Moore Threads CEO Zhang Jianzhong in Dialogue at Peking University Guanghua: Analyzing the Breakthrough and Long March of Domestic GPUs

On the evening of October 27, the first event in the "Guanghua TechTalk" dialogue series was held in the lecture hall of Building 2 at Peking University's Guanghua School of Management. The event invited Zhang Jianzhong, founder, chairman, and chief executive officer of Moore Threads, to come to Guanghua and engage in an in-depth dialogue with Tian Xuan, dean of Peking University's Guanghua School of Management and Peking University Boya Distinguished Professor, to discuss together the computing foundation of the AI era, the breakthrough path for domestic chips, and the long-termism of hard-tech innovation. More than 300 faculty, students, and alumni attended in person.

Keynote sharing:

Doing difficult but correct things

At the start of the event, Zhang Jianzhong gave a keynote sharing. Starting from "GPU is more than graphics processing," he explained the uniqueness of the "Universal GPU": from 3D graphics rendering to AI training and inference, physical simulation, and scientific computing, almost all computations that can be expressed mathematically can be completed by a universal GPU. In Zhang Jianzhong's view, AI is like a new industrial revolution and also a version iteration of humanity's "operating system." With AI, everyone's mental and physical capabilities can be greatly extended, and the ability to explore the unknown is thereby amplified tenfold or even more.

Zhang Jianzhong emphasized that AI is inherently everyone's assistant, not an opponent competing with humanity for resources. Rather than being anxious about "whether AI will take my job," it is better to learn to master AI and coexist with it, treating it as a good friend, a good teacher, and an excellent colleague. He proposed that AI will ultimately evolve from a technology into infrastructure as ubiquitous as water and electricity.

At the same time, he said that competition in the AI era is essentially competition over "soil," and computing power is the fertilizer that nourishes industrial innovation—whoever has the most fertile computing foundation can cultivate the best innovation. China has elevated AI to a national strategy, forming a highly unified support atmosphere across policy, capital, and society. This is a precious innovation soil and also the deep reason why Moore Threads chose to officially begin operations on October 1, 2020.

He particularly emphasized the power of "belief": only by believing in AI and believing in the future can consensus be built and investment be sustained. This is precisely the original aspiration with which Moore Threads, since its founding, has upheld the vision of "accelerating a better world" and has been determined to build universal GPUs.

Dean dialogue:

How to persist in difficult but correct things

In the subsequent "Dean Dialogue" session, Tian Xuan and Zhang Jianzhong engaged in a one-hour in-depth conversation, with topics extending from career choices and industrial breakthroughs to organizational management and the capital market.

Speaking about why he started a business, Zhang Jianzhong, who worked for fifteen years at a leading overseas company and once led a team to build Greater China into the world's most important GPU market, deeply realized that key technologies must be mastered in one's own hands.

In October 2023, Moore Threads, only three years after its founding, encountered external restrictions. Recalling the situation at the time, Zhang Jianzhong said that being "choked at the neck" instead forced the company to rise to the challenge, restructure upstream and downstream cooperation domestically, and ultimately achieve independent controllability in key links from chip design to manufacturing. This is exactly the meaning of "difficult but correct."

In terms of organizational management, Moore Threads has been established for six years and has more than a thousand employees, nearly 80 percent of whom are R&D personnel. Zhang Jianzhong said, first, shared belief: "Success comes from passion; believe, and you will see"; second, alignment mechanisms: the purpose of strategic alignment is not to eliminate differences but to make execution consistent; third, information flow should not be filtered by hierarchy, so that frontline information can reach the decision-making level directly.

Regarding the capital market, Zhang Jianzhong said, "Going public is a report card, not the finish line. With clear goals, do what should be done." In his view, enterprises should answer the real questions: Has the technology progressed? Are the products competitive? Has the organization filled the shortcomings that should be filled?

Tian Xuan shared his observations from the perspective of research on venture capital and the capital market. He said that research on venture capital often finds that "the more knowledgeable people are, the less they dare to act," because insiders are more aware of technical difficulty and industrial gaps; Zhang Jianzhong started a business after fifteen years at a leading overseas company and is one of the few samples who transformed industrial experience into entrepreneurial judgment.

Combining his own research, Tian Xuan emphasized that tolerance for failure is highly correlated with innovation incentives. Only when trial and error have an institutional outlet will enterprises dare to pursue disruptive exploration; if one failure means a veto, organizations will tend toward conservatism. The capital market, in serving science and technology innovation enterprises, likewise needs to understand the long-cycle characteristics of technological innovation and find a balance among information disclosure, investor communication, and institutional tolerance. Letting society as a whole share innovation risks is not lowering standards, but respecting the objective laws of hard-tech investment.

Interactive Q&A:

In the interactive session, in response to a question from a student with a business background about "how to position oneself in the AI era," Zhang Jianzhong said that the objects of management in the AI era are expanding from "people" to "agents," and this is precisely where business schools can play a role—learning management, collaboration, and resource allocation will become scarce capabilities. Tian Xuan gave his answer: when AI can quickly provide answers, what is truly irreplaceable is the ability to ask questions, the ability to judge, and the ability to empathize, communicate, and capture "soft information." This is precisely the unique value that business schools should forge for students.

On the topic of Chinese AI going global, an alumnus engaged in game exports asked a question on site. Zhang Jianzhong judged that the going-global of computing power and tokens is an inevitable trend. With lower production costs, a more complete industrial chain, and advantages in new energy, Chinese AI going global has natural competitiveness, and countries in the Global South will become a broad incremental market; content creation is precisely the field where AI excels most, and imagination will be scarcer than computing power.

On the evening of the event, Moore Threads also provided students on site with recruitment special benefits such as a "green channel for interviews" for submitting resumes.

After the conversation ended, Tian Xuan, on behalf of the Guanghua School of Management, presented a commemorative gift to Zhang Jianzhong. The two parties concluded the dialogue with a group photo of the entire audience, while reflections on "doing what is difficult but right" continue in the hearts of every young person devoted to hard-tech innovation.

Key Takeaways | Highlights of Moore Threads CEO Zhang Jianzhong's Views

01 | GPU Is More Than Graphics Processing

The full name of GPU is Graphics Processing Unit. But today its meaning is no longer limited to graphics processing: from 3D rendering and video encoding and decoding to AI training and inference, scientific computing, and physical simulation, a single chip carries computing power for multiple scenarios. Zhang Jianzhong said that a full-function foundation that does not limit imagination is the computing infrastructure that can withstand the test of time.

02 | We Are in a New Version of the Human Operating System

This wave of AI is a singularity moment in which technological breakthroughs, a knowledge explosion, an entrepreneurial atmosphere, and the liberation of productivity arrive simultaneously. Zhang Jianzhong analogized it to "version iteration of the human operating system"—mental power can be expanded with the help of cloud AI, and physical power can be extended toward physical AI. This is not a tool upgrade but species evolution: under the new system, the upper limit of human cognition is raised, and the core of the gap between people is no longer simply memory and computing ability, but the courage to dare to trial and error, the imagination facing the unknown, and the desire to keep moving forward.

03 | AI Will Become Infrastructure

Tool, friend, good teacher, colleague, and partner—the last one is environment. This is the order Zhang Jianzhong laid out for the stages of the relationship between AI and people. By the final stage, you no longer feel its existence—it becomes infrastructure like water and electricity; you do not think of it unless the power goes out. He said many people have long forgotten the existence of water and electricity, and that is exactly what infrastructure should be like. To judge whether a technology has truly landed, see whether it can naturally integrate into daily life and support our production and life without being felt.

04 | How Precious Is China's AI "Soil"?

"Rarely does a country elevate a technology to the height of a national strategy." Zhang Jianzhong said that China's unique advantages in developing AI first come from top-level consensus. From the national level to local governments, strategic attention to the AI industry has been formed from top to bottom; coherent and unified industrial guidance policies and consistently stable support directions are very rare globally. Beyond the policy level, this innovative soil also comes from the tolerance and respect of the whole society. Doing chips, building large models, and developing robots—the hard-tech track has long cycles and high failure risks. Only when the whole society is willing to give innovation room for trial and error and recognize the value of frontline workers in research and industry can innovation take root and sprout.

05 | Crisis Instead Forces Independent Reconstruction

Moore Threads encountered external restrictions three years after its founding, which forced the team to break away from the original path, seek domestic alternatives, and reconstruct the complete chain. The bottleneck predicament accelerated the coordinated integration of domestic upstream and downstream supporting links. Hard-tech self-reliance will not be smooth sailing, and external pressure objectively accelerates the maturation of local industry. The real barrier is not a single chip but a complete system composed of chips, software, tools, and ecosystem together, which is also the direction Moore Threads continues to tackle.

06 | Lower the Threshold So Ordinary People Can Use AI Computing Power

The computing power industry should not serve only big companies and research institutions; popularization of the ecosystem is key. To this end, Moore Threads has accelerated its layout of edge-side AI products, launching the AI computing laptop AIBOOK for the general public and the AI hub AICUBE for families, with the goal of achieving "inclusive" computing power: enabling elderly people, children, and ordinary people without technical backgrounds to access AI computing power at low cost and with a low threshold. The prosperity of the computing power industry is not only a contest of chip performance parameters but also the prosperity of the application ecosystem. Only by enabling ordinary people and developers in all industries to use computing power conveniently can rich local AI applications be spawned and a positive cycle formed, thereby revitalizing the entire domestic computing power ecosystem.

07 | The Essence of Management Is Communication; Everything Is Efficiency

In Zhang Jianzhong's view, the essence of management is communication, the transformation of information; the smoothness of information flow directly determines the efficiency of the organization. He said bluntly that the ultimate deliverable of management is efficiency, and everything comes down to efficiency in the end. In traditional hierarchical management, information must be summarized and filtered layer by layer and then transmitted downward, just as a neural network performs a screening at each layer; by the time it reaches the end, it is often distorted, which is precisely the biggest source of internal friction in an organization. He used the metric he is most familiar with as an analogy: how high the computing power utilization rate can reach, and what percentage of the operators' capability is brought into play in your hands. Leading a team is actually asking the same thing: how much information is lost in your hands. Moore Threads' approach is that administrative levels may be set, but information flow must be aligned; progress and ideas from frontline R&D can reach management directly, and through continuous "alignment" to ensure consistent pace, using the smoothest information flow to eliminate internal friction and go straight to results.

08 | Imagination Will Be Scarcer Than Computing Power

This was Zhang Jianzhong's closing for the whole audience. When AI drives execution costs very low, the bottleneck shifts from resources to ideas, and imagination will be scarcer than computing power. At the same time, he extended from an organizational perspective: what the AI era lacks most is not executors who write code but managers who can manage agents. In the future, an AI company may have only a small number of core positions coordinating hundreds of agents, and what is needed to steer this system is precisely the ability to divide labor, collaborate, and circulate information. In the AI era, the value of people is not dissolved; instead, it continues to rise toward creativity and management.