Question · Q2 2025
Johanna Ma asked for an understanding of the cost structure differences between AI tools and traditional internet products, specifically which costs have significantly decreased and which have increased, and how to interpret the overall ROI for AI tools.
Answer
Chairman and CEO Fu Sheng explained that while a clear ROI for AI tools isn't fully established due to limited large-scale deployment, the biggest cost reduction comes from programming, where AI significantly lowers R&D costs and accelerates development. He acknowledged that calling large language models (LLMs) for AI products is more expensive than traditional internet products. However, he noted three trends: LLM token costs are rapidly decreasing (over 90% since GPT-3.5), user willingness to pay for AI products is strong, and commercial models are evolving. He expressed confidence that as LLM costs continue to fall and user adoption grows, the ROI for AI tools will become favorable, drawing a parallel to the past reduction in bandwidth costs for online video. He believes the break-even point for many AI applications, especially in programming, is rapidly approaching.
Ask follow-up questions
Fintool can predict
CMCM's earnings beat/miss a week before the call