Unitree, CXMT among Chinese new-gen firms charging forward amid collaborative ecosystem
In this year’s sweltering summer, China’s technology landscape arrived at an inflection point. In the span of less than a month, two domestic tech companies made global headlines with stock market debuts that would have been unthinkable a decade ago.
On Aug 10, Unitree Robotics — hailed as “China’s first humanoid robot stock” — opened subscriptions for its initial public offering on Shanghai’s STAR Market. The company priced its IPO at 150.80 yuan ($22), aiming to raise about 6.1 billion yuan in total proceeds — a staggering multiple for a company that, like many in the humanoid robotics space, is still scaling commercial deployments. The company surged as much as 629 percent in its stock market debut on Aug 19, giving it a market capitalization of nearly 445 billion yuan ($62 billion), more than seven times its valuation at the IPO price.
Three weeks earlier, on July 27, the Chinese mainland’s largest memory chipmaker ChangXin Memory Technologies (CXMT) had an even more headline-making debut on the same exchange. Shares of the tech giant surged 465.82 percent to close at 49 yuan per share, propelling its market value to more than 3 trillion yuan — making it the most valuable company on the A-share market on its first trading day.
But a look beyond the head-turning numbers reveals a far more consequential tale unfolding.
This is not about one firm’s heroics but a replicable, self-reinforcing innovation ecosystem emerging across China’s technology landscape. It spans semiconductor design, chip fabrication, artificial intelligence (AI) frameworks, and industrial integration. And it is beginning to challenge the long-standing assumption that tech supremacy flows only from a handful of breakthrough nodes controlled by a few foreign players, said experts and corporate executives.
Such a robust innovation ecosystem can be seen in everything from Huawei’s newly proposed Tau Scaling Law — which reimagines chip development processes for the global semiconductor industry — to the dizzying pace of five major Chinese AI models released in just eight weeks.
The flurry also includes a Shenzhen, Guangdong province-based supply chain that can source 30,000 robot components within half a day, and the seamless integration of traditional manufacturing with cutting-edge embodied intelligence.
“Foreign technology export controls are forcing China to solve a structural problem that Silicon Valley never had to,” said Xiang Ligang, director-general of the Information Consumption Alliance — a telecom industry association.
“China is no longer chasing a single ‘silver bullet’. Instead, it is building an orchestrated stack where every layer, from silicon to software, improves in lockstep — and that may prove more resilient,” Xiang said.
Take Unitree Robotics. Its listing partners offer a clue to the systemic shift. Its strategic investors include not only the National Council for Social Security Fund and DeepSeek, but also the investment arms of China National Petroleum Corp, China Southern Power Grid, China Telecom, Tencent, and CITIC Securities.
A humanoid robot is no longer a niche gadget — it is a node connecting energy infrastructure, telecom networks, financial services, and AI research. Behind a single robot, dozens of interconnected sectors come together: AI models, chips, sensors, precision reducers, motors, batteries, new materials, industrial software, and intelligent manufacturing equipment.
Meanwhile, CXMT’s debut delivered an even louder statement. Wu Hao, a portfolio manager at Founder Fubon Fund, described CXMT’s significance in clear terms: “It completes the most critical piece of a puzzle for the A-share market’s memory sector. It gives China its first homegrown DRAM powerhouse with global heft, sharpening the investment case across the memory value chain.”
What Wu did not say explicitly — but industry watchers have begun to recognize — is that CXMT’s rise is inseparable from the wider ecosystem. The memory value chain — equipment, materials, design, manufacturing, and packaging — is no longer a series of isolated Chinese firms fighting their own battles. They are being wired together.
In late May, as CXMT was preparing for its IPO, a different kind of breakthrough was presented at an international semiconductor conference. He Tingbo, president of Huawei’s semiconductor business department, unveiled the Tau Scaling Law — a theory that experts are quietly describing as the most significant rethinking of chip evolution since Gordon Moore articulated his eponymous law in 1965.
Moore’s Law, which observed that the number of transistors on silicon chips tends to double about every two years, drove six decades of progress by relentlessly shrinking the transistor.
However, with features now approaching atomic scale, that path is hitting physical and economic hurdles. Huawei’s Tau Scaling Law takes a different approach. Instead of obsessing over smaller transistors, it aims to reduce processing times at every level using transistor switching, circuit delays, chip architecture, interconnect topology, and even data center workloads.
Zhou Jianjun, a professor at Shanghai Jiao Tong University’s School of Integrated Circuits, said the theory offers a fresh road map. “Chip manufacturing no longer needs to rely so heavily on cutting-edge lithography tools, and the strategic role of advanced packaging continues to grow,” he said.
The background is instructive. China’s chip sector has faced mounting pressure, with the United States blocking shipments of extreme ultraviolet lithography tools and advanced chipmaking equipment from top suppliers.
Huawei said the Tau Scaling Law will allow it to produce a chip with transistor density equivalent to 1.4 nanometers by 2031. By comparison, TSMC, the world’s largest advanced chip foundry, plans to roll out 1.4 nm products in 2028.
Zhou Hongyi, founder of 360 Security Group, framed it as a long-awaited answer to a pointed question: “When others block China from cutting-edge chip technologies, does China still have a second path? Chinese semiconductors are moving from single-point breakthroughs to system-level collaboration across the entire industrial chain.”
If the Tau Scaling Law represents systematic thinking at the chip level, then the recent AI landscape is the evidence that the same logic works at the application layer.
Between June and early August — just eight weeks — China’s AI labs released five major models in rapid succession: Alibaba’s Qwen3.8-Max, Moonshot AI’s Kimi K3, DeepSeek-V4-Flash, Zhipu AI’s GLM-5.2, and ByteDance’s Seedance 2.5. On July 16, Kimi K3 debuted with a 2.8-trillion-parameter MoE (Mixture-of-Experts) large language model and a 1-million-token context window — making it the largest open-source model ever published globally.
The numbers bear this out. The Stanford AI Index Report 2026 showed the comprehensive performance gap between top Chinese and US models has narrowed to just 2.7 percent. The cost differential is even more telling. For equivalent workloads, some Chinese models cost a fraction of what US closed-source flagships charge.
But the deeper point is that these model breakthroughs are not accidents of isolated genius. They are downstream outcomes of a hardware-software feedback loop that is growing tighter by the month. Chinese AI labs are not waiting just for US-made chips — they are training more on domestic accelerators, whose market share in China’s AI accelerator market surged to 41 percent in 2025, according to International Data Corp, while Nvidia’s share fell to 55 percent.
The training data, the model architectures and the deployment scenarios are being optimized in parallel with domestic hardware, creating a flywheel that accelerates with each iteration, experts added.
Walk into an industrial park in Shenzhen, and you will see a 62-kilogram robot performing continuous backflips, side flips, and a spinning jump that launches it 87 centimeters off the ground. The leaping automaton is not a laboratory curiosity. It is the product of a supply chain so tightly integrated that the company can source all 30,000 of its components within half a day. That turnaround allows eight product iterations in just six months — a pace many overseas competitors cannot match.
Zhang Yin, vice-chairman and secretary-general of the Guangdong Artificial Intelligence and Robotics Industry Alliance, put it bluntly: “The Greater Bay Area (Guangdong-Hong Kong-Macao GBA) is the only region in the world that has both AI technology and mechatronics (derived from combination of “mechanics” and “electronics”) technology, and the ability to deeply integrate the two.”
This integration — between traditional manufacturing and cutting-edge intelligence — is the secret sauce that many outside observers miss. Wei Liurong, chief engineer of the policy and economics research institute at the China Academy of Information and Communications Technology, offered a vivid analogy.
“Traditional industries provide the ‘body’ and the ‘scenario’ for embodied intelligence. A humanoid robot needs high-precision reducers, servo motors, torque sensors, and lightweight structural components — the core manufacturing capabilities for these components reside precisely in traditional manufacturing enterprises,” Wei said.
At the same time, she added, traditional factories, warehouses, and mines are exactly where robots can first go to work after leaving the laboratory.
“The complexity of these scenarios and the feedback of real-world data are essential for robots to evolve from being able to walk to being able to work. If you only focus on the large language model for embodied intelligence while neglecting the foundation of traditional precision manufacturing, the robot will suffer from ‘cerebral palsy’ — a very smart brain with unresponsive limbs, unable to pass tests of precision, durability, and stability.”
Chen Yanbin, director of the Institute of Industrial Economics at the Chinese Academy of Social Sciences, said China’s manufacturing scale ranked first globally for the 16th consecutive year in 2025.
“The strategic underpinning for this has been its complete industrial system,” Chen said.
Hong Qunlian, a researcher at the Chinese Academy of Macroeconomic Research under the National Development and Reform Commission, placed this shift in a broader context.
“As a new round of technological revolution accelerates, global economic competition has shifted from product rivalry and corporate battles to a competition between entire industrial ecosystems. Whether China can seize opportunities and meet challenges by building a modern industrial system aligned with this technological revolution will determine our strategic position in future development,” Hong said.
Contact the writers at masi@chinadaily.com.cn