Moonshot seeks 30% K3 revenue share

This clause shows Moonshot is trying to keep cloud platforms from turning K3 into a commodity while keeping most of the economics for themselves. By releasing the weights, Moonshot gets broad distribution, but the separate license lets it charge large model hosts that sell K3 through their own clouds and APIs. That matters because Microsoft, Amazon, and Google own the customer checkout, the GPU fleet, and often the broader enterprise relationship.1 2 3 4

• The license is aimed at model as a service operators, not ordinary enterprise users. If a host and its affiliates make more than $20M over any 12 month period, commercial use requires a separate agreement with Moonshot. That gives Moonshot leverage over hyperscalers without blocking smaller developers from self hosting.2 3 4
• The practical fight is over who captures inference revenue. A cloud provider can package K3 inside an existing platform, bill the customer, and bundle it with storage, security, and developer tools. Seeking up to 30% means Moonshot is trying to get paid like a model owner even when someone else runs the sales channel and infrastructure.1 2
• This is a different play from fully permissive open models. Moonshot still sells its own API and apps, but the K3 license adds a second monetization path, enterprise cloud licensing. Comparable open weight companies use licensing restrictions to turn developer adoption into paid enterprise deals rather than relying only on cheap token sales.4 5 6

Going forward, the key question is whether Moonshot can make K3 important enough that hyperscalers accept sharing economics instead of steering demand to rival open models. If that works, Moonshot becomes more than an app and API vendor. It becomes a model owner that gets paid wherever K3 is sold.1 2 4