Post-2000s Generation from Tsinghua University Announce Financing as a Team

The protagonist of this financing round is LiberAI (Beijing Jiangxian Technology Co., Ltd.), a "physical world model" company founded in December 2025. After HSG entered the angel round in March 2026, it successively added four additional investments in just a few months, completing the rare pace of "four consecutive investments from the angel round" for the same company.

ZhenFund, which also increased its stake continuously in the same period, has made a total of three consecutive investments starting from the seed round. Shunwei Capital quickly became the lead investor in the Pre-A round after its first move in the angel+ round. Multiple first-tier institutions have intensively bet on the same target in a very short period of time, which fully demonstrates the high popularity of this track.

The founder of this company is Liu Songming, a post-2000 PhD candidate in the Department of Computer Science of Tsinghua University. He once won the Tsinghua University Special Scholarship (only 10 people are selected each year) and has ranked first in his grade for many years. His supervisor is Professor Zhu Jun, a scholar in the field of video generation models. His research directions cover artificial intelligence, embodied intelligence and world models, and he has published many first-author papers at top conferences such as ICML and NeurIPS.

Liu Songming is undoubtedly the top-tier profile of the "young geniuses" that VCs are chasing right now. This is not an isolated case recently. In 2026, a group of post-2000s entrepreneurs from Tsinghua University are queuing up to announce their financing: Hu Yucheng's LatentVerse, Qin Shentao's OriginFlow, Min Yuheng's Zero-Power Robotics, Tang Jiankai's Microvision Tech, Wang Chen Yuhang's PrismLight Innovation, Xue Chuanyi's Singularity Escape, and many others.

The resumes of this group of people are surprisingly similar: born after 2000, from Tsinghua University, currently studying or just graduated not long ago, first author of papers at top academic conferences, and their supervisors are leading figures in the industry. Stories of young people being chased by capital happen every year, but it has never happened before that profiles as precise as "post-2000s from Tsinghua system" appear in such an organized and large-scale way like this year.

Less than one year, 6 rounds of financing totaling 1 billion yuan

Liu Songming being targeted by top-tier institutions is not just because he has a glamorous resume. What is more convincing is that before founding LiberAI, he had already delivered industry-recognized landmark achievements on the highly competitive track of "embodied intelligence foundation model":

RDT-1B released in 2024 is the world's first robot foundation model using the paradigm of "large-scale pre-training + diffusion Transformer", which solves the problems of data scarcity and multimodal behavior distribution in dual-arm robot operation, leading the PI-0 model of Silicon Valley company Physical Intelligence by about three weeks. In 2026, RDT-2, the earliest embodied model using the large-scale UMI (unembodied human data) pre-training paradigm, was released.

What these two papers prove is not product capability, but direction judgment. In an industry where technology has not yet converged, direction judgment is precisely the most valuable asset, because it determines which path all subsequent capital will take.

The core proposition of LiberAI is that the root cause of the insufficient generalization of current embodied models lies in the lack of modeling of force and causal relationships. If the model cannot understand that force is the cause and motion state is the effect, it cannot complete fine long-range physical operations like humans. Focusing on this point, the company has put forward two concepts: modal alignment and inductive bias. The former uses the generalization of Internet videos to feed back the scarce real machine data, while the latter directly writes physical laws into the model structure.

Faced with the technological paths that have not yet converged, top-tier institutions do not wait for products. As long as they confirm that the founder's academic judgment and engineering delivery capability have passed the initial screening, they can lock in positions with extremely high-frequency additional investments to avoid losing the qualification to enter the market in subsequent rounds.

The financing rhythm also confirms this kind of position locking. In less than a year, the company has completed 6 rounds of financing, totaling more than 1 billion yuan, and its valuation has entered the range of several billion yuan. The specific rounds are as follows:

In December 2025, ZhenFund exclusively invested in the seed round; in March 2026, HSG entered the angel round; in May, the angel+ round of about 500 million yuan was jointly participated by four parties: ZhenFund, HSG, Shunwei Capital, and Longzhu Capital of Meituan; in June, the Pre-A round of several hundred million yuan was led by Shunwei Capital, with Cathay Innovation, Yuanhe Origin, and Muhua Tech Innovation participating as followers; in July, the Pre-A+ round saw 360, China Development Bank Capital, Sinovation Ventures, CMC Capital, and Binfu Capital enter the market; in September, the Pre-A++ round of several hundred million yuan welcomed new investors including Huaying Capital, Zhongguancun Science City, Dacheng Capital, and Gaoxin Capital.

A group of "post-2000s Tsinghua entrepreneurs" complete intensive financing

In addition to Liu Songming's LiberAI as a representative, a group of post-2000s Tsinghua entrepreneurs are queuing up to officially announce their financing.

The company that started earlier than LiberAI is located in Hefei. Zero-Power Robotics was spun out from the AI&Robot Intelligent Robotics Laboratory of Tsinghua University in January 2025. Its founder Min Yuheng was born in 2000, and he studied in the Mingyue Lake Science and Innovation Experimental Class of Chongqing University for his bachelor's degree, which is a project incubated by Li Zexiang's system, before going to Tsinghua University to pursue a master's degree in artificial intelligence.

This company completed 5 rounds of financing in a year and a half: tens of millions of yuan in the angel round in February 2025; angel+ in May of the same year, angel++ in July, with Baohe Venture Capital, Tongchuang Weiye, Lihe Tech Innovation, and Shuimu Fund entering the market one after another; the Pre-A round of over 100 million yuan in April 2026, led by Runze Group.

Zero-Power Robotics completed 4 rounds of financing in one year and two months: it completed the angel round of tens of millions of yuan in February 2025, with investors including Hefei Seed Fund, Baohe District Pilot Fund, etc.; it completed the angel+ round of tens of millions of yuan in May 2025, with Lihe Venture Capital, Tongchuang Weiye, and Baohe Venture Capital entering the market; it completed the angel++ round of hundreds of millions of yuan in July 2025, with Shuimu Fund, Mizuho Lihe, and Ralph Venture Capital adding positions. In April 2026, Zero-Power Robotics completed another A-round financing of over 100 million yuan, led by Runze Group, with Ningbo Dongli, Relay Angel, and Pinghu Zexin participating as followers.

What is rare among post-2000s startups is that less than two years after its establishment, both academic papers and revenue from operations can be seen. According to reports, its robots have been deployed at Hefei Xinqiao International Airport and the OLE supermarket in Shenzhen MixC, with order value exceeding 100 million yuan.

LatentVerse proves another thing: when the technical route is not yet determined, the founder's supervisor is the core value of the target. Founder Hu Yucheng is a PhD student at the Institute for Interdisciplinary Information Sciences of Tsinghua University, and his supervisor is Chen Jianyu, the founder of StarMotion Era. The company was established in May 2026, and in August it officially announced a seed round of several hundred million yuan, with GL Ventures, Qingliu Capital, Innoangel, and Agibot Robotics entering the market.

Qin Shentao of OriginFlow pushed this valuation logic to a more extreme level. He was born in 2001, completed his bachelor's degree at Harbin Institute of Technology, and is pursuing his PhD at Tsinghua University. Since this spring, BlueRun Ventures, Oasis Capital, 58 Industrial Investment, and Lisi Capital have entered the market one after another, completing three rounds of financing in five months, totaling more than 500 million yuan. BlueRun led the angel round and continuously increased its stake for three rounds. Zhu Tianyu, its managing partner, later recalled that it only took 30 minutes for them to decide to invest in him.

Not every post-2000s entrepreneur is in the top leadership position. Lingsheng Tech officially announced a 100-million-yuan level A-round financing on October 8, jointly invested by Huafang Capital, Daohe Yuanqi, Suwen Electric Energy, and a leading industrial player in the embodied intelligence field. The post-2000s talent in this company serves as CTO: Jiang Yuhua, who completed his bachelor's degree in Tsinghua University Qian Xuesen Honors Class and his PhD in the Department of Automation. He won the Tsinghua University Special Scholarship which is awarded to only 10 people every year, and also won the Best Paper Award of IEEE WCNC 2024.

The list can go on. The founding team of Vangt AI is all made up of post-2000s Tsinghua talents, focusing on end-side inference engines and modules. 5Y Capital and Frees Fund invested in its angel round and angel+ round; Microvision Tech completed a financing of several million US dollars, founder Tang Jiankai is a PhD candidate at Tsinghua University's Xinya College and Department of Computer Science, who has long been deeply engaged in machine vision and physiological computing, and has published more than 20 achievements in top journals such as Science Advances. The founding team of PrismLight Innovation is all post-2000s, CEO Wang Chen Yuhang completed his bachelor's and master's degrees in the Department of Electronic Engineering of Tsinghua University, specializing in robot tactile sensing; Xue Chuanyi of Singularity Escape is a Tsinghua University direct-track PhD candidate, first author of a NeurIPS paper, and completed a 10-million-yuan level seed round in August, jointly led by StarLink Capital and Shuimu Venture Capital.

There is another company that needs to be mentioned separately. Fu Zhi, the founder of Gongji Tech, was admitted to Tsinghua University in 2018 and later became a Tsinghua University Person of the Year. His business of computing power scheduling is called "DiDi in the computing power industry". This company was established as early as 2023, with initial capital of 300,000 US dollars from personal investment of Lu Qi. Now it has accumulated financing of more than 20 million yuan, with a valuation of nearly 400 million yuan, and its annual revenue in 2025 was 30 million yuan. It was two years ahead of this wave of boom, and the market value rose sharply this year.

Of course, apart from Tsinghua University, post-2000s entrepreneurs from Peking University, University of Science and Technology of China, Shanghai Jiao Tong University and other universities also appear in the financing list of tracks such as embodied intelligence, world models, AI Infra and end-side inference, but their density is not as high as that of the Tsinghua system — Tsinghua is the most prominent cluster in this wave, but not the whole picture.

The admission ticket has changed, and the rules have also changed

Funds that formulate strategies around this group of post-2000s Tsinghua entrepreneurs have also achieved extremely outstanding returns.

I learned about an AI fund that specifically formulates strategies around Tsinghua University, which has become one of the fastest-growing VCs in this wave. It does not pursue full-track coverage, but takes Tsinghua laboratories, supervisor networks and alumni startup pools as its project sources, invests in the first round with small amounts and high frequency, and uses coverage density to exchange for hit rate. The internal priority of evaluating founders is also changing: first look at papers and open source projects, then look at supervisor endorsement and laboratory status, and finally look at the business plan. Founded on the eve of the AI era, this strategy has allowed it to capture multiple unicorn targets, and its management scale has expanded rapidly this year.

Let's look at the recent development of Shuimu Tsinghua Alumni Seed Fund, the first domestic university alumni fund that put forward the slogan of "writing the first check for Tsinghua alumni".

As of the first half of this year, it has invested in more than 150 companies in total, and the proportion of its role as the first institutional investor exceeds 60%. In the first half of 2026 alone, it has invested in more than 60 projects, 70% of which are the first round. In June this year, its Shanghai fund completed fundraising of 535 million yuan, and a growth-stage fund with a scale of 1 billion yuan has also been launched, which is planned to complete fundraising before the end of the year.

Innoangel Fund was initiated by Tsinghua alumni Li Zhu, Liu Huaiyu and others in 2013, and also takes the "Tsinghua system" as its distinct label. More than 60% of the project founders in its second science and technology innovation fund come from Tsinghua University. In 2026, the third Innoangel Science and Technology Innovation Fund completed its first close with a scale of 1.5 billion yuan.

In addition, Qiming Venture Partners even launched a Y Transformer, which only invests in founders born after 1998, only invests in the first round, with a single investment of about 600,000 US dollars. And in the same wave of stake grabbing, according to a report by LatePost, some institutions even sent investment intention letters with the amount field left blank, which means "you can set the terms as you like".

Capital has never been so densely gathered towards the same group of top young people as it is now. This naturally leads to three questions: Why post-2000s? Why now? Why Tsinghua University?

First of all, the admission ticket for entrepreneurship has changed. In the last wave of Internet entrepreneurship, business models, channels, supply chains and management radius were valued, and the primary market was popular for investing in serial entrepreneurs and senior executives from large factories. These rules were unfavorable for young people at that time. However, embodied intelligence and world models have reversed the rules, and the valued capabilities have been transformed into academic papers, competition achievements, foundation models and supervisor endorsements.

Antler's report shows that the average age of AI unicorn founders has dropped from 40 in 2020 to 29 in 2024. This group of young people has no path dependence, and they are natively learning Transformer in their academic studies.

Secondly, the definition of production efficiency has also changed. The AI era is not a competition for talents, but a competition for geniuses. The top 1000 people in this industry have made more than 95% of the contributions. Why are capital willing to give a 20-something PhD student a valuation of 1 billion yuan? Because at this stage, the scarcest resource is the judgment of technical routes. If a top researcher chooses the right direction, his efficiency may be ten times or even a hundred times that of others.

Another important variable is the cost structure. In the past, to set up a technology company, you had to build production lines, lay channels, and hire sales teams; now a 10-person team can rent computing power, connect APIs, release a model on GitHub and X, and get global developers and orders within a week. Model is product, model is channel. The fixed cost of entrepreneurship has dropped to a level that PhD students can afford.

When the three factors of people, efficiency and cost reverse at the same time, the whole entrepreneurship playbook changes.

When the playbook changes, the first rule of investment — investing in people — has been amplified to an unprecedented level. Behind this lies a layer of industrial anxiety: in this round, only large models have truly formed a closed loop, and technologies in fields such as embodied intelligence, world models, quantum technology and nuclear fusion have not yet converged. On the demand side, hundreds of embodied intelligence companies and brain-computer interface companies are competing on the same field, and no one dares to say which route will definitely win; on the capital side, investment has been fully advanced, with valuations and listings all moving forward in schedule. The collision of uncertain technology and certain capital leads to only one conclusion: when the route cannot be bet accurately, people become the only anchor.

Of course, with the influx of hot money, the risk of bubbles is also brewing.

A post-2000s PhD entrepreneur from a top university who obtained multiple rounds of financing this year told me that when he was raising money, the investor replied to him: "The amount of money you want is too little, I can give you twice as much, but you need to tell a larger business story, otherwise I will not invest." This forced him to take the path with higher returns but also higher risks.

This is not the situation of him alone. When the valuation is supported by "the next person who is willing to take over" rather than by delivery and repurchase, the faster the financing speed, the faster the bubble accumulates. Blindly pursuing high valuation is essentially overdrawing the growth space of the next three to five years in advance. The more people chasing it, the wider the gap between the price and the fundamentals will be. The most dangerous part of the bubble is not when it will burst, but that before