Behind Zhipu AI's "Over 2.5 Billion Yuan Financing"

[baoshi.rao]

Recently, the most talked-about topic in the "large model" circle is the financing announcements of two major AI startups—Baichuan Intelligence securing $300 million in funding, elevating its valuation to unicorn status, while Zhipu AI raised over 2.5 billion yuan, drawing attention with its 10-billion-yuan valuation.

The collaboration between Baichuan and Xiaomi has garnered widespread industry attention, with Lei Jun's involvement further spotlighting Zhipu AI. Beyond these two highly discussed companies, another large model startup, "Moonshot AI," has also become a favorite among investors. With backing from Sequoia Capital and ZhenFund, Moonshot AI has already raised over $200 million.

As the year draws to a close, the capital competition in the "large model" sector is beginning to show results, with Zhipu AI stepping into the limelight with the highest financing amount and valuation. Due to its long-standing low-profile academic style, Zhipu AI had not garnered much attention outside the industry until this high-profile financing announcement, which ignited the venture capital community.

Faced with such large-scale financing, we can't help but ask: What are Zhipu AI's advantages, and why have so many investors and institutions chosen it? This also prompts us to reflect on the future development of the AI industry. With the rise of next-generation technologies like cognitive intelligence, what new opportunities and challenges will the AI industry face? What will the future landscape look like?

1. Zhipu AI: The "Chinese OpenAI" Backed by Capital Markets

Not only that, but the performance of their products is strikingly similar. For example, in Stanford's evaluation of mainstream global large models, Zhipu AI's GLM-130B, developed in 2022, was the only Asian model selected, matching OpenAI's GPT-3 in accuracy and maliciousness while outperforming all other models in robustness and calibration error.

Moreover, as AI technology becomes increasingly widespread, ensuring its ethical and moral implications has become a global focus. Both OpenAI and Zhipu AI have begun emphasizing AI ethics research and practice, striving to develop technologies that adhere to ethical standards.

Financially, the two companies are also alike. Zhipu AI leads the domestic AI sector in funding, while OpenAI accounted for over 70% of total AI-related financing in the U.S. last year, making it the most favored by investors. The capital market is notoriously ruthless, and investment decisions are based on a comprehensive evaluation of technological trends, market potential, and the company's overall capabilities. When multiple investors pour money into a single company, it signals strong future prospects.

2. The "Industrialization of Large Models" Trend Revealed by Investor Backgrounds

From a societal perspective, the industrialization of large models has profound implications. On one hand, applications like smart homes and intelligent transportation improve quality of life, making it more convenient and secure. On the other hand, large models also introduce new challenges, such as data privacy and AI ethics, which require collective societal efforts to address.

From a commercial standpoint, large models are expanding beyond the internet into traditional sectors like finance, healthcare, education, manufacturing, and services. These industries possess vast data resources but have struggled with data processing and intelligent applications. Large models, with their high precision, efficiency, and reliability, offer new solutions to enhance productivity, reduce costs, better understand customer needs, predict market trends, and drive innovation. This accelerates the digital transformation of traditional industries, making commercialization a key driver of large model industrialization.

Of course, industrializing large models requires support from the entire ecosystem, including hardware, software, data, and talent. As the ecosystem matures, industrialization will gain more comprehensive support, and stakeholders will reap greater benefits.

3. The Next Frontier in AI: "Cognitive Intelligence"

Computational intelligence is the foundation of AI, enabling machines to perform calculations and store data—tasks beyond human capacity, such as memorizing thousands of numbers. Perceptual intelligence goes further, allowing machines to understand human language, recognize images, and make decisions based on sensor data, as seen in facial recognition systems.

Cognitive intelligence, the advanced stage of AI, is the key bottleneck for breakthroughs and large-scale industrial applications. It requires machines to think, reason, learn, and adapt like humans, with applications in smart customer service, smart homes, and autonomous driving. Unlike perceptual intelligence, cognitive intelligence emphasizes depth and simulation of human cognition, demanding significant R&D investment. During the National People's Congress, representative Liu Qingfeng urged, "China must accelerate the development of cognitive intelligence large models and promote their application across industries—this is an urgent priority."

In business, cognitive intelligence is expanding rapidly. Leveraging data, machine learning, and natural language processing, it helps companies simulate human decision-making, providing accurate and reliable insights. As data grows and technology advances, data-driven decision-making will become even more prominent.

Meanwhile, evolving consumer demands are pushing businesses toward personalized services. Cognitive intelligence can deeply analyze consumer behavior to deliver tailored products and experiences, a trend already evident in e-commerce and finance, driving digital transformation across sectors.

However, cognitive intelligence faces challenges. Data privacy is a major concern, requiring robust measures to ensure security. Algorithmic bias and risks also pose issues, as human-designed models may inadvertently introduce errors, impacting decisions and user experiences. Strengthening design and validation processes is crucial to mitigate these risks.

Additionally, improving the interpretability and flexibility of cognitive intelligence remains a challenge. While progress has been made, complex tasks still require human-like adaptability and reasoning, limiting applications in critical fields like healthcare and finance. Collaboration between human experts and AI is essential to enhance solution accuracy and effectiveness.

Facing the current state of cognitive intelligence R&D, large manufacturers like Baidu and Alibaba actually have more advantages in technological breakthroughs—they typically possess vast data resources, where the richness and quality of data often play a crucial role in model accuracy and performance. At the same time, they can invest more manpower and resources to research and develop cognitive intelligence technologies, attract top talent from related fields, and leverage their brand trust and market share to more easily attract partners and distributors for product application testing and industry chain collaboration. This enables deeper research, faster iteration, and a higher likelihood of achieving technological breakthroughs.

However, no matter what, we must not forget one thing: whether it's large manufacturers or startups, in this exploratory phase of cognitive intelligence, everyone is in the position of "the little horse crossing the river." To secure a place in this fiercely competitive market environment, continuous technological innovation and enhancing one's own capabilities will always be the primary factor in reaching the other shore of success.