A recent analysis from Chinese tech circles highlights a growing crisis in the AI industry: while companies like DeepSeek and Xiaomi slash prices to attract users, the total cost of token consumption is exploding. The core issue is that lower prices drive higher usage, which in turn requires massive investment in compute and energy infrastructure. This creates a 'token burn' cycle where revenue per token drops but total costs rise, squeezing margins. For overseas developers and founders, this signals a need to rethink pricing strategies, optimize inference efficiency, and explore alternative business models like usage caps or enterprise tiers. The trend is particularly relevant for those building AI-native applications or investing in AI startups, as it suggests that the current race to the bottom may be unsustainable without significant breakthroughs in model efficiency or hardware.
Despite aggressive price cuts by AI companies like DeepSeek and Xiaomi, the cost of serving tokens is skyrocketing due to increased usage and infrastructure demands. This paradox threatens profitability and raises questions about the long-term viability of current AI business models.