Everyone knows what a AAA game is, right? If it were a movie, you’d call it a “blockbuster”: a big, glossy, heavily marketed production, expected to sell big numbers. Then there are AA games, mid-budget games from mid-sized teams — often nostalgically thought of as an endangered species. This term is fuzzier, but not as fuzzy as “indie,” the contentious umbrella for smaller, niche projects by small studios or solo developers, whether they were produced independently of a publisher or not.
We all know instinctively what we mean when we use these terms, but we don’t always agree, and any attempt to actually define them quickly gets mired in confusion and edge cases. This is exacerbated by the fact that production budgets, unlike in the film industry, are shrouded in secrecy and often unknowable.
HushCrasher, a new newsletter from game industry consultant Julie Belzanne and data scientist Antoine Mayerowitz, has attempted to solve this problem and create a new “taxonomy of video game production scope” using only hard, publicly available numbers: the install size of a game, and the number of people cited in the credits. These are used to divide games into four categories: AAA, AA, and two categories invented by HushCrasher: Midi (usually larger indie productions, like Hades) and Kei (the smallest games, including “solodev” titles, like Balatro).
Belzanne and Mayerowitz are hard data zealots. They describe HushCrasher as “a scream of rage against intuition-based theories” and say that “vibe-analysis is clearly out of the question.” They gathered data from all games published on Steam since 2006 and cross-referenced it with credits information from MobyGames (excluding special thanks and playtesters from the credits). Then they used cluster analysis to determine the size and shape of their four distinct groups.
The groups themselves are easy to grasp, especially with the help of the exemplars picked out by HushCrasher: Red Dead Redemption 2 for AAA, Clair Obscur: Expedition 33 for AA, Outer Wilds for Midi, and Undertale for Kei. A typical Kei game (named after Japan’s class of tiny cars) will have between 1 and 50 people in the credits, and come in under 5 GB. Midi games are up to 10 GB, and the credits might go into three figures. AA games are mostly in the 10-50 GB zone, and have credits in the hundreds, up to even 1,000 or so. AAA games are usually around the 100 GB mark or more, and have thousands of people credited.
It’s a clever classification system that’s delightfully clear. It’s also easy to use yourself using HushCrasher’s guide and chart. I was eager to test the limits of the system, so I spent a happy, nerdy couple of hours plugging recent titles into it to see what classifications would be applied, and what I could learn — both about the game industry, and the HushCrasher system itself. Here are some takeaways.
Nintendo is an anomaly
Using file size as a guide to a game’s production scope has one notable flaw: It can be affected by platform. Specifically, the Nintendo Switch’s lower specs and game cartridge storage format enforce much smaller game sizes that can radically change a game’s categorization.
You might think The Legend of Zelda: Tears of the Kingdom, vast as it is, is the definition of a AAA game, but in fact it’s AA according to HushCrasher’s system, and it’s not even close. With more than 1,000 people credited, it’s quite a large team for the category, but it’s only 18 GB. Mario Kart World (the first $79.99 video game of the current era) is AA too. Switch 2 flagship Donkey Kong Bananza is actually at the smaller end of AA, while Kirby and the Forgotten Land is Midi, mixed in with bigger indie games.
So there’s some platform bias here. But it’s also an indication of how conservatively Nintendo runs its business, reining in the scope (and, presumably, cost) of games that it sells at the same price as the more bloated competition.
True AAA is rarer than you think, and AA is much more common
It turns out that reports of AA’s death have been greatly exaggerated; it seems like a healthy category, especially in 2025, the year of Clair Obscur, Split Fiction, Avowed, Kingdom Come: Deliverance 2, Like a Dragon: Pirate Yakuza in Hawaii, Nintendo’s Switch 2 launch titles, and many more. These are all AA games according to the HushCrasher system, and they are among the best-reviewed and best-selling games of the year to date. Even Borderlands 4 is a AA game (MobyGames doesn’t have credits available for it yet, but its slender 28 GB install size guarantees its place in the category). Last year’s Game of the Year-winner Astro Bot is a classic example of a AA game, with around 400 people in the credits.
But 2025 has so far only seen two brand-new AAA releases: Assassin’s Creed Shadows (at least 5,000 people credited!) and Death Stranding 2: On the Beach. The latter is at the smaller end of this category, too. Meanwhile, Monster Hunter Wilds, at 75 GB and around 2,000 credits, is right on the cusp between AA and AAA.
The scarcity of AAA games, and resurgence of AA, is perhaps a sign of the industry reacting to the rapidly rising cost of development. Arguably, though, HushCrasher has set the border between AA and AAA a little too high. Often, I looked up perceived AAA games and found them in the lower category — but sometimes it goes the other way. Baldur’s Gate 3, everyone’s favorite underdog role-playing game? It’s actually full AAA, as its monster file size and 2,000-strong credits attest.
In indie-world, things are still messy
HushCrasher’s brave attempt to re-categorize the smaller end of the game production spectrum is only a qualified success — or at least, it still crashes uncomfortably up against perception.
According to Belzanne and Mayerowitz’s analysis, the vast majority of Steam releases are Kei games, but most of these go ignored. Indeed, it’s quite hard to find examples of the form that break through. But this could also be down to some distortion that results from using credits as a metric.
Even “solodev” games, made by single individuals or tiny teams, can have dozens of people in the credits if they have support from a publisher, or if they have a voice cast. Hollow Knight: Silksong, for example, is generally understood to be the work of the Team Cherry duo plus a couple of friends, and it’s self-published. But it actually has around 80 credited people, many of them musicians, making it a Midi game. The Ecuadorian indie game Despelote has a publisher and a large voice cast, and it’s 8 GB, so it’s Midi too. Blue Prince is a similar story. Balatro, the sole work of developer LocalThunk, is a true Kei game thanks to its tiny install size. But, due to publisher support, it has 80 credited people, and only just clings on to its place in the category as a result.
In a way, this is fair. Having publisher support — or, in Silksong’s case, the luxury of a long development time, well funded by the success of a predecessor — naturally increases the scope, marketing reach, and financial safety of a production. But it’s also arguable that, at the smallest scale, a raw credit count doesn’t proportionally reflect the number of people who put a substantial amount of work into a game.
Regardless, Belzanne and Mayerowitz have come up with a simple and very compelling way to reclassify games, and to understand the industry that makes them, all within an elegant two axes. They promise to get into actual production budgets next, and I’ll be fascinated to read about that, but I almost don’t want to introduce the complication. Things can only get messier from here.
