achieved — a reminder that a strong setting outlives a bad launch if the setting itself was never the problem. And Phantom Liberty, the September 2023 expansion built for current-generation hardware only, gave the game an ending that argued for itself rather than apologising: a spy thriller with Idris Elba’s Solomon Reed carrying real weight against Keanu Reeves’ Johnny Silverhand, tightly scoped rather than sprawling, shipped alongside the 2.0 systems rework rather than bolted onto the old ones. I wrote about how that expansion actually closed the argument the launch had opened — it’s the rare DLC that functions as a studio’s closing statement rather than a victory lap.

The Witcher 3 got its own coda in the same spirit: a free next-generation update in December 2022, four years after the base game’s expansions, rebuilding Geralt’s world for hardware that didn’t exist when it shipped. It shipped free, at a moment the studio had every commercial excuse to charge for it — the same instinct that built GOG.com, applied again.

None of this erases the underlying story about labour. The industry’s broader habit of treating unsustainable schedules as a heroic last mile rather than a planning failure is a pattern I’ve written about more generally in the myth of the heroic deadline, and CD Projekt Red’s Cyberpunk cycle is one of the clearest, best-documented cases of it. A studio owning its mistake in public and rebuilding the product doesn’t retroactively justify how the product got made; it’s simply the more useful thing to have happened than the alternative, which is most studios’ actual default: ship the broken version, patch quietly, say nothing.

What the rebuild bought

CD Projekt Red also restructured internally through this period — trimming headcount in 2022 as part of a wider reorganisation, closing down peripheral projects including the Boston studio’s earlier scope, and consolidating around two franchises rather than the broader slate it had floated mid-decade. The studio’s next moves are a direct bet on the credibility it rebuilt: The Witcher 4, codenamed Polaris, was revealed in October 2024 running on Unreal Engine 5 — the studio’s first mainline title to abandon its own REDengine — with Ciri confirmed as the protagonist rather than Geralt. A Cyberpunk sequel, developed in part by the Boston studio built around the former Molasses Flood team, is in early production under the codename Orion.

Both bets are the same wager: that an audience which watched this studio break a promise and then spend three years proving it understood why that mattered will extend credit again. It’s a bet no studio gets to make twice. CD Projekt Red is currently living on the first extension, and the next launch — not the next patch — is where we find out if it holds.

The GOG instinct, applied to a mistake

It’s worth naming what actually connects the Witcher-era goodwill to the Cyberpunk-era recovery, because it’s a single, consistent business habit rather than two unrelated stories. GOG.com, launched in 2008, sold a promise that the studio would let you own what you bought — no always-online check, no client mandatory to press play — at a moment when the rest of PC publishing was moving hard toward locked storefronts. That promise cost CD Projekt real revenue for years before digital storefronts made DRM-free distribution commercially viable at scale. The Witcher 3’s two large expansions, priced and scoped like genuine game-length content rather than reskinned side quests, cost the studio the easier margin of smaller, cheaper add-ons sold more often. The free next-gen update cost it the obvious monetisation of a paid remaster four years after the fact.

None of those decisions were forced. Each one traded a short-term commercial upside for a longer relationship with the audience buying the next game — and that’s precisely the currency CD Projekt Red spent when Cyberpunk 2077 shipped broken. A studio with no banked trust does not get three years of patches and a from-scratch systems rework read charitably; it gets written off, the way plenty of launches that never recover are written off. The 2.0 update and Phantom Liberty worked as a redemption arc specifically because the studio had already demonstrated, repeatedly and at real cost, that “we’ll fix it properly” was a sentence it had kept before. That history is also the reason the crunch revelations landed as betrayal rather than routine industry news — the studio had spent a decade telling players and press it was different, and for one project, under real pressure, it wasn’t. Both halves of that reputation are true at once, and a career read of this studio has to hold them together rather than picking whichever one is more convenient.