Keep the "Lobster" a secret. A post-00s doctor took a leave of absence from school to start a business and got financing again.
The outbreak of the "Lobster" security incident made Lin Xiuchun, the post - 2000s CEO of "Jinghua Secret Computing", clearly feel the change in the situation.
In recent weeks, investment institutions that were once somewhat unfamiliar with the concept of "confidential computing" have started to approach Lin Xiuchun actively. "Previously, we had to chase after others to explain what we were doing. Now, they are chasing us to ask if this thing can keep 'Lobster' data confidential," Lin Xiuchun said.
What surprised the team even more was that leading large - model manufacturers also came for exchanges. These "top - tier" players in the domestic AI circle made appointments with Lin Xiuchun. They not only wanted to learn about the technology from him but also wanted to know if this solution could be integrated into their products.
The dual enthusiasm from the capital market and the industrial end has brought more attention to this young company. Recently, China Venture Capital News exclusively learned that Jinghua Secret Computing has completed an angel + round of financing of tens of millions of RMB. The lead investor is Shengjing Jiacheng Venture Capital, and listed companies Guolian Co., Ltd., Boyan Technology, and old shareholder Inno Angel Fund participated in the follow - on investment.
Jinghua Secret Computing was established in September 2024 and focuses on the cutting - edge technology of high - performance AI confidential computing.
The concept of confidential computing may sound obscure. We can use an analogy to understand it: Traditional computing uploads plaintext data to the public network, allowing the server to "open - eyed" view your plaintext data. In contrast, confidential computing uploads ciphertext and an "operation manual" to the public network. It enables the server to perform blind calculations on a pile of ciphertext "blind - eyed" according to this operation manual. After the calculation, the result is sent back to the local, and you can decrypt it with the key. Throughout the process, the data remains in a ciphertext state in every link of transmission, storage, calculation, and the hardware environment, truly achieving data security.
The funds from this round of financing will be mainly used for the continuous R & D of the high - performance AI confidential computing platform, the iterative verification of confidential - state dedicated computing chips, and the accelerated commercialization in high - sensitivity scenarios of core data such as government, finance, medical, and legal fields.
A post - 2000s Peking University doctor suspends studies to start a business and secures two rounds of financing
Lin Xiuchun was born in 2000 and suspended his doctoral studies at Peking University to start a business. He chose to start a business at this time because the team had made a major technological breakthrough.
Professor Ren Ju, the technical advisor of Jinghua Secret Computing, is a leading figure in the industry. He is the director of the Institute of Human - Computer Interaction and Media Integration at the Department of Computer Science, Tsinghua University, and a national high - level talent.
Confidential computing used to have the nickname of "a dud that hasn't gone off in twenty years" because it was too slow. Blindly calculating a large amount of ciphertext according to an operation manual with an astronomical encryption level would cause a 1000 - to 10,000 - fold time loss in model training. However, the Jinghua Secret Computing team started to tackle this problem in 2019. They gave up the general - purpose computing function of confidential computing, specifically optimized the operators related to AI intelligent computing, and then carried out engineering transformation on the AI - related operators. Finally, they achieved an available level and realized "usable but invisible" with zero - precision loss in the entire process of data transmission, storage, and calculation.
In the past, the common solution in the industry was "trusted computing", which created an isolated "safe house" at the chip level through hardware such as NVIDIA GPU - TEE. However, this solution has two fatal limitations: Firstly, the technology is firmly bound to overseas graphics cards and cannot be compatible with domestic manufacturers. Secondly, the security is guaranteed by a single overseas manufacturer. 'Relying on others for the root of trust is itself a huge potential risk.'
The solution of Jinghua Secret Computing does not rely on NVIDIA's GPU - TEE at all and is fully compatible with domestic graphics cards. "Any domestic graphics card or a graphics card without TEE can run in our framework," Lin Xiuchun said. This means that from relying on overseas commercial hardware to relying on open cryptography standards and fully independent computing power, Jinghua Secret Computing has completely taken control of the "root of trust" for security in the future intelligent era.
This technology also enabled Jinghua Secret Computing to win the first place in a startup competition in 2025. It stood out among hundreds of technology teams from various industries across the country, including artificial intelligence, embodied intelligence, biomedicine, and aerospace, and thus came into the view of investment institutions on a larger scale.
Before participating in the startup competition, the company had just completed the seed round, and the investor was Inno Angel Fund. Lin Xiuchun introduced that at that time, the company had four or five term sheets, and some institutions offered a much higher valuation than Inno. However, he ultimately chose Inno.
The main reason behind this is that they are from the same academic lineage. "There are about a dozen classmates in our team from the Department of Computer Science, Tsinghua University, and several of them have won the '84 Scholarship'," Lin Xiuchun said. Entrepreneurs from the Department of Computer Science, Tsinghua University, have a natural sense of closeness to Inno Angel Fund.
In this round of financing, the most eye - catching are the two listed companies, Guolian Co., Ltd. and Boyan Technology. They are not just financial investors but also "upstream and downstream partners" bringing industrial resources and business synergy.
Guolian Co., Ltd. is a B2B e - commerce and industrial Internet platform company. It was established in 2002 and listed on the main board of the Shanghai Stock Exchange in 2019. It has a large number of physical enterprises undergoing intelligent transformation in its hands. These enterprises are potential customers for confidential computing.
Boyan Technology's main business is to help Chinese enterprises adapt their products to overseas markets, pass compliance reviews in various countries, and build local operation systems. In this process, the security issue of cross - border data flow is the most headache for customers. And this is exactly where Jinghua Secret Computing comes in handy.
This round of financing came faster than the team planned. The team originally wanted to wait for this year's commercialization before seeking financing to obtain a higher valuation. However, with the efforts of Shengjing Jiacheng, which brought in the two listed companies, Guolian Co., Ltd. and Boyan Technology, this round of financing was successfully completed.
Regarding this investment, Liu Haofei, the founding partner of Shengjing Jiacheng Venture Capital said: No matter how much technology advances, efficiency will always be restricted by security, and security capabilities must keep up with technological development. Therefore, when AI applications are in full bloom, it will also be the time when AI security measures are mature enough. Jinghua Secret Computing has significantly shortened the time of traditional confidential computing in the production environment, making it very likely that confidential computing will become the mainstream path for privacy computing in the era of general artificial intelligence.
Currently, the company has a clear path for commercialization: Firstly, for government and enterprises, it focuses on high - level local private - domain deployment, building a moat for confidential computing covering the entire process of 'input - inference - output'. Secondly, for developers and the C - end, it features a confidential - state inference platform and the "Lobster Security Guard", supporting plug - and - play across platforms and multiple terminals. Thirdly, for high - value private - domain data, it uses a confidential - state training engine to truly enable the data to be leased and sold in ciphertext form. Founder Lin Xiuchun revealed that the company is striding towards the industrial implementation of confidential computing.
"Get on the table and wait for the wind"
In Lin Xiuchun's business perception, privacy protection can be divided into three major stages.
In the first stage, he calls it the traditional Internet era. People would hand over data when posting or searching. The rampant "cyber sleuthing" inferred others' life trajectories through marginal information and finally profited through means such as fraud and threats - at least not directly related to economic interests.
In the second stage, he calls it the (pre -) large - model era. Facing the AI dialog box, people start to upload hundreds or thousands of times more sensitive files, meeting minutes, and work content to the public network for AI processing. So everyone has to face a real problem: Do we choose security or efficiency?
In the third stage, he calls it the (post -) large - model era. The popularity of self - thinking agents like OpenClaw has completely detonated people's anxiety about AI privacy. In the past, even if you were worried, at least you knew what files you had uploaded and what things you should never hand over. But now, a local "Lobster" can secretly take a lot of your sensitive data to the public network for inference without your knowledge, which is very terrifying.
He said that in this era, once a data breach occurs, the first victim is completely unaware. Secondly, sensitive privacy is embedded in the database of the large model and will be remembered by the Internet forever. Thirdly, targeted cyber sleuthing based on the data breach of large - model Q&A makes the leaked data come not only from social media but also from all your work and life content stored on your computer.
"Imagine that someone you dislike can pay to buy all the data of your interactions with AI," Lin Xiuchun said. "All it takes is one phenomenal data security breach event."
This is the problem to be solved in the post - Lobster era: When 'Lobster' has the permission to read all your data, and when AI is no longer a tool you actively feed but a 'digital employee' lurking in your system 24/7, what should you do? At that time, confidential computing is not an option but a must - answer question.
Lin Xiuchun believes that almost no one needs "24 - hour privacy", but everyone needs that "critical 10 minutes".
"When you usually ask the large model 'Which coffee is good?' or 'Who is more powerful, Li Bai or Du Fu?', these questions are not private at all. But anyone may, at some point in their lives, due to work or life reasons, need 10 to 20 minutes of absolute privacy - using the large model to handle some truly sensitive issues," he explained. For example, a government official needs to draft a highly confidential internal reference; a corporate founder needs to discuss an undisclosed merger and acquisition plan; an ordinary person wants to consult a medical condition that is hard to talk about. In these scenarios, people need a powerful and private model to help them solve problems.
He said that the biggest difficulty now is that most people don't have a real sense of how insecure the public network is. People are used to "going naked" and think that privacy breach is someone else's problem or have a fluke mentality - anyway, it hasn't happened to me yet.
His strategy is "get on the table and wait for the wind". First, penetrate into some high - privacy scenarios such as finance, medical, and government affairs, and let a small number of people with rigid needs start using it.
"Look, before the outbreak of the 'Panda Burning Incense' virus, many network security companies in China were on the verge of collapse. Then, overnight, they all took off," Lin Xiuchun said. The situation with AI is even more exaggerated - almost all sensitive data is being uploaded to the public network. Even a small data breach event will cause widespread panic. "As long as we are at the table at that time, we will be victorious."
Post - 2000s CEO: Entrepreneurship is a partnership between "optimists" and "cautious people"
Lin Xiuchun has many labels: post - 2000s, Peking University doctor, and suspending studies to start a business. In fact, Lin Xiuchun, who has always been passionate about computer science, made a far - sighted choice when choosing his major - he studied business and economics and obtained a doctorate.
Lin Xiuchun won the National Scholarship during his undergraduate studies and published top - tier journals as the first author. When applying for postgraduate studies, he faced two choices: one was to continue to delve into the field of computer science, and the other was to pursue a master's degree in finance at Peking University Guanghua School of Management. Finally, the advice of an elder made him determined: "If you really want to achieve something, it's best to understand both technology and business."
So Lin Xiuchun went to Peking University Guanghua to study business.
Caption: Lin Xiuchun
Regarding the matter of "post - 2000s entrepreneurship", Lin Xiuchun has a clear understanding and poured cold water on it: "I think it still depends on the individual and whether what you do matches your endowments. The advantage of post - 2000s is their perception of new consumption and cutting - edge technologies, but the disadvantages are also obvious - many people don't have years of work experience and are prone to making mistakes."
Lin Xiuchun's solution is "partnership entrepreneurship". He said that if the problems of decision - making and division of labor are solved, it is entirely possible to have both. For example, a senior CEO is best paired with one or two young and innovative partners who are responsible for technological breakthroughs at the forefront or the operation and sales of new tracks. A young and technology - savvy CEO should be paired with some middle - aged people with experience, business knowledge, and both success and failure experiences in the team, and give the elders full right to give advice in major decision - making.
Lin Xiuchun emphasized that it is the right to give advice, not the right to make decisions. "Many young CEOs may be suppressed by their senior partners. But from a business logic perspective, whoever is in charge of the company should make decisions and take responsibility."
There is such a partner in the team - a senior executive vice - president who has worked in Accenture, Alibaba, and Huawei for many years. After Lin Xiuchun's multiple invitations, this successful executive decided to quit his high - paying job at Huawei and join this startup company. "A partnership based on age, taking advantage of each other's strengths, is the best solution."
If you want to know the difference between a post - 2000s CEO and an older CEO, perhaps the following "small thing" can explain it a bit:
Two post - 2000s interns in the company designed a bear wearing sunglasses for the company's brand image. When combined with "Lobster", it looks a bit absurd and abstract. If it were a 40 - year - old CEO, he would definitely frown and ask what this is. He wouldn't understand it and wouldn't like it. But Lin Xiuchun supported them. "Let post - 2000s design things for post - 2000s, and let young people perceive the future they want," he said.
This article is from the WeChat official account "China Venture Capital News", author: Riemann. It is published by 36Kr with authorization.