Michael Guo(@Michaelzsguo):Kimi K3 exposed the real bottleneck in open AI: open weights do not mean easy inference. Moonshot paused new subscriptions as demand overwhelmed capacity, and recommends 64+ accelerator supernodes for K3. This also sheds light on what the model-serving ecosystem may look like once K3 releases its weights. It turns out we already have a thriving industry: @FireworksAI_HQ , @baseten and @togethercompute carry $38.8B in combined valuations, with revenue, bookings and token volumes growing rapidly.

Michael Guo

@Michaelzsguo

Building AI agents and AI-native orgs. Demystifying AI in practice. EN/中文 (Selected build notes, experiments, and practical tips at the website link.)

Joined January 2022

404 Following 5.6K Followers

Michael Guo@Michaelzsguo

2026.07.20 12:55

Kimi K3 exposed the real bottleneck in open AI: open weights do not mean easy inference. Moonshot paused new subscriptions as demand overwhelmed capacity, and recommends 64+ accelerator supernodes for K3. This also sheds light on what the model-serving ecosystem may look like once K3 releases its weights. It turns out we already have a thriving industry: @FireworksAI_HQ , @baseten and @togethercompute carry $38.8B in combined valuations, with revenue, bookings and token volumes growing rapidly.