The competition to develop the next groundbreaking large language model (LLM) is heating up, and a new contender from China is generating buzz with a significant funding milestone. Moonshot AI, an innovative artificial intelligence startup launched less than a year ago, has reportedly secured over $1 billion in a Series B funding round, as reported by various Chinese media outlets. If confirmed, this funding round would elevate Moonshot AI’s valuation to $2.5 billion, making it the largest single funding round on record for Chinese LLM developers.
Operating under the name YueZhiAnMian in China, Moonshot AI is prioritizing the advancement of large language models. Its main focus is pioneering long-form context processing, a challenge that has perplexed many in the industry. The startup has rapidly initiated its ambitious projects to tackle this challenge.
In March of last year, timed with the 50th anniversary of Pink Floyd’s iconic album The Dark Side of the Moon—the inspiration behind its name—Moonshot launched a 100 billion-parameter LLM. By October, the company debuted its first chatbot in China, Kimi, boasting the ability to process 200,000 Chinese characters in a single exchange—an impressive feat that eclipses the 25,000 character limit of OpenAI’s GPT-4-32K.
We reached out to Yang Zhilin, the AI researcher and academic co-founder of Moonshot alongside Zhou Xinyu and Wu Yuxin, for comments and will provide updates if he responds.
Reports indicate that the funding comes from a selection of prominent investors, including e-commerce titan Alibaba and HongShan, the venture capital firm formerly known as Sequoia China, as highlighted by the South China Morning Post. Other notable participants include Meituan, a Chinese “super app,” and Xiaohongshu, often referred to as China's version of Instagram, according to the Chinese tech outlet LatePost. Bloomberg later confirmed that Alibaba led the $1 billion round with participation from Monolith Management.
Previously, Moonshot raised $200 million from HongShan and Zhen Fund at a valuation of $300 million, according to PitchBook Data. HongShan declined to comment on the new reports, and Alibaba has not yet responded to our inquiry, although we have reached out to Moonshot for further information.
If confirmed, HongShan’s participation is noteworthy, considering Sequoia Capital's announcement last year to separate its Asian operations amid escalating geopolitical tensions. This restructuring is expected to be completed by March 2024. During this transition, the China operation has faced scrutiny from U.S. authorities regarding its AI dealings, contributing to the cautious approach of this historically prominent investor.
The roster of investors is a distinguished lineup of leading tech names, reflecting a shift in financial backing for Chinese startups, as major U.S. firms like Microsoft, Google, and Amazon have made significant investments in LLM startups like OpenAI and Anthropic. Chinese tech giants are now racing to establish their own robust AI capabilities for future endeavors. This strategy allows them to strengthen their stakes in promising companies while continuing to develop their in-house solutions.
While OpenAI currently dominates the global LLM landscape, there is no definitive leader in China, leading many investors to diversify their portfolios within the sector. Alibaba, for example, is also investing in Baichuan, founded by Xiaochuan Wang—a notable figure in search engine technology, which raised $350 million last year and surpassed a $1 billion valuation—and Zhipu AI, among others.
Meanwhile, Alibaba’s competitor, Tencent, has invested in Baichuan, Zhipu, MiniMax, and Light Years Beyond, indicating that China’s internet conglomerates are now pivotal investors in the country's LLM market.
Despite the substantial funding, the $1 billion investment in a startup less than a year old is significant, reflecting the investors' confidence in the founding team's pedigree. Yang Zhilin boasts an impressive background, holding a PhD in computer science from Carnegie Mellon University where he worked under AI expert Ruslan Salakhutdinov, a former head of AI research at Apple. Prior to that, he studied at Tsinghua University under Jie Tang and has held positions at Google Brain and Meta AI.
Yang is also developing another AI venture, Recurrent.AI, aimed at enhancing sales efficiency, which raised approximately $60 million as of 2021, with the business still operational. Notably, Yang was a key contributor to Transformer-XL, a significant advancement in LLM architecture that enables natural language processing beyond fixed-length contexts, integral to Moonshot’s objectives.
Moonshot's emphasis on processing longer inputs and outputs paves the way for tapping into numerous untapped use cases for text-based applications, such as legal documentation, creative writing, and detailed financial analysis. Kimi Chat is reported to be trained on data up to January 2024.
Moonshot is not the only Chinese player focused on lengthy context processing; Baichuan announced its Baichuan2-192K model in October, which is said to handle around 350,000 Chinese characters in a single context window.
Although global fundraising remains tight, this round signifies a readiness among deep-pocketed investors to capitalize on promising opportunities. Despite a broader AI boom—projected to attract around $200 billion by 2025, according to Goldman Sachs—the funding climate in China appears comparatively sluggish. In 2023, China witnessed approximately 232 AI investments, marking a 38% decrease from the previous year, with total fundraising for AI firms around $2 billion—down 70% from 2022.
Ranking
