Kimi Secures Over $3.5 Billion in Funding, Valuation Rises to $35 Billion, Pre-IPO Round Launched Early

Original Title: "Kimi Secures Over $3.5 Billion in Funding, Valuation Rises to $35 Billion, Pre-IPO Round Launched Early"
Original Author: Dongcha Beating

Kimi has just released the model weights for K3, and the latest round of funding for Moonlight has also been finalized.

According to exclusive information from the STAR Market Daily, Moonlight has completed its Series F funding round, raising over $3.5 billion, with a post-money valuation reaching $35 billion.

This round of funding was not initially planned to raise so much money. Reports indicate that due to investor subscriptions exceeding the original target by more than three times, Moonlight closed its Series F round early. The Series G funding, originally scheduled to start in August, has also begun ahead of schedule.

The Series G round will be the Pre-IPO round for Moonlight, with a pre-money valuation already rising to $50 billion.

A week ago, market rumors suggested that Moonlight would complete a round of funding with a pre-money valuation of about $31.5 billion before launching its final private placement in August. Now, with Series F ultimately raising over $3.5 billion and a post-money valuation of $35 billion, the next round will not wait until August. Both the money and valuation have come faster than originally planned.

Since its establishment in 2023, Moonlight has increased its valuation from $30 million to $35 billion in just over three years. The ongoing Pre-IPO round has already set the next price tag at $50 billion.

Over Ten Rounds of Financing in Three Years,

Valuation Rises from $30 Million to $35 Billion

Moonlight was founded in April 2023 by Yang Zhiling, Zhang Yutao, Zhou Xinyu, and Wu Yuxin.

About two months after its establishment, it completed an angel round of financing exceeding $200 million, with a post-money valuation of approximately $30 million. For a startup focused on a product that has not yet been officially launched, this was already a rare early-stage financing.

What truly brought Moonlight to the center of the capital table was the A+ round of financing completed in February 2024.

This round raised over $1 billion, led by Alibaba, with participation from Sequoia China, Xiaohongshu, Meituan, and other institutions and industrial capital, pushing Moonlight's valuation to about $2.5 billion. Six months later, the Series B financing exceeded $300 million, and the company's valuation continued to rise to $3.3 billion.

However, after that round of financing in 2024, Moonlight's fundraising pace slowed down for a while.

During that year, the Chinese large model market underwent significant changes. Major companies like ByteDance and Alibaba continued to lower model prices, while DeepSeek rapidly rose with open-source and cost efficiency. The competition in general chat products gradually shifted from user growth to model capabilities, inference costs, and commercial revenue. Kimi had once gained attention with long text capabilities, but relying solely on one product label made it difficult to sustain a higher valuation.

It wasn't until the end of 2025 that Moonlight completed a $500 million Series C financing, with a post-money valuation reaching $4.3 billion. After that, the company's fundraising clearly accelerated.

In the first two months of 2026, Moonlight completed multiple rounds of financing, with its valuation rising from the previous $4.3 billion to $10 billion, and further reaching $18 billion.

In May this year, the company completed approximately $2 billion in Series D financing, with a post-money valuation reaching $20 billion. Participants were no longer limited to internet companies and market-oriented investment institutions; China Mobile, Guozhi Investment, CPE Yuanfeng, and several state-owned funds also began to enter the shareholder list.

Just after completing the Series D round, a new round of financing was launched in June. At that time, market rumors indicated that the pre-money valuation had reached $31.5 billion. Now, this round has ultimately closed with over $3.5 billion in funding, raising Moonlight's post-money valuation to $35 billion.

Looking back at this financing curve, the most noticeable change occurred in the past six months.

At the end of 2025, Moonlight's valuation was still $4.3 billion. More than half a year later, this figure has reached $35 billion, an increase of over seven times. If the next round is completed with a pre-money valuation of $50 billion, Moonlight's valuation increase over just over three years will exceed 160 times.

The reasons given by capital are becoming increasingly direct. In the past, investors bet on Yang Zhiling and a technical team from Tsinghua University; now they are betting on whether Moonlight can become one of the few companies in China that remain at the forefront of global model development.

After K3, Moonlight Returns to the Center of the Table

The timing of this round of financing closing early coincided perfectly with the release of Kimi K3.

On July 16, Moonlight released Kimi K3. On July 27, the company further opened the complete model weights, technical reports, and some key infrastructure technologies for K3.

K3 is a mixed expert model with a total parameter count of 28 trillion, activating 104 billion parameters per inference, supporting 1 million tokens of context, and possessing capabilities in vision, programming, reasoning, and long-term tool invocation.

In its technical report, Moonlight stated that K3 has reached the forefront level in tasks such as long-range programming, agents, knowledge, reasoning, and vision. Its overall capabilities are now very close to the strongest closed-source models, Claude Fable 5 and GPT-5.6 Sol.

Rankings can change at any time, and model companies must also keep an eye on their own testing. The more important impact of K3 is that it pushes the boundaries of the capabilities of open-weight models further forward.

For a long time, open-source models have primarily played the role of catch-up. Closed-source companies are responsible for training the strongest models, and six months or a year later, the open-source community replicates some of those capabilities at a lower cost.

K3 changed this time gap. A Chinese company has directly opened a model with capabilities close to the global forefront, allowing other developers to download, modify, and deploy it, as well as continue training their own models based on it.

Of course, the model has not become cheaper as a result. The weight files for K3 exceed 1.5 TB, and even loading the complete model requires multiple high-end accelerator cards. The hardware investment for production environment deployment may be quite high. Its "openness" primarily addresses whether enterprises can master the model and escape reliance on a single API supplier, and does not mean that everyone can run it on their home computers.

Even so, open weights will still impact the business of closed-source companies.

When a sufficiently powerful model can be downloaded, enterprises have the opportunity to deploy it on their own servers, keep data internal, customize the model, and no longer have to pay according to the token prices of a particular company. The barriers built by closed-source models based on their leading capabilities will also be diluted more quickly.

Thus, after the release of K3, the debate quickly transcended technical rankings and entered the policy and commercial interest disputes within the U.S. AI industry.

Behind the Open Source and Closed Source Debate

On July 24, companies and institutions including OpenAI, Google, Microsoft, NVIDIA, AMD, Meta, and Hugging Face publicly supported open-weight models and opposed the U.S. government's eagerness to restrict AI models available for download and deployment.

Anthropic and Amazon did not appear on the signing list. This led to external speculation about whether Anthropic hoped to limit the development of open models under the guise of safety and national security, in order to shield its closed-source business from competitors.

On the same day that K3's weights were opened, Anthropic CEO Dario Amodei published an article responding to the controversy.

He stated that Anthropic has never advocated for a complete ban on open-weight models, and that open models without dangerous capabilities are a public good that can lower usage costs and create value for enterprises, developers, and researchers.

However, Amodei's core stance has not changed.

He still advocates for continued restrictions on the flow of advanced chips and chip manufacturing equipment to China, cracking down on so-called "industrial-scale model distillation," and requiring sufficiently powerful models to undergo mandatory safety testing before release, regardless of whether these models are open or closed source.

On the surface, this debate revolves around safety, but underneath it lies a very real business issue.

Closed-source companies wish to continue controlling the strongest models, charging fees through APIs and subscription services; open models allow capabilities to spread to more companies, lowering prices and shortening the distance between newcomers and leading companies. The faster the opening, the harder it becomes for a few companies to keep model capabilities locked in their servers for the long term.

K3 comes from China, is large enough, and its capabilities are already close to the forefront. For supporters of open models, K3 proves that the open route can still produce top-tier models; for Anthropic, it also means that the capability advantage built by American companies at great expense may spread more quickly.

As a result, Moonlight has gained a capital value it did not have before.

It is no longer just an AI application company with a large number of Chinese users, nor merely a foundational model startup waiting for its business model to mature. After the release of K3, it began to be revalued within the context of global open-source versus closed-source, the U.S. versus China, and the competition between model capabilities and computational power control.

The $3.5 billion funding and $35 billion valuation are the answers provided by the capital market.

Original Link