(TMTPOST)—Zhipu AI, a unicorn company valued at 20 billion yuan (US$ 2.77 billion), has recently participated in the first round of financing for the AI infrastructure software company Beijing Qingcheng Jizhi Technology.
According to Tianyancha, Qingcheng Jizhi recently completed a new round of registration changes, adding entities such as China Gold Guanbo Shanghai Private Equity Investment Fund Enterprise (CICC Capital) and Beijing Zhipu Huazhang Technology Co., Ltd. (Zhipu AI) as its shareholders, with Zhipu AI holding a 2.1673% stake.
Prior to this, Zhipu AI had also invested in distributed storage company Yanrong Technology, Qingcheng Jizhi's competitor Infini AI, multimodal model company Shengshu AI, and many other companies in the AI industry chain.
As of now, Zhipu AI has invested in over ten startups, with a single investment scale reaching the million-yuan level.
Established in December 2023, Qingcheng Jizhi is dedicated to building efficient AI system software, empowering domestic computing power, and supporting the development of large-scale models. The company's co-founder, chairman, and CEO Tang Xiongchao, along with the core technical staff, are all doctoral graduates from the Department of Computer Science at Tsinghua University, with the current research and development team comprising approximately 20 people.
By the end of 2024, Qingcheng Jizhi is expected to expand to a team of 50 people, with research and development personnel accounting for over 80%.
According to official disclosures, Qingcheng Jizhi has a full-stack research and development team that fully masters technologies related to bottom-layer compilers, upper-layer parallel acceleration, and other aspects.
The company has multiple core capabilities in supporting domestic intelligent computing power in system software, covering key technological areas such as parallel systems, computing frameworks, communication libraries, operator libraries, AI compilers, programming languages, scheduling systems, storage systems, memory management, and fault tolerance. It can achieve efficient end-to-end model training and inference services and provide ultimate performance optimization capabilities.
In response to the problem of multiple types of domestic AI chips with low efficiency, the company has independently developed an intelligent compiler that can generate efficient code for various chips to fully utilize various domestic intelligent computing chips, achieve code portability and performance portability, eliminate user migration costs, and reduce the threshold for the landing deployment of artificial intelligence in various fields.
The company’s technology belongs to the AI infra system. AI Infra refers to the infrastructure needed to build AI, covering the AI middleware infrastructure that connects computing power and applications, including data preparation, model training, model deployment, and application integration. It not only involves computing power but also includes optimization at the big data level.
AI Infra includes AI basic framework technology, involving various underlying facilities in the fields of large-scale model training, deployment, and basic software stack, aiming to optimize computing power algorithms and promote application landing.
It plays a crucial role in connecting hardware, software, and applications in the AI industry, promoting the development and application of AI technology. At present, there are only a few companies in China that can do AI Infra technology, including Infini AI and Qingcheng Jizhi - and behind both companies is Zhipu AI.
In 2023, Zhipu AI jointly initiated the CCF-Zhipu Large Model Fund and SMP-Zhipu Large Model Interdisciplinary Fund with the China Computer Federation (CCF) and the Social Media Processing Special Committee of the China Society of Information. These funds were used to support research and interdisciplinary innovation in large models, covering large model theory, algorithms, models, applications, and other related research. Last year, the two funds provided research support for 41 research projects involving more than 30 universities nationwide, with a cumulative total of over ten million yuan in cash and computing resources.
In January 2024, Zhipu AI announced the establishment of an open-source large model fund. At the same time, Zhipu also announced the establishment of one billion yuan entrepreneurial fund, the "Z Plan" Entrepreneurship Fund, in collaboration with its ecosystem partners, to support original innovation in large models, covering large model algorithms, underlying operators, chip optimization, industry large models, and super applications, among other directions.
While engaging in large-scale AI models, Zhipu AI is also investing in numerous AI industry chain enterprises, showcasing a unique strategy amidst the current large model enthusiasm. Various signs indicate that Zhipu AI is trying to emulate Sam Altman and his OpenAI.
Zhipu AI CEO Zhang Peng said that Zhipu is one of the few leading technology companies in China that can compete with OpenAI. The company hopes to work together with partners across the industry chain, developer communities, and academia in the AI field to contribute to the future of Chinese AI.