Take‑Two eyes AI to trim GTA 6’s multibillion development bill
- Take-Two CEO Strauss Zelnick said on May 8 the company is exploring AI and other tools to curb rising game-development costs. - His clearest line was blunt: Take-Two “can’t deal” with exponential cost growth, and may not even handle linear growth either. - That matters because GTA 6 arrives November 19, 2026 — and blockbuster budgets are starting to look structurally unsustainable.
Video games are running into a scale problem. The biggest releases now take so many people, so many years, and so much money that even a hit can start to look financially absurd. That is the backdrop for Strauss Zelnick’s latest comments — not “AI will make GTA 6,” but something more practical. Take-Two wants technology to help keep giant games possible at all. ### What actually changed? On May 8, Zelnick said Take-Two is exploring whether AI can help the company “get there” as it tries to make games more efficiently, though he immediately framed “technology” as the better umbrella term. The point was cost control, not replacing ambition. He also said the company’s bet is not to give players “less” or make games “lighter, shorter, worse.” (gamespot.com) ### Why is cost suddenly the whole story? Because Zelnick put the problem in unusually direct terms. He said the business cannot handle exponential growth in production costs and probably cannot even handle linear growth. That is a striking admission from the company publishing Grand Theft Auto VI — the kind of game that can justify a gigantic budget precisely because it is one of a tiny number of true mass-market blockbusters. (gamespot.com) ### Is this about GTA 6 specifically? Yes and no. GTA 6 is the obvious example because it is widely expected to be one of the most expensive games ever made, and Zelnick has already described it as expensive. But his argument was broader — if costs keep climbing, some titles simply will not get made. So the real issue is the pipeline around GTA 6, BioShock, Borderlands, 2K sports games, and whatever comes after them. (gamespot.com) ### What kind of AI is he talking about? Probably not the flashy version people imagine. Zelnick also said Take-Two has not yet observed clear AI-driven cost savings, and maybe never will. That pushes the likely use cases toward quieter production tools — things like workflow automation, iteration help, testing support, simulation systems, or smarter NPC and world-behavior tooling. Basically, back-end leverage rather than “press button, get GTA.” (gamespot.com) ### Why does he keep saying “technology” instead of “AI”? Because “AI” now means two different things at once. One meaning is useful studio software that speeds up parts of production. The other is the hype claim that generative models can spit out an entire blockbuster game. Zelnick has been much more comfortable with the first idea than the second. Turns out that distinction matters — especially for a studio culture built on handcrafted scale. (gamespot.com) ### Why does this matter beyond Take-Two? Because Take-Two is not some fringe publisher experimenting in public. It owns Rockstar, 2K, and Zynga, and it is heading into GTA 6 with its next earnings report scheduled for May 21, 2026. When a company at that size says current production economics may break the model, that is a warning for the rest of the industry — particularly studios without a GTA-sized payoff at the end. (gamespot.com) ### Does this mean game prices go up too? Maybe. GameSpot notes that some analysts have already floated an $80 price for GTA 6 as a way to help reset the market. But Zelnick’s comments point to a different lever as well — if players will not accept endless price hikes, publishers have to attack the cost base instead. That is where AI and adjacent tooling become attractive, even if the gains are incremental. (ir.take2games.com) ### Bottom line The real news is not that Take-Two wants AI in the abstract. It is that one of gaming’s biggest publishers is openly saying the blockbuster model needs new economics. GTA 6 may still be big enough to brute-force that problem. The catch is everything else probably is not. (gamespot.com)