Moonshot pivots Kimi to enterprise distribution

This marked Moonshot’s move from buying consumer traffic to inserting Kimi into other companies’ products, which is where durable AI revenue is being captured. After DeepSeek R1 reset the consumer market in January 2025, Moonshot stopped leaning on ads, released Kimi K2 as open weights in July 2025, and used inference partners to get Kimi into coding and agent workflows where customers pay for repeated production use, not one off chat sessions.1 2 3 4 5

• Moonshot had already spent heavily to win consumer attention, then cut that motion after Kimi’s user base fell as DeepSeek gained share. Its own business model section ties the July 2025 K2 release to a distribution shift through Fireworks AI and Together AI, with enterprise engineers supporting deployments for business customers.1 2
• The distribution change also changed who controls the customer. On Together and Fireworks, a US company can buy Kimi through the provider’s API, infrastructure, billing, and data handling setup, which makes Kimi easier to adopt for coding and legal agents but leaves the platform partner owning most of the account relationship.1 3 4 5
• This is a different strategic bet from DeepSeek’s. Moonshot scaled up to K3 at 2.8T parameters and pushed broad model licensing, while DeepSeek focused on lower active parameters and lower inference cost. That means Moonshot is chasing premium agent performance, while DeepSeek is optimized to win more workloads on price and serving efficiency.1 2

Going forward, the real contest is less about who has the most popular chatbot and more about which model becomes the default engine inside enterprise agents. If Moonshot keeps landing day zero placements on hosts like Together and Bedrock, Kimi can keep gaining share in production workflows even when consumer attention and benchmark leadership move elsewhere.1 3 5 6