๐ญ๐ฐSCMP TechnologyโขStalecollected in 1m
Moonshot AI Targets $30 Billion Valuation in Funding Round
๐กUnderstand the capital scale and valuation trends of China's leading LLM developers in the current AI race.
โก 30-Second TL;DR
What Changed
Moonshot AI is raising between US$1 billion and US$2 billion.
Why It Matters
This massive valuation signals continued investor confidence in Chinese LLM developers despite geopolitical headwinds. It suggests that capital-intensive model training remains the primary focus for top-tier Chinese AI startups.
What To Do Next
Monitor Moonshot AI's API documentation for updates to their long-context models to see if they offer competitive advantages for your data-heavy applications.
Who should care:Founders & Product Leaders
Key Points
- โขMoonshot AI is raising between US$1 billion and US$2 billion.
- โขThe new valuation target is 50% higher than its previous US$20 billion valuation.
- โขThe funding reflects the aggressive capital race among Chinese AI firms.
๐ง Deep Insight
Web-grounded analysis with 22 cited sources.
๐ Enhanced Key Takeaways
- โขMoonshot AI's annual recurring revenue (ARR) surpassed $200 million in April 2026, doubling from $100 million in March, primarily from Kimi chatbot subscriptions and API usage.
- โขThe company is reportedly dismantling its offshore corporate structure to prepare for a potential initial public offering (IPO) on the Hong Kong Stock Exchange.
- โขThis latest funding round marks Moonshot AI's third financing in six months, demonstrating an aggressive capital acquisition strategy in China's competitive AI landscape.
- โขMoonshot AI's valuation has seen a seven-fold increase from just over $4 billion in December 2025 to $20 billion after a Meituan-led round in May 2026, before targeting $30 billion in the current talks.
- โขMoonshot AI was founded in March 2023 by Yang Zhilin, Zhou Xinyu, and Wu Yuxin, all Tsinghua University alumni, with Yang Zhilin previously working at Meta AI and Google Brain.
๐ Competitor Analysisโธ Show
| Competitor | Valuation (approx.) | Key Products/Features |
|---|---|---|
| Moonshot AI | $20B (May 2026, targeting $30B) | Kimi Chat (chatbot), Kimi K2.6 (1T MoE, 256K context, Agent Swarm, multimodal), Kimi Code |
| Zhipu AI | ~$80B (as of June 2026) | GLM-5 model, chatbot tools for businesses |
| MiniMax | ~$20B (as of June 2026) | M2 model (top open model on Artificial Analysis's overall intelligence index), multimedia tools (AI companions, video generators) |
| DeepSeek | ~$10B (April 2026, seeking $50B in debut round) | DeepSeek-R1, DeepSeek-V3.2 (open-weight models), DeepSeek Sparse Attention, low training cost |
| Alibaba (Qwen) | N/A (part of larger Alibaba Group) | Qwen series (open-source LLMs), Qwen3 Max Thinking (claimed to outperform US rivals) |
๐ ๏ธ Technical Deep Dive
- Model Architecture: Kimi K2.6 is a 1-trillion parameter Mixture-of-Experts (MoE) model, with 32 billion active parameters per token. It uses 384 experts, activating 8 per token (plus 1 shared expert) across 61 layers.
- Context Window: Kimi K2.6 supports a 262,144 token (256K) context window, utilizing Multi-Head Latent Attention (MLA). Earlier versions like Kimi (original) supported 128,000 tokens.
- Multimodality: Kimi K2.5 and K2.6 are native multimodal agentic models, trained on 15 trillion+ visual and textual tokens from the start. They accept text, code, images, and video inputs via the 400M-parameter MoonViT encoder.
- Agentic Capabilities: Kimi K2.6 features an "Agent Swarm" system that scales to 300 domain-specialized sub-agents, capable of executing up to 4,000 coordinated steps in a single autonomous run. It supports long-horizon autonomous execution, including continuous runs of over 12 hours with thousands of tool calls.
- Coding Focus: Kimi K2.6 is optimized for software engineering across Rust, Go, and Python, handling tasks from front-end generation to DevOps and performance optimization. It can transform text prompts and visual mockups into production-ready interfaces.
- Training & Optimization: Moonshot AI uses a custom optimizer called MuonClip (featuring QK-Clip weight clipping) to stabilize training at the trillion-parameter scale.
- Operational Modes (Kimi K2.5): Instant for fast responses, Thinking for step-by-step analysis, Agent for autonomous workflows with 200-300 tool calls.
๐ฎ Future ImplicationsAI analysis grounded in cited sources
Moonshot AI's aggressive fundraising will intensify the global AI arms race, particularly in the development of large language models and agentic AI.
Securing substantial capital allows Moonshot AI to invest heavily in GPU capacity and accelerate model training, directly challenging both domestic and international competitors.
The company's move towards a Hong Kong IPO signals a strategic shift to secure diverse capital and potentially establish a more stable long-term funding mechanism outside of private rounds.
Dismantling its offshore structure and exploring a Hong Kong listing indicates a deliberate effort to navigate regulatory landscapes and access public markets for sustained growth.
Moonshot AI's focus on advanced agentic capabilities and multimodal models will drive innovation in complex, multi-step AI applications, potentially leading to more autonomous AI systems.
Models like Kimi K2.6 with Agent Swarm technology and long-horizon execution capabilities are designed for sophisticated tasks beyond simple chat, pushing the boundaries of AI utility.
โณ Timeline
2023-03
Moonshot AI founded by Yang Zhilin, Zhou Xinyu, and Wu Yuxin.
2023-10
Kimi chatbot officially released and began closed-beta testing, supporting 128,000 tokens of lossless context.
2024-02
Alibaba Group led a $1 billion funding round, valuing Moonshot AI at $2.5 billion.
2025-07
Moonshot AI released Kimi K2, a 1 trillion parameter Mixture-of-Experts (MoE) model.
2026-01
Moonshot AI released Kimi K2.5, a 1 trillion parameter MoE model with native multimodal capabilities and Agent Swarm technology.
2026-05
Moonshot AI raised approximately $2 billion in a round led by Meituan's venture arm, valuing the company at over $20 billion.
๐ Sources (22)
Factual claims are grounded in the sources below. Forward-looking analysis is AI-generated interpretation.
- economictimes.com
- techinasia.com
- mlq.ai
- pulse2.com
- japantimes.co.jp
- rootdata.com
- techfundingnews.com
- wikipedia.org
- businessmodelcanvastemplate.com
- llm-bento.com
- deepinfra.com
- kimik2ai.com
- nvidia.com
- punchng.com
- scmp.com
- wikipedia.org
- understandingai.org
- codecademy.com
- medium.com
- wikipedia.org
- intuitionlabs.ai
- aibase.com
๐ฐ
Weekly AI Recap
Read this week's curated digest of top AI events โ
๐Related Updates
AI-curated news aggregator. All content rights belong to original publishers.
Original source: SCMP Technology โ