Woofun AI reports that Huang Zhenxin, head of Kimi B-end at Moonshot AI, acknowledged a bubble in the AI industry while affirming solid fundamentals and clear enterprise ROI. He attributed recent API price increases to global computing power shortages and insufficient chip capacity for token demand growth.

Huang emphasized evaluating model cost-performance via cache hit rates, noting Kimi’s rate exceeds 90%. He stated that Kimi will continue architectural innovations to sustain the Scaling Law, with its Muon optimizer widely adopted by mainstream models.