Moonshot AI is riding a wave of enterprise and open-source momentum for its Kimi K3 model.

Moonshot AI is riding a wave of enterprise and open-source momentum for its Kimi K3 model. On July 30, 2026, Technode reported that Kimi K3 is now available through Fireworks AI on Microsoft Foundry, giving Azure customers a managed path to deploy the model via OpenAI-compatible APIs. This expands Moonshot's enterprise distribution without requiring customers to manage GPU infrastructure. Separately, Technode cited Bloomberg reporting that Moonshot has closed a ~$3.5 billion round at a $35 billion valuation, with annual recurring revenue reportedly growing from $100M (March) to $300M (June). Pandaily's coverage quantifies the open-source impact: Kimi K3 captured 41% of global open-source model downloads within two days of launch and drove 4x overseas paid user growth. Pandaily also frames Moonshot's infrastructure engineering as the real moat. A Hacker News Show HN even demonstrated running full Kimi K3 on 29GB RAM, suggesting unusual inference efficiency that could shift GPU deployment patterns. These signals—enterprise channel expansion, massive open-source adoption, and heavy funding—position Moonshot as a rising force in AI models and infrastructure, with potential knock-on effects on inference GPU demand, cloud provider workloads, and open-source model economics.

What the sources said:

• "Fireworks lists Kimi among more than 20 open models served through Foundry." — Technode, https://technode.com/2026/07/30/moonshot-ais-kimi-k3-reaches-enterprise-users-through-fireworks-on-microsoft-foundry/
• "The round was roughly twice the upper end of the $1 billion to $2 billion target the company had initially set." — Technode/Bloomberg, https://technode.com/2026/07/30/moonshot-ai-reportedly-closes-3-5-billion-round-at-35-billion-valuation/
• "Kimi K3 ... 41% of World Open-Source Downloads and 4x Paid User Growth Overseas." — Pandaily, https://pandaily.com/kimi-k3-open-source-global-craze-jul2026
• "Show HN: Run Full Kimi K3 with 29 GB of RAM." — Hacker News, https://news.ycombinator.com/item?id=49106591