The developer of "Teach My Little Sister How to Drive," a driving instruction simulation featuring a dynamic, generative AI-powered non-player character, is reportedly spending $1,000 per day to maintain the game's demo. This substantial daily cost has led the small four-person team at Easy Fox to take out a bank loan to keep the demo operational.
The game's popularity has surged, with the developer noting a more than twentyfold increase in demo players over the past month. This increased engagement directly translates to higher operational costs, as the game constantly queries AI models to power the NPC's interactions, reactions, and disobedience.
To address these expenses, the developer plans to incorporate anticipated AI usage costs into the pricing of the full game release. Players will not face separate charges for AI token consumption after purchase. However, the financial strain is significant enough that the team may be forced to end the demo earlier than planned if current trends continue.
Originally launched in February, the demo gained notable traction partly due to popular streamers like think noodles. The game has also found a significant audience in Asia. For its AI functionalities, "Teach My Little Sister How to Drive" currently utilizes Google Gemini and ChatGPT, with plans to explore support for local AI models on capable PCs.
This situation highlights the growing real-world costs associated with generative AI. Recent reports have shown other companies facing unexpectedly high AI token expenditures, with one mystery company reportedly spending $500 million in a single month, and another experiencing $1.3 million in OpenAI API costs. The broader gaming industry has also seen discussions around AI, with companies like Capcom clarifying their stance on AI in development and publishers defending its use, even as teams work to replace AI-generated content with human performances. Activision also faced scrutiny regarding AI usage in "Call of Duty" marketing materials.