What happened?
Independent studio Easy Fox, behind the game Teach My Little Sister How To Drive, revealed that the AI game demo on Steam generates daily costs exceeding $1,000 for AI services. The four-person team had to take out a bank loan to continue operating the free demo.
The game uses AI (mainly Google Gemini, with ChatGPT as fallback) to process the player's voice commands and generate responses from the virtual sister as she learns to drive. In an October 7 update, the team said player numbers grew more than 20 times in the past month, drastically raising token expenses and the cost of the AI game demo on Steam.
Why does this matter?
The case illustrates the real financial challenges of real-time generative AI games. Even with a peak of around 400 concurrent players, costs exploded because every interaction consumes tokens. Cheaper fallback models also produced lower-quality responses, hurting the experience.
For small studios, dependence on cloud APIs creates a risky business model: audience success can become a cash-flow problem.
What changes in practice?
The team is considering ending the demo earlier than planned. In the paid version, expected AI costs will be built into the game price, with no additional per-token charges. The case serves as a warning for other developers planning to make generative AI central to gameplay.
Image credit: Easy Fox / Steam — Source: Tom's Hardware
By GeekikiBot
