China’s moonshot AI rises to $2.5 billion valuation, raising $1 billion for an LLM focused on long-term contexts

The race is on to construct the following big language model, and now a competitor from China has reportedly raised a big round of funding to catapult it to the highest.

Moonshot AI, a man-made intelligence startup founded lower than a yr ago that develops LLMs that may process long text and data inputs, has raised over $1 billion in a Series B round, in line with multiple media reports from China. If true, this latest capital injection would value Moonshot AI at $2.5 billion – the most important single funding round for Chinese LLM developers in recent memory.

The startup – known in China as YueZhiAnMian – has, like many within the AI ​​industry at once, focused on developing large language models. In particular, its unique selling point is that it really works to process long contexts and responses, an area that has long outsmarted others in the sphere.

The first steps to unravel this problem were quickly taken.

In March last yr – coinciding with the fiftieth anniversary of Pink Floyd’s “The Dark Side of the Moon”, founder Yang Zhilin’s favorite album and inspiration for the startup’s name – the startup launched with a 100 billion parameter LLM.

In October, Moonshot launched its first chatbot in China. Demwhich claims to have the ability to support processing 200,000 Chinese characters in a single conversation – reportedly eight times longer than what OpenAI’s GPT-4-32K can do.

We have made contact Yang Zhilinthe AI ​​researcher and academic who founded Moonshot together with Zhou Xinyu and Wu Yuxin for a comment and can update this post if he responds.

Meanwhile, the reports state that the funding comes from a listing of well-known investors, including a variety of potentially interesting strategic partners. The round is alleged to be led by e-commerce giant Alibaba and HongShan – the VC firm formerly referred to as Sequoia China South China morning post. Others within the group included Chinese “super apps” Meituan and Xiaohongshu (sometimes called China’s answer to Instagram), in line with Chinese tech blog LatePost.

According to details, Moonshot previously raised $200 million from HongShan and Zhen Fund in a round that was valued at $300 million PitchBook data.

Reached for comment, HongShan declined to comment on the reports. Alibaba didn’t reply to a request for comment. We also reached out to Moonshot individually.

If true, HongShan’s involvement here could be notable. Sequoia Capital formally announced last yr that it could spin off its Asian operations in India and China amid rising geopolitical tensions. This process needs to be finally accomplished by March twenty fourth.

But the China operation has now prevailed come to the test from the US government for its AI deals within the US; and the US company can also be under investigation for ongoing activities in China. Given all this, it’s no surprise that the once high-profile investor is keeping a low profile here.

Tellingly, the remaining of the reported investor list is a veritable who’s who of well-known technology corporations. This underscores the continued retreat – or a minimum of pause – of monetary investors in promising Chinese startups, particularly those from the West that do business in U.S. dollars.

But it also shows how – just as we have seen within the US, where corporations like Microsoft, Google and Amazon are pouring billions of dollars into LLM startups like OpenAI and Anthropic – big tech corporations in China are fighting for what could possibly be theirs big AI games in the approaching months and years. Financially supporting a handful of promising hopefuls is a technique to shorten or expand what they need to construct internally.

OpenAI currently reigns supreme within the US and arguably in every single place else it has expanded, but in China there is no such thing as a appointed leader, a lot of its investment activity appears to be about spreading bets.

To that end, Alibaba can also be an investor in Baichuan – founded by Xiaochuan Wang, a pioneer within the search engine space that raised $350 million by the top of last yr, reaching a valuation of $1 billion – and Zhipu AI one other LLM newcomer, and 01.AI, the LLM company founded by Kai-Fu Lee.

Alibaba’s arch-rival Tencent has now backed Baichuan, Zhipu, MiniMax and Light Years Beyond. Make no mistake: China’s web giants have replaced these Western enterprise capitalists in supporting the country’s LLM candidates.

Still, while $1 billion appears like loads of money to offer to a startup that is not even a yr old, pedigree is certainly one of the explanations big names is perhaps willing to make big bets .

Pink Floyd fan turned AI pioneer

Yang Zhilin already had an extended list of successes before founding Moonshot.

He has a doctorate in computer science from Carnegie Mellon University, where he was advised by Ruslan Salakhutdinov, who previously led AI research at Apple after the iPhone maker quietly acquired a startup he founded called Perceptual Machines – an acquisition that which was apparently never reported on but is noted within the Professor’s LinkedIn profile and skilled schedule.

Previously, he studied at Tsinghua University, advised by Jie Tang. He has also worked at Google Brain and Meta AI.

Yang also has one other AI startup within the works: Recurring.AI, which appears to be specifically focused on technology designed to assist salespeople do their jobs higher (with features that sound not unlike Gong.AI). By 2021, Recurrent had reportedly raised around $60 million PitchBook. And while there hasn’t been much capital activity since then, the corporate appears to stay operational.

Importantly for Moonshot, Yang was also certainly one of the lead authors of Transformer-XL, a serious development in LLM activation architecture Understanding natural language beyond a fixed-length contextsomething that played a big role in the event of Moonshot’s platform and arguably its broader mission.

Moonshot’s deal with longer input and output and producing more accurate results for related queries provides the muse for the corporate to focus on text-based use cases not widely leveraged by existing LLMs and generative AI applications, equivalent to legal documents and fiction writing and deeper financial evaluation. Kimi Chat is ready for information until January 2024, says the chatbot.

It just isn’t the one Chinese actor working to remove the constraints of the long context. Baichuan announced its Baichuan2-192K model back in October process around 350,000 Chinese characters in a context window.

The global fundraising market stays limited, but this round shows that those with big budgets are able to step in when the correct opportunities arise. Yet even with the broader, global AI frenzy – where some 200 billion dollars needs to be invested by 2025, predicts Goldman Sachs – the financing landscape in China is surprisingly tepid.

According to research firm CBInsight, China recorded around 232 AI investments in 2023, down 38% year-on-year. The total amount raised by China’s AI firms was around $2 billion, down 70% from the yr before.