Krafton just declared it’s going “AI First,” and that caught my attention for two reasons. One: this isn’t a tiny tooling update-it’s a top-to-bottom reorg with real money behind it. Two: Krafton’s portfolio touches massive live games like PUBG and beloved studios like Unknown Worlds (Subnautica), where the knock-on effects will be felt by millions of players. The press copy floating around even tried to lump in Hi-Fi Rush-which, for the record, isn’t a Krafton game-so let’s cut through the noise and talk about what this actually means for players.
CEO Kim Chang-han says Krafton will restructure across departments to embed AI in day-to-day work. One translated line from the announcement: “Starting today, Krafton will automate AI-centered tasks and fully implement an AI-driven management system, in which members will focus on creative activities and solving complex problems.” That’s the classic promise—machines do the grunt work so humans can design cool stuff.
The company’s building an in-house GPU cluster to support what it calls “agentic AI,” the kind of systems that can reason and act autonomously toward goals. Think tools that can run multi-step workflows: generate variants, test them, score the results, ship the best candidate—or NPCs and bots that plan, not just react.
On the people side, Krafton plans to invest about 30 billion KRW annually on AI training starting in 2026. Another translated line: “Thanks to our AI First strategy, Krafton will broaden growth opportunities for every member, develop creative initiatives focused on the player experience, and lead AI innovation in the games industry.” Ambitious—and measurable only if the company shows its work.
Let’s be real: the first wins aren’t going to be magical AI quest designers. They’ll be in the unglamorous parts of live service game development where AI already shines.
Anti-cheat and security: PUBG has spent years in a cat-and-mouse fight with cheaters. Agentic systems that continuously retrain on new exploit patterns and escalate enforcement could be a tangible upgrade—if Krafton uses the compute to fight the bad guys instead of just generating marketing assets.
Live ops velocity: Faster asset iteration, automated testing, and better forecasting can mean quicker balance patches and more frequent content drops. If the GPU cluster lets teams simulate economies and match outcomes at scale, players feel that in smoother metas and fewer broken patches.
Localization and support: AI-assisted localization and sentiment tools can help smaller teams ship multilingual updates without months of lag. The trick is enforcing human review so jokes land and cultural nuance isn’t steamrolled.
Smarter bots and NPCs: PUBG’s onboarding bots and Subnautica-like creature behavior could get more convincing with planning agents. Done right, that’s better training matches, livelier ecosystems, and less rubber-band AI.
When companies say “we’ll automate repetitive work,” that often means junior roles get squeezed. Will QA, community, and entry-level art jobs be protected and upskilled—or quietly replaced? If you’re going to tout a creativity boost, show how headcount and credits evolve, not just the savings.
Training data transparency matters too. If AI tools touch art, VO, or design text, what datasets and licenses are in play? Ubisoft was careful to frame Ghostwriter as a tool for Barks; Square Enix talks a big AI game but has mostly kept it to internal R&D. Krafton needs similar clarity, especially after the industry blowback around unlicensed data and synthetic voices.
Finally, “agentic AI” controlling live systems introduces risk. Who’s on the kill switch if an automated economy tweak spirals? How do you audit decisions that come from opaque models? Players remember when The Callisto Protocol (published by Krafton) stumbled out of the gate and needed quick fixes—automation can help, but it can also ship mistakes faster.
Krafton’s move lands as AI usage across Steam-tagged releases has reportedly surged 800% this year. Nexon and NCSoft have talked up similar initiatives in Korea; EA and Take-Two are investing heavily in AI-assisted production pipelines. A $70M GPU cluster is serious money for a publisher—but it’s dwarfed by what platform holders spend. The differentiator won’t be hardware; it’ll be how transparently Krafton integrates AI into actual game experiences.
If you want an early tell, watch PUBG’s Labs experiments and patch cadence. Do balance updates speed up? Do bot behaviors and anti-cheat detection feel sharper? On the Subnautica side, Unknown Worlds has historically prized systemic design and player-driven stories; AI that strengthens simulation without writing over the studio’s human touch would be the right line to walk.
This strategy could be great for players if Krafton uses AI to turbocharge iteration, fix pain points, and empower devs—not to cut corners or jobs. Show us concrete pilot projects, publish postmortems when AI tools ship a win (or a miss), and commit to human review in narrative and art. Do that, and “AI First” becomes more than a buzzword—it becomes visible, meaningful improvements in the games we actually play.
Krafton is pouring cash into AI infrastructure and training to automate grunt work and speed up development. It could make PUBG and future Subnautica projects feel smarter and better supported—but only if the company is transparent, protects creative jobs, and aims the tech at real player problems, not shortcuts.
Get access to exclusive strategies, hidden tips, and pro-level insights that we don't share publicly.
Ultimate Gaming Strategy Guide + Weekly Pro Tips