Everyone is talking about open-sourcing AI models, but few discuss the cost of actually deploying them. According to estimates shared in Kimi's official community, running Kimi locally in China could require 64 Nvidia H200 GPUs (if available), 8TB enterprise SSD storage, liquid cooling, power infrastructure and engineering staff. Total costs could exceed $4.2 million-and still only serve a limited number of users. Open source does not mean free. Meanwhile, Moonshot AI is reportedly seeking a $50 billion pre-IPO valuation. The company needs to move quickly as the market reassesses AI valuations and the sustainability of the Anthropic-style business model.
Kimi 官方社区估算,在中国本地运行 Kimi 需 64 块 Nvidia H200 GPU、8TB 企业级 SSD、液冷及电力设施等,总成本超 420 万美元,且仅能服务有限用户。开源不等于免费。同时,月之暗面(Moonshot AI)正寻求 500 亿美元上市前估值,需在市场重估 AI 估值及 Anthropic 式商业模式可持续性之际快速行动。
Everyone is talking about open-sourcing AI models, but few discuss the cost of actually deploying them. According to estimates shared in Kimi's official community, running Kimi locally in China could require 64 Nvidia H200 GPUs (if available), 8TB enterprise SSD storage, liquid cooling, power infrastructure and engineering staff. Total costs could exceed $4.2 million-and still only serve a limited number of users. Open source does not mean free. Meanwhile, Moonshot AI is reportedly seeking a $50 billion pre-IPO valuation. The company needs to move quickly as the market reassesses AI valuations and the sustainability of the Anthropic-style business model.