Masahiro Sakurai on AI: More Games, Fewer Worth Playing

Masahiro Sakurai has spent most of his career making games that had no template to copy. Kirby floats instead of fighting fair. Super Smash Bros. replaced health bars with a knockback percentage. Neither was what the industry expected in its genre, and both outlived the games that were.

So when Sakurai says generative AI is about to make the industry worse at originality, the reasoning is worth more attention than the headline. He laid it out in an interview with Denfaminico Gamer in October 2026, and it's a more specific argument than "AI is bad."

What Sakurai Actually Said

His starting point is that genuinely new ideas are hard to sell, not just hard to find.

Developers, he argues, can no longer describe a new game without comparing it to something that already exists — a shorthand that makes a pitch easy to understand and quietly removes the reason to make it. The hard part isn't having an unusual idea. It's convincing a team and a publisher that an unusual idea will work.

AI, in his view, makes that worse rather than better. The number of games will keep climbing while the number of games anyone actually wants stays flat.

For example, if Super Smash Bros. had followed the rules of a typical fighting game – 1-on-1 with guarding, attacking, and special moves, and a health system where the match ends with a KO – then I don't think the first Super Smash Bros. would have even been released. Even if it had been released, it would likely have been treated as a novelty item, and the history of Super Smash Bros. would have ended there.

He put the rest of it plainly: "AI will increase the number" of games released, but "I think most of them will become unprofitable titles." On the output itself — "When it comes to things created with AI… or things mass-produced with poor quality, I think most of them will simply be reproductions of systems seen in various other games."

The Numbers Behind It

Sakurai didn't pull the scale of the problem out of thin air. He used Steam as his yardstick.

When the Game Designers Award was established in 2010, roughly 275 games were being released on Steam per year. That figure now sits above 19,000 annually. His expectation is that AI pushes it higher still, because the bottleneck was never the idea — it was the labour of building it.

Two other data points make the same shape. At the October 2026 Steam Next Fest, 36% of the showcased games carried an AI content disclaimer, up from the previous event. And a report previewed at Tokyo Game Show 2026 found more than 85% of Japanese game developers now use generative AI in some part of their work.

That last number matters for a reason that's easy to miss: heavy AI adoption and a fear of sameness are not contradictory. They're the same phenomenon seen from two ends.

The Ueda Test

Sakurai reaches for Fumito Ueda — the designer behind ICO and Shadow of the Colossus — to draw the line between imitation and invention.

An AI can study every game Ueda has shipped and produce something that looks like his work. It has the full catalogue. What it can't do is guess what Ueda will think of next, because that idea doesn't exist yet anywhere in the training data. Ueda's next game is a surprise to everyone, including the people who have played all his previous ones.

That's the whole distinction Sakurai is drawing. Models interpolate; originality extrapolates. And the industry's incentive structure — publishers backing sequels, storefronts flooded with lookalikes, teams too stretched to gamble — already leans toward interpolation before a single prompt is written.

What Originality Cost in 1999

The reason his Smash example lands is that it isn't retroactive myth-making.

Sakurai devised the concept in 1998 with Satoru Iwata, then at HAL Laboratory, working on it outside office hours. The prototype was called Dragon King: The Fighting Game, and it had no Nintendo characters in it at all — four generic fighters on a screen. The goal was specifically to build a four-player brawler as a counterpoint to the 1-on-1 fighting games that dominated the market.

Nintendo characters came later, and for a reason that sounds almost mundane: Sakurai wanted "atmosphere." Line up generic fighter one, two, three and the cast blurs together; put Mario, Link and Pikachu in the same ring and players understand the game before they touch a controller.

The deeper break was the win condition. No health bar, no KO — you raise an opponent's damage percentage until they fly off the stage. It's the rule that makes four-player chaos readable, and it's the rule a committee optimising for "a fighting game" would have rejected.

Super Smash Bros. shipped in Japan in January 1999 with a small budget and little promotion, was originally planned as a Japan-only release, and became one of the best-selling games on the Nintendo 64. Sakurai still had the original prototype in 2022 and demonstrated it publicly, thirty years into a career built on that one decision.

The Kirby Counter-Example

The same instinct shows up earlier and smaller.

Sakurai created Kirby at 19 and directed Kirby's Dream Land in 1992, then Kirby's Adventure in 1993 and Kirby Super Star in 1996. The design premise was accessibility in the blunt sense: someone who had never played an action game should be able to finish it. Kirby's float is that philosophy turned into a mechanic — instead of demanding precision, the game lets you drift over the obstacle that beat you.

That's a design choice, and it's also a commercial one. It widened the audience for a genre that was busy narrowing its own.

Why This Is an Argument for Old Games

Take Sakurai's framing seriously and it has an uncomfortable implication for anyone choosing what to play this weekend.

If annual output climbs past 19,000 while most of it reproduces systems you've already played, then discovery gets more expensive. Sorting signal from noise costs time you'd rather spend playing. The games that survived twenty or thirty years of that sorting process have already passed the filter — not because they were old, but because enough people kept choosing them after the novelty wore off.

None of this requires taking a side on AI as a tool. Sakurai's point is narrower and harder to argue with: a market that makes production cheaper will produce more, and "more" has never been the same thing as "better" or "new."

Where to Start

The obvious place is the game he used as his own example: the N64 original, where the no-health-bar rule was new and four players on one cartridge was still an event.

On the other side of his career, there's the Kirby series — start with Kirby's Adventure on NES for the game that introduced Copy Ability, or Kirby Super Star on SNES for the one that turned those abilities into a full moveset.

Both run in a browser without an install. Browse the Nintendo 64 library and start a game — Play Classic Games. No Downloads. Just Play.