- Type
- Artificial intelligence company
- Founded
- March 2023, Beijing, China
- Founders
- Yang Zhilin, Zhou Xinyu, Wu Yuxin
- Chief executive
- Yang Zhilin
- Key products
- Kimi chatbot; Kimi K-series open-weight models
- Reported valuation
- US$35 billion (July 2026)
- Related
- DeepSeek, Zhipu AI, Qwen, MiniMax
- Type
- Artificial intelligence company
- Founded
- March 2023, Beijing, China
- Founders
- Yang Zhilin, Zhou Xinyu, Wu Yuxin
- Chief executive
- Yang Zhilin
- Key products
- Kimi chatbot; Kimi K-series open-weight models
- Reported valuation
- US$35 billion (July 2026)
- Related
- DeepSeek, Zhipu AI, Qwen, MiniMax
Moonshot AI is a Chinese artificial intelligence company based in Beijing, founded in March 2023 by Tsinghua University alumni Yang Zhilin, Zhou Xinyu and Wu Yuxin. The company develops the Kimi family of large language models and the Kimi chatbot, and is known for releasing frontier-scale models under open weights — including Kimi K3, a 2.8-trillion-parameter system released in July 2026 that the company describes as the largest open-weight model ever published.[1][6]
Moonshot is commonly grouped with DeepSeek, Zhipu AI and MiniMax among China's leading independent model developers. It grew from an early consumer-chatbot success into a frontier laboratory whose funding rounds, open-weight releases and planned Hong Kong listing made it one of the most closely watched AI companies in Asia during 2026.[2][3]
History
Moonshot AI was founded in March 2023 by Yang Zhilin — a machine-learning researcher who had worked at Google Brain and Meta AI after studying at Tsinghua University and Carnegie Mellon University — together with fellow Tsinghua alumni Zhou Xinyu and Wu Yuxin. The company raised early backing from Chinese venture investors, and Yang has described its goal as building foundation models on the path toward artificial general intelligence.[6]
In October 2023 the company launched Kimi, its first chatbot, which could process up to 200,000 Chinese characters in a single conversation — among the longest context windows available at the time — and became popular for analysing long documents, contracts and research reports. Kimi was named after Yang's English nickname. The company subsequently attracted investment from major Chinese technology firms including Alibaba and Tencent, reaching a multi-billion-dollar valuation by 2024.[3][6]
Moonshot moved into frontier-scale open releases in July 2025 with Kimi K2, a one-trillion-parameter mixture-of-experts model whose weights were published freely, making it one of the largest open-weight models then available; a reasoning-focused K2 Thinking variant followed in late 2025, and the company reported fast-growing revenue from paid subscriptions and API usage.[5][3]
The pace accelerated in 2026. Moonshot raised roughly US$700 million at a US$10 billion valuation in February, about US$2 billion at US$20 billion in May — when its annualised recurring revenue was reported to have passed US$200 million — and a larger-than-expected US$3.5 billion round in July at a reported US$35 billion valuation, riding momentum from the Kimi K3 release. Demand for K3 was so intense that the company paused new chatbot subscriptions on 20 July 2026, and reports said it was preparing an initial public offering in Hong Kong within roughly six months, alongside talks for a final pre-IPO round that would value it at US$50 billion.[2][3][4]
Key Technologies
The Kimi K-series models are sparse mixture-of-experts systems. Kimi K2 contains about one trillion total parameters while activating roughly 32 billion per token; Kimi K3 raised the total to 2.8 trillion parameters, which Moonshot says makes it the first open model to reach that scale. K3 features a context window of up to one million tokens, accepts native multimodal input, and is optimised for agentic and coding tasks; the company published the full model weights on 27 July 2026, allowing organisations to inspect, fine-tune and self-host the model without using its API.[1][5]
Open-weight publication is central to Moonshot's strategy. By releasing frontier-scale models under permissive licences, the company spreads its technology across the global developer ecosystem — including hosting platforms and self-managed infrastructure — in direct contrast to the closed release model of leading American laboratories. Analysts have described this approach as both a distribution strategy and a hedge, letting Moonshot build global mindshare in markets where its consumer app cannot easily operate.[1][7]
The company's commercial products include the Kimi assistant for consumers, a developer API platform, and the open model releases themselves, which are distributed through the Hugging Face ecosystem and used inside third-party products such as coding agents and search tools.[3][7]
Applications and Impact
Kimi is used by consumers for long-document analysis, writing and research, and by developers as the basis for agentic workflows, coding assistants and retrieval systems; its long context and strong reported performance on coding benchmarks made it a common choice for software engineering tasks. The subscription pause of July 2026 illustrated the gap between open-model popularity and serving capacity, as demand outstripped the company's compute supply ahead of its fundraising and listing plans.[1][4]
Moonshot's releases intensified competition in the open-weight segment that also includes DeepSeek and Alibaba's Qwen, putting pressure on the pricing and moat assumptions of closed frontier laboratories. Commentators have framed K3 in particular as evidence that Chinese developers can match near-frontier capability while giving away the weights, and major hosting and data platforms began offering the model soon after release.[2][7]
>See Also
Malaysian developers and enterprises access Moonshot's technology in two main ways: through the Kimi assistant and API, and by self-hosting the open-weight K2 and K3 models. Self-hosting is attractive to organisations that want capable models without per-token fees or routine data transfers to overseas APIs, and Malaysian research groups and startups have used open models of this class in natural-language processing work alongside domestic efforts such as ILMU and MaLLaM. Running frontier-scale models locally still requires substantial GPU infrastructure, so most Malaysian teams combine cloud-hosted API access with smaller self-hosted variants.[3][7]
For regulated organisations, any use of Kimi services raises considerations under the Personal Data Protection Act 2010 (PDPA), since prompts and outputs are processed on servers outside Malaysia and may contain personal data; the National AI Office and the Malaysia Digital Economy Corporation encourage evaluation of data residency, provenance and security before adopting foreign AI platforms. Malaysia's deepening technology and data-centre ties with China have made Chinese model stacks a practical part of the local landscape, even as governance frameworks for foreign AI services continue to develop.[4][6]
References
- ↑Moonshot AI. (2026). Kimi K3: open frontier intelligence (tech blog). https://www.kimi.ai/blog/kimi-k3
- ↑Bloomberg. (2026). Moonshot AI surpasses funding goal to hit $35 billion value. https://www.bloomberg.com/news/articles/2026-07-29/china-s-moonshot-ai-passes-funding-goal-to-hit-35-billion-value
- ↑TechCrunch. (2026). China's Moonshot AI raises $2B at $20B valuation as demand for open source AI skyrockets. https://techcrunch.com/2026/05/07/chinas-moonshot-ai-raises-2b-at-20b-valuation-as-demand-for-open-source-ai-skyrockets/
- ↑Reuters. (2026). China's Moonshot pauses Kimi subscriptions amid hot demand, IPO push. https://www.reuters.com/legal/transactional/chinas-moonshot-pauses-kimi-subscriptions-amid-hot-demand-ipo-push-2026-07-20/
- ↑HPCwire. (2025). China's Moonshot AI releases trillion parameter model Kimi K2. https://www.hpcwire.com/2025/07/16/chinas-moonshot-ai-releases-trillion-parameter-model-kimi-k2/
- ↑The Wire China. (2025). Who is Moonshot AI? https://www.thewirechina.com/2025/09/28/who-is-moonshot-ai/
- ↑Scientific American. (2026). China's Kimi K3 and the rise of open-weight AI models. https://www.scientificamerican.com/article/china-kimi-k3-and-the-rise-of-open-weight-ai-models/