The artificial intelligence (AI) landscape is in constant flux, with companies continuously adapting to new technologies, market demands, and competitive pressures. Once hailed as the future, several AI unicorns have navigated significant transformations, initial public offerings (IPOs), and funding rounds, each charting a unique path forward. This article delves into the current state of 11 such companies, examining their strategic shifts, financial performances, and future prospects in a rapidly evolving industry.
The resurgence of AI large language model (LLM) enterprises as industry focal points is indicative of this dynamic environment. Recent moves, such as Cambricon’s substantial investment in LLM computing power with a proposed fundraising of nearly 5 billion yuan, and Zhipu AI’s announcement of a 3 billion yuan financing round alongside IPO plans, highlight the growing interest and investment in this sector. Simultaneously, China’s LLM industry is experiencing vigorous growth, with over 320 registered generative AI large models as of March 2025, signaling the commencement of intense market competition.
Amidst this competition, six startups—Baichuan Intelligent, 01.AI, Zhipu AI, MiniMax, Moonshot AI, and StepFun—have emerged as the “Six Little Tigers” in the LLM field due to their advanced technological capabilities and considerable market potential, with valuations exceeding several billion dollars each. As these new forces gain momentum, attention is also directed towards the “AI Four Little Dragons”—SenseTime, Megvii, CloudWalk, and Yitu Technology—which once spearheaded China’s AI revolution. These companies, formerly prominent AI unicorns, now face a new wave of technological change and market rivalry. How are these former industry leaders faring today? It is foreseeable that the increasing maturity and widespread application of LLM technologies will lead to a profound reshuffling of China’s AI industry.
Cambricon: Scaling Up LLM Computing Power with Substantial Funding
Founded in 2016 and headquartered in Beijing, Cambricon is dedicated to providing high-performance, low-power chip solutions for AI applications. Established by brothers Chen Tianshi and Chen Yunji, the company’s products are extensively used in smart terminals, data centers, and intelligent vehicles, establishing it as a leading enterprise in China’s AI chip sector.
Cambricon recently unveiled its latest private placement plan, seeking to raise no more than 4.98 billion yuan. The proceeds are earmarked for chip and software platform projects targeting LLMs, as well as for supplementing working capital.
The chip platform project for LLMs is slated to receive an investment of 2.9 billion yuan, aimed at developing a series of chip products tailored for various LLM task scenarios. This includes chips for LLM training, reasoning, multimodal reasoning, and switch chips for LLM requirements. The software platform project, with a planned investment of 1.6 billion yuan, will construct a software platform for LLMs, enhancing the openness and usability of the company’s software ecosystem. This will better adapt to mainstream frameworks such as PyTorch and TensorFlow, reduce developer migration costs, and strengthen ecosystem stickiness.
Cambricon’s financial report for 2024 reveals revenue of 1.174 billion yuan, a 65.56% year-on-year increase. The net loss attributable to the parent company was 452 million yuan, a 46.69% reduction from the previous year. Looking at the first-quarter performance of 2025, the company achieved revenue of 1.111 billion yuan, a 4230.22% year-on-year increase, and a net profit attributable to the parent company of 355 million yuan, a 256.82% year-on-year increase. This marks Cambricon’s second consecutive profitable quarter.
Notably, Cambricon’s cloud product line showed remarkable performance in the 2024 financial report, generating revenue of 1.166 billion yuan, an 1187.78% year-on-year increase. Cambricon attributed the demand growth in its cloud product line to the computing power needs driven by the domestic implementation of LLM applications, particularly the accelerated adoption of AI applications led by DeepSeek.
SenseTime: Shifting Focus to Generative AI and Restructuring Business
SenseTime recently disclosed its 2024 financial report, reporting revenue of 3.77 billion yuan, a 10.8% year-on-year increase. The gross profit margin was 42.9%, with losses narrowing by 33.7% year-on-year. The company’s internal business structure underwent significant adjustments, with generative AI becoming the core driver of revenue growth, reaching over 2.4 billion yuan for the year, a 103.1% year-on-year increase, and increasing its share of total revenue from 34.8% in 2023 to 63.7%.
Conversely, SenseTime’s traditional business segments faced challenges. Visual AI revenue was 1.112 billion yuan, a 39.5% year-on-year decrease, with its revenue share dropping from 53.9% in 2023 to 29.5%. Smart car revenue was 256 million yuan, a 33.2% year-on-year decrease, with its revenue share falling from 11.3% to 6.8%. These changes indicate SenseTime’s rapid transition from traditional AI to generative AI.
SenseTime has consistently invested in technological innovation and infrastructure construction, providing solid support for business growth. In 2024, SenseTime’s large-scale computing power increased by 92%, with a total operational computing power scale of 23,000 PetaFlops, and a preliminary nationwide layout was completed. This robust computing power base provides strong support for the training and optimization of generative AI models, further enhancing SenseTime’s technological barriers in the AI field.
In 2024, SenseTime completed two equity placements, providing ample financial support for the company’s business development. In June, SenseTime completed an approximately 2 billion Hong Kong dollar equity fundraising, attracting strategic investors and leading funds. In December, SenseTime again raised approximately 2.8 billion Hong Kong dollars through the placement of new Class B shares. These funds will primarily support the company’s core business development, including building AI clouds, expanding the scale of large-scale computing power, supporting generative AI (such as large model research and product development), and general operating funds.
To better respond to the opportunities and challenges brought by the AI 2.0 era, SenseTime completed a strategic organizational restructuring on December 4 of last year, forming a “1+X” new structure. “1” represents SenseTime Group’s core business, focusing on building an industry-leading AI cloud to achieve seamless integration of large-scale computing power, basic models, and AI applications. “X” represents SenseTime Group’s restructured ecosystem matrix, including intelligent vehicle “Jueying,” family robot “Yuanrobot,” smart healthcare, and smart retail. Each ecosystem company will establish an independent CEO, responsible for business development, while also obtaining more flexible incentive mechanisms and financing channels. Currently, these independent segments are advancing financing matters.
The rapid development of generative AI has become SenseTime’s core narrative. In 2023, SenseTime began to shift its business focus to generative AI, reorganizing its business into three segments: generative AI, traditional AI, and intelligent vehicles. At the end of 2024, SenseTime further reorganized its business into generative AI, intelligent vehicles, and visual AI. This series of adjustments indicates that SenseTime is fully betting on generative AI to seize the vast market opportunities brought by this emerging technology.
Megvii: Deepening AIoT Ecosystem and Venturing into Smart Cars
Megvii, a leading AI enterprise in China, has recently made frequent moves in strategic financing, business layout, and organizational structure adjustments, demonstrating its ambition and strategic layout in the AIoT (Artificial Intelligence of Things) field.
In April 2025, Megvii completed a strategic financing round of several hundred million yuan. Investors included Shanghai Yunzhao Venture Capital Co., Ltd., under Ant Group. These funds will primarily be used to build an AIoT ecosystem, strengthen the self-developed visual perception algorithm engine and AIoT operating system, and promote the deep integration of AI and IoT in consumer, city, and supply chain scenarios.
Megvii’s business layout focuses on three core scenarios: consumer IoT, city IoT, and supply chain IoT. Through its self-developed Brain++ platform, it has achieved the full-process and scaled supply of AI technology. The platform integrates the deep learning framework MegEngine, the cloud computing platform MegCompute, and the data management platform MegData, supporting the comprehensive integration of AI technology from algorithm production to application.
Megvii also recently completed an industrial and commercial change, increasing its registered capital from approximately 5.715 million yuan to approximately 33.942 million yuan. This change is not a new financing but to dismantle the VIE structure built for a Hong Kong IPO. This adjustment indicates that Megvii’s layout in the capital market is changing, and there may be more strategic actions in the future.
Notably, Megvii’s co-founder, chairman, and CEO Yin Qi joined Chongqing Automobile Company Lifan Technology (now renamed Qianli Technology) with part of his team after terminating Megvii’s five-year listing process at the end of November 2024, becoming the company’s chairman. This move shows Megvii’s strategic intentions in the smart car field, with Yin Qi mainly promoting the research and development of new technologies, including intelligent driving.
CloudWalk Technology: Seeking Transformation and Betting on Industry LLMs with Huawei Ascend
CloudWalk Technology, a leading AI enterprise in China, recently released its first-quarter report for 2025, showing that the company is still in a critical period of transformation, facing the dual challenges of declining performance and transformation breakthrough.
In the first quarter of 2025, CloudWalk Technology achieved total operating revenue of 37.2332 million yuan, a 31.56% year-on-year decrease; net profit was -124 million yuan, with losses decreasing by 22.86% year-on-year. Although losses have narrowed, the overall revenue decline is significant, reflecting the company’s considerable pressure in business adjustment and market expansion.
In the full year of 2024, CloudWalk Technology achieved operating revenue of approximately 398 million yuan, a 36.69% year-on-year decrease; the net profit attributable to the parent company was -696 million yuan, with losses expanding by 8.12% year-on-year. The decline in company revenue is mainly due to the proactive promotion of strategic adjustments in product portfolio and customer structure, reducing low-value-added business units, focusing on high-quality customer groups, and promoting structural optimization of business quality. However, this adjustment inevitably led to a contraction in the scale of new orders.
Faced with performance pressure, CloudWalk Technology has embarked on a “cut-off for survival” transformation path. In 2024, CloudWalk Technology deepened its cooperation with Huawei Ascend, jointly launching the “Calm LLM Training and Reasoning Integrated Machine,” and developing industry LLM applications based on the Ascend hardware base. In addition, CloudWalk Technology plans to modularize technology in vertical fields such as smart cities and finance, launch subscription services, and develop AI assistants and creative tools for individual users.
Despite facing many challenges, CloudWalk Technology continues to conduct technology research and development and product releases. In April 2025, CloudWalk Technology released the DeepSeek Prover-V2 model, demonstrating its technological accumulation and innovation capabilities in the AI field.
CloudWalk Technology’s transformation path is full of challenges but also contains opportunities. Through cooperation with Huawei Ascend and continuous investment in the field of LLMs, CloudWalk Technology is expected to find new growth points in the track of AI and IoT integration. However, the company needs to further optimize its cost structure and enhance its profitability while improving business quality to win the confidence of the capital market.
Yitu Technology: Seeking Breakthroughs in Medical AI and Innovating with Self-Developed Chips
Founded in 2012 and headquartered in Shanghai, Yitu Technology is a leading AI company in China. The company focuses on core AI technologies such as computer vision, speech recognition, and natural language processing, and is committed to promoting the implementation of AI technology in security, finance, healthcare, and other fields.
In 2022, Yitu Technology independently spun off its smart healthcare business unit to focus on developing the “digital doctor” system. The newly launched AI-assisted diagnostic platform and its chip team took a different approach, launching the world’s first “cloud-edge-end” full-scenario AI chip. In a smart manufacturing park in Jiangsu, Yitu’s industrial quality inspection system achieves micron-level defect detection, with a false detection rate controlled below 0.02%.
Yitu Technology has achieved significant results in several AI technology fields, especially in computer vision, speech recognition, and natural language processing. Its face recognition technology has repeatedly won the global first place in international authoritative evaluations. In recent years, Yitu Technology has launched several innovative technologies and products, including the Yitu Tianwen Large Model, Yitu Wanxiang, and the world’s first “cloud-edge-end” full-scenario AI chip.
Financially, according to Xinhua Finance, from 2017 to 2019, Yitu Technology achieved operating revenues of 69 million yuan, 304 million yuan, and 717 million yuan, respectively, with revenue growth exceeding 9 times in three years. However, the scale of losses is also expanding. From 2017 to 2019, Yitu Technology’s net profit attributable to the parent company was -1.166 billion yuan, -1.161 billion yuan, and -3.642 billion yuan, respectively.
Baichuan Intelligent: All In on AI Healthcare, Focusing on Vertical Fields
Baichuan Intelligent is a high-tech enterprise focused on AI technology research and development and application, committed to promoting innovation and breakthroughs in multiple fields through large language model (LLM) technology.
In the second half of 2024, Baichuan Intelligent clearly defined its strategic direction of “All in AI Healthcare,” focusing on the medical track and stopping the pre-training of new ultra-large-scale general LLMs. This strategic adjustment aims to concentrate resources to promote the deep application of AI technology in the medical field.
In July of last year, Baichuan Intelligent completed a Series A financing round, with a total financing amount of 5 billion yuan and a valuation rising to 20 billion yuan. The financing attracted many well-known investors, including leading manufacturers such as Alibaba, Tencent, and Xiaomi, as well as the Beijing AI Industry Investment Fund and the Shanghai AI Industry Investment Fund. After completing the Series A financing, Baichuan Intelligent plans to launch a Series B financing with a valuation of 20 billion yuan to further expand its influence in the AI field.
Notably, in March 2025, Baichuan Intelligent’s two co-founders, Chen Weipeng and Hong Tao, successively resigned. Chen Weipeng is preparing for an AI coding-oriented startup project, while Hong Tao’s resignation is related to Baichuan Intelligent’s strategic adjustment to focus on core medical businesses.
01.AI: Strategic Adjustment from General LLMs to Industry Applications
01.AI was founded in May 2023 by Kai-Fu Lee, chairman of Innovation Works and former global vice president of Microsoft, focusing on the research and development and application of large language models (LLM) and related AI technologies. The company’s core team members come from top technology companies such as Google, Microsoft, IBM, and Baidu, with strong technical research and development capabilities and rich industry experience.
In August 2024, 01.AI completed a financing round of several hundred million US dollars, with participants including an international strategic investor, a Southeast Asian consortium, and Alibaba. Previously, 01.AI had also received several hundred million US dollars in angel round financing from Alibaba Cloud, and the company’s valuation has exceeded 1 billion US dollars, becoming a unicorn enterprise in the AI field. This round of financing will be used to further promote technology research and development and market expansion, especially in the layout of enterprise-level large models and AI applications.
In 2025, 01.AI announced that it would shift its development focus from pursuing ultra-large models to developing small-parameter, moderate-sized industry models, paying more attention to commercialization and industry applications. CEO Kai-Fu Lee clearly stated that the company no longer regards AGI (Artificial General Intelligence) as a short-term goal but focuses on the pre-training and deployment of lightweight models. This strategic adjustment makes 01.AI the first “AI Six Little Tigers” member to fully embrace DeepSeek technology.
In January 2025, 01.AI and Alibaba Cloud reached a strategic cooperation to establish the “Industrial Large Model Joint Laboratory,” aiming to accelerate the application of large models in industrial scenarios. The two parties will conduct in-depth cooperation in technology, computing power, business, and talent to promote the implementation of AI technology in finance, manufacturing, transportation, energy, and other fields.
On March 17, 2025, 01.AI officially released the “Wanji Enterprise Large Model One-Stop Platform” and provided enterprise-level DeepSeek deployment customization solutions. The platform supports localized reasoning and data isolation, suitable for industries with extremely high requirements for data security such as finance and government affairs.
In the domestic market, 01.AI will focus on To B (business-oriented) business, launching the closed-source model Yi-Large, which has served many global Fortune 500 companies. In 2024, the company’s overall revenue exceeded 100 million yuan, with B-end business accounting for as high as 70%, and C-end business also approaching profitability.
In the overseas market, 01.AI continues to explore To C business. Its AI office tool PopAi has been online for 9 months, with nearly 10 million users and a return on investment approaching 1. In addition, the company plans to spin off an AI application company “Oasis,” focusing on application development in games and other directions, and seeking external financing.
Zhipu AI: Launching IPO After Multiple Rounds of Financing, Accelerating Commercialization
Zhipu AI is a high-tech enterprise focused on AI technology research and development and application, incubated by the Knowledge Engineering Laboratory of the Department of Computer Science at Tsinghua University. The company’s core team members are all from Tsinghua University, with CEO Zhang Peng, Chairman Liu Debing, and President Wang Shaolan all being Tsinghua alumni. Zhipu AI is committed to developing largelanguage models (LLM) and related AI technologies, promoting the application of AI in multiple fields.
Zhipu AI has completed multiple rounds of financing in the past year, with financing scale and valuation repeatedly reaching new highs. In December 2024, Zhipu AI completed a financing round of 3 billion yuan, with new investors including several strategic investors and state-owned capital institutions, and old shareholders such as Legend Capital continued to follow suit. In March 2025, Zhipu AI completed a strategic financing of over 1 billion yuan, with investors including Hangzhou Chengtou and Shangcheng Capital, among other Hangzhou municipal state-owned capital. In addition, Zhipu AI also received a 500 million yuan Series E+ financing in March 2025, invested by Zhuhai Huafa Group.
As of now, Zhipu AI has completed at least 15 rounds of financing, with a total financing amount exceeding 16 billion yuan and a post-investment valuation exceeding 20 billion yuan. In April 2025, Zhipu AI officially completed the IPO guidance filing with the Beijing Securities Regulatory Bureau, becoming the first enterprise in China’s “Large Model Six Little Tigers” to launch the listing process.
Zhipu AI has made significant progress in technology research and development and product releases. In 2024, Zhipu AI successively released multiple large model products, including GLM-4, GLM-4-Plus, CogVideoX, GLM-4-VideoCall, and GLM-4-Voice. These models perform excellently in multimodal generation, deep reasoning, video generation, and other fields, with some performance indicators approaching or reaching international leading levels.
In 2025, Zhipu AI continues to increase its open-source efforts, announcing the open-source CogAgent-9B, the base model of GLM-PC. In addition, Zhipu AI also launched the deep reasoning model GLM-Zero-Preview, which performs excellently in mathematical logic and complex problem reasoning.
Zhipu AI’s commercialization process is also accelerating. In 2024, Zhipu AI’s commercial revenue increased by over 100% compared to 2023, with the MaaS platform bigmodel.cn attracting 700,000 enterprise and developer users. The C-end product Zhipu Qingyan App has over 25 million users, with annualized revenue exceeding 10 million. In addition, Zhipu AI has also partnered with Samsung mobile phones, bringing Agent experiences to Samsung’s latest phones.
Zhipu AI’s strategic layout focuses on technological innovation and commercialization. The company will continue to promote the research and development of large models, especially in the fields of multimodal and deep reasoning. At the same time, Zhipu AI is also increasing its exploration of the To C market, planning to launch more applications for individual users.
MiniMax: Leading Technology, Multimodal LLMs Empowering B and C-End Applications
MiniMax is a high-tech company focused on general artificial intelligence (AGI) and large language model (LLM) research and development, founded in December 2021. The company was founded by Yan Junjie, former vice president of SenseTime, and its core team members have strong technical backgrounds, committed to promoting the widespread application of AI through independently developed general large models.
MiniMax completed a Series B financing of 600 million US dollars in March 2024, with a valuation reaching 2.5 billion US dollars, becoming a unicorn enterprise in the large model field. The financing was led by Alibaba, demonstrating its high recognition of its technical strength and market prospects. Previously, MiniMax had completed at least three rounds of financing, including multiple rounds of support from well-known investment institutions.
MiniMax has made significant progress in large model technology. The company took the lead in realizing the combination of Linear Attention mechanism and MoE (Mixture of Experts) technology, developing large models represented by the MiniMax-01 series. In January 2024, MiniMax fully released the large language model abab6, and in April of the same year, it launched the trillion-parameter MoE large language model abab6.5. These models perform excellently in various core capability tests, approaching global top large language models such as GPT-4.
MiniMax’s multimodal large model has comprehensive capabilities at the world’s first echelon, supporting multiple modalities such as text, voice, music, images, and video, providing strong technical support for the efficient implementation of AI Agent and many industry scenarios.
MiniMax has a wide layout in both the B-end and C-end markets. On the B-end, the MiniMax open platform provides enterprises and developers with safe, flexible, and reliable API services, helping to quickly build AI applications. On the C-end, the company has launched multiple native applications, such as the productivity tool “Seashell AI” and the immersive AI content community “Starry Field.”
Moonshot AI: Long Text Smart Assistant Kimi Shows Promise, Exploring Content Communities
Moonshot AI is a high-tech company focused on general artificial intelligence (AGI) and large model technology research and development, founded in March 2023. The company was founded by Yang Zhilin, a graduate of Tsinghua University, and its core team members have participated in the research and development of multiple large models such as Google Gemini, Google Bard, Pangu NLP, and Wu Dao, with a deep technical background.
Multiple industry sources show that Moonshot AI has completed more than five rounds of financing since 2024, with investors including Alibaba, Sequoia China, Xiaohongshu, and Gaorong Venture Capital.
Moonshot AI has achieved significant technical breakthroughs in the field of large models. In October 2023, the company launched its first smart assistant product supporting long text input, Kimi Chat (later renamed Kimi Smart Assistant). In 2025, Moonshot AI continued to increase its technological innovation efforts, releasing multiple new models, including Kimi-Latest/Kimi-Thinking-Preview/Kimi-Prover.
Moonshot AI has also made important progress in commercialization and strategic layout. In April 2025, the company announced a content partnership with Caixin Media to provide users with more professional and reliable financial content. In addition, Moonshot AI has partnered with mobile phone manufacturers such as Huawei to equip the Kimi Harmony version on Huawei’s Pura X phones.
The latest news shows that Moonshot AI is grayscale testing its first content community product. This move is considered an important attempt to enhance user stickiness and explore commercialization paths. By building a content community, Moonshot AI hopes to better understand user needs and lay the foundation for future commercial monetization.
StepFun: Deeply Cultivating Multimodal Large Models, Empowering Diverse Industry Applications
StepFun is a high-tech company focused on general artificial intelligence (AGI) and multimodal large model technology research and development, founded in April 2023 and headquartered in Shanghai. The company was founded by Dr. Jiang Daxin, former corporate vice president of Microsoft, and its core team members come from top technology companies such as Microsoft, ByteDance, and Google.
StepFun takes “positive feedback on innovation from high talent density” as its core competitiveness, committed to promoting multimodal technology integration and AGI technology breakthroughs, and actively exploring implementation applications in multiple industries such as smart terminals, smart cars, finance, and content creation.
StepFun has achieved significant technical breakthroughs in the field of multimodal large models, and has released a total of 15 multimodal large models, repeatedly ranking first in China in the multimodal field on domestic and foreign authoritative lists. The company’s self-developed Step series model matrix covers multiple directions such as language, image, video, and voice, with the characteristics of comprehensive layout, leading performance, and rapid iteration. The main models include Step-2/Step-1V/Step-Video-T2V/Step-R1-V-Mini and others.
StepFun actively promotes the application of multimodal large models in multiple industries, covering smart terminals, fintech, content creation, smart manufacturing, and other fields. In the smart terminal field, it has cooperated with Geely Automobile, OPPO, and others to launch车载智能系统and mobile phone multimodal search functions; in fintech, the company has launched the securities industry’s first trillion-parameter model “Junhong Lingxi” to achieve intelligent interpretation of financial data; and in content creation, it supports IP interactive game development and 4K film-level video generation.