Moonshot AI targets $1b raise, eyes $18b valuation

Moonshot AI is seeking to raise as much as US$1 billion in an expanded funding round that would value the Chinese startup at about US$18 billion, people familiar with the matter said.

Moonshot AI, the developer of the Kimi chatbot, began talks after raising more than US$700 million earlier this year at a US$10 billion valuation, people familiar with the matter said.

Backers including Alibaba, Tencent, and 5Y Capital increased their bets at the US$10 billion level, Bloomberg News reported last month.

Moonshot launched Kimi Claw, powered by its Kimi K2.5 model, and, according to one of the people, monthly sales after the launch exceeded last year’s total revenue.

Rivals Zhipu and MiniMax have recently traded at valuations between US$30 billion and US$40 billion in Hong Kong.

🔗 Source: Bloomberg

🧠 Food for thought

Implications, context, and why it matters.

Moonshot valuation leans on performance claims and real-world usage

Moonshot AI’s Kimi K2.5 model, described as powering Kimi Claw, is billed as a leading open model, with one index ranking it as the strongest model not made by OpenAI, Google, or Anthropic 1.
Venture capitalist Chamath Palihapitiya said his firm 8090 has shifted many workloads to Kimi K2 on Groq because it was “way more performant” and “a ton cheaper” than OpenAI and Anthropic 1.
Use extends beyond Moonshot’s own apps, with Chinese models from firms such as Moonshot making up nearly two thirds of token usage among the top five models on OpenRouter in a recent ranking 2.
On the technical side, the model has more than 1 trillion parameters, yet that scale makes it hard for many teams to run on their own hardware 1.

Moonshot rapid rise also faces allegations of illicit access and distillation

Anthropic alleges Moonshot generated over 3.4 million exchanges with Claude using hundreds of fraudulent accounts 3.
Anthropic also says the activity aimed to pull out and rebuild Claude’s reasoning traces, with request metadata that matched public profiles of senior Moonshot staff 3.
Similar claims that also name DeepSeek and MiniMax have stirred debate over whether rapid gains came from independent work or from capabilities copied from US models, with Anthropic arguing extraction can deliver an edge in a fraction of the time and cost 3.
Anthropic argues distillation attacks can weaken export controls by pulling capabilities from API outputs at scale, while still needing access to advanced chips, and this remains Anthropic’s framing rather than an established finding 3.