Capital · out
compute buildout to meet K3 demand + subsidized/rate-limited free-tier access; no disclosed capex or training-cost figure (contrast DeepSeek's disclosed ~$5.6M V3 run — Moonshot has not published an equivalent for K2/K3)
Sourced figures 1
- TradingView/Reuters + 36Kr via Google News K3 demand strained compute capacity, forcing a subscription pause — implies heavy training/serving spend not itself quantified
the working
Threads 2
A claim is one metric we assert about a node — the value is the summary, this page is the receipt. Sourced from kestrel; supersedes-history not yet shown. Back to Moonshot AI.
