Alibaba

ByteDance

CATL

Huawei

Meituan

Tencent

Xiaomi

Most Followed

Market News

Listings (IPO)

Earnings Reports

Venture Roundup

Big Tech

Automotives

Retail and Commerce

Collections

News

Features

Insights

Press Releases

Anta

DeepSeek

Geely

Li Auto

Mixue Bingcheng

Pop Mart

Going Global

Latest

Collaborate

Earnings

Auto

Retail & Commerce

When most in the industry think of Unitree Robotics, they see a company focused on building robot hardware. But at the <a href="https://kr-asia.com/taking-the-pulse-of-chinas-humanoid-robot-push-at-wrc-2025">World Robot Conference (WRC)</a>, founder Wang Xingxing offered a different narrative.
During his keynote at WRC, Wang dedicated a large portion of his talk to large models, algorithms, and data. It was a shift that didn’t go unnoticed. His comments sparked debate, particularly his critique of the vision-language-action (VLA) framework driving many of today’s embodied robots.
Wang didn't mince words: he called the VLA architecture “relatively dumb.”
The main problem, in his view, is data, or the lack thereof. VLA models require vast, high-quality datasets to function effectively in the real world. While the scarcity of such data is widely acknowledged, many companies have pursued brute-force methods: gathering real-world robot data, generating simulation data, or building specialized data collection infrastructure.
Wang believes that emphasis is misplaced.
“People are paying too much attention to foundational data,” he said. “What we really need to improve is the architecture of embodied models. Right now, they are neither good enough nor unified enough.”
Historically, Wang has emphasized Unitree’s strength in hardware, not in building a robot “brain.” That messaging led many to assume the company wasn’t seriously investing in <a href="/industry/ai">artificial intelligence</a>.
But speaking to 36Kr during WRC, Wang clarified: “Our model team is actually quite large, it’s just smaller compared to the big AI firms.”
Still, Wang argued that size isn’t everything.
“It doesn’t take the most money or the biggest team to build world-class technology,” Wang said. “Innovation doesn’t only happen inside tech giants. Smaller teams can develop better models, though the pressure is much higher.”
Unitree’s model strategy remains flexible. While Wang is skeptical of current VLA approaches, he noted that Unitree is also experimenting with training AI on top of VLA frameworks. His core argument is that simply feeding more data into immature models is inefficient. A well-designed embodied model, he suggested, might perform well with only a limited amount of high-quality data.
The company’s bigger bet appears to be on video-driven models. Last year, Google introduced a world model trained on video, and Wang said Unitree was exploring similar ideas around the same time.
In this approach, a video generation model creates footage of a robot completing a task, such as tidying a room. That video is then used to guide a real robot in performing the same actions.
Wang believes this method may surpass VLA-based strategies in speed and convergence. Still, it comes with tradeoffs: generating high-resolution video is GPU-intensive.
Despite the technical demands, Wang sees a path forward. He envisions low-cost, large-scale, distributed computing clusters built specifically for <a href="/industry/robotics">robotics</a>. “If you’ve got 100 robots in a factory,” he said, “it makes sense to deploy a distributed server cluster inside the plant. The robots need low-latency communication anyway.”
Unitree’s robots, known for dancing at Lunar New Year galas and performing in high-octane battle scenes at this year’s World Artificial Intelligence Conference and WRC, have gained a reputation for showmanship rather than practical utility.
Meanwhile, newer entrants are focused on putting robots to work: tightening screws, folding laundry, making beds. That contrast has led some observers to dismiss Unitree’s robots as more spectacle than substance.
Wang disagrees:
<blockquote>“Right now, putting robots to work in homes or factories isn’t very realistic. Performance and demonstrations are just more achievable at this stage.”</blockquote>
He added that the team working on practical robotics is the largest within Unitree.
So why aren’t those use cases more visible?
“Developing practical robots is a massive challenge for AI models,” Wang said. “And honestly, we’re still not where we want to be.”
Wang’s vision for useful robots extends beyond narrow tasks like folding clothes or cooking. He sees them as general-purpose and multifunctional, such as being capable of pouring tea in a factory, then performing onstage.
Asked when robotics might see a breakthrough equivalent to AI’s “ChatGPT moment,” Wang offered a timeline: “The fastest would be two to three years. Slowest, maybe three to five. But this wave of embodied intelligence won’t take more than ten.”
What might that moment look like?
Picture a venue full of humanoid robots, walking freely. You stop one, give it a command, and it gets the job done. That, Wang said, would mark the tipping point.
KrASIA Connection features translated and adapted content that was originally published by 36Kr. This <a href="https://36kr.com/p/3420014135987845">article</a> was written by Qiu Xiaofen for 36Kr.

Unitree’s founder disputes VLA consensus, backs video-trained models for robotics

Artificial Intelligence

Robotics

Deep Tech

Embodied Intelligence

Foundation and Large Models

China

In a factory workshop in Chongqing, rows of mini tillers, generators, and energy storage devices are packed for shipment to destinations worldwide. Behind these seemingly ordinary machines is a manufacturing hub with outsized influence: the Chongqing general-purpose machinery (GPM) industrial belt produces and sells more than 14 million units annually—half of China’s total output—with mini tillers alone accounting for two-thirds of national production. These products serve traditional markets in eastern and central Europe and are increasingly reaching Southeast Asia and South America.
Once regarded as China’s de facto capital of mini tillers, this southwestern manufacturing cluster is transforming from a low-profile OEM (original equipment manufacturer) into a global innovation center for GPM.
“In the past, we’ve always been at the bottom of the smile curve. Now we’re stretching to both ends,” said Zhao Ming, deputy general manager of the general machinery division at Loncin Motor. He referred to the familiar model illustrating the challenge for Chinese manufacturing: low value-added assembly in the middle, with high-value R&amp;D and brand marketing concentrated in foreign companies.
Chongqing’s GPM industry has progressed through four phases: building a technical foundation during the Third Front Movement, expanding through private enterprise growth, integrating foreign capital for technology upgrades, and now, pursuing smart, green manufacturing.
The city has established a complete industrial chain, from parts production to final assembly. In agricultural machinery, Chongqing produces roughly two-thirds of China’s mini tillers. In GPM overall, annual production exceeds 14 million units, or half of the nation’s This integrated chain has become a competitive advantage internationally. In 2024, Chongqing’s electromechanical exports reached RMB 454.97 billion (USD 63.7 billion), up 8.4% year-on-year, with GPM products representing a notable share.
Unlike fast-moving consumer goods, GPM products are durable industrial equipment, which makes overseas expansion more challenging. “It’s not like we can create a hit product that sells thousands a day,” Zhao said. “Unit prices are high, and customers buy out of necessity.”
After-sales service poses an even greater hurdle. “Our products require users to have a basic level of mechanical knowledge,” said Xie Fuman, head of the smart terminal division and e-commerce marketing center at Senci. Generators and portable energy storage units require oil and gasoline, as well as regular maintenance—unlike plug-and-play electronics.
“After-sales problems can be life-or-death,” Zhao stressed. “High return rates can instantly wipe out profits.” Building service networks has therefore become a strategic priority for Chongqing’s GPM exporters.
To address these challenges, companies are testing solutions such as video-based guides, secondhand resale channels, and third-party service partnerships.
In response to global industrial shifts, Chongqing firms are also upgrading products. “Generators under 2,000 watts are gradually being replaced by energy storage products,” Zhao said. These devices are quieter, emission-free, and—thanks to advances in battery and chip technology—charge faster and last longer than traditional generators.
Artificial intelligence integration is emerging as another trend. While adoption is constrained by incomplete data, many in the industry see intelligent features as inevitable. “Energy storage devices will eventually become seamless backup power for homes,” Xie said. “When a power outage happens, users won’t even notice as the supply will switch over automatically.”
Senci, one of Chongqing’s major exporters, reported Amazon sales exceeding USD 40 million in 2024, with B2B transactions up more than 20%. Loncin Motor joined Amazon in 2024 and, within about three months, recorded RMB 4 million (USD 560,000) in additional sales. B2B orders from France, Spain, and Italy accounted for more than 10%.
Looking ahead, Chongqing companies are taking a measured view. “In the next five years, there’s unlikely to be a new product that can completely disrupt the GPM market,” said Haimaitong general manager Zhao Xin. He noted that competitiveness will depend on meeting user needs and avoiding closed-off development, ensuring that market feedback drives design.
Xie sees potential in untapped niche applications that could drive future growth. She also underscored the importance of brand building: winning customer loyalty through reliable products and service so that buyers return for upgrades and accessories.
Chongqing’s GPM export story has developed gradually, built on craftsmanship, market awareness, and sustained investment in innovation. These efforts have earned the trust of customers worldwide.
As Zhao put it: “Once we truly embrace independent innovation, the products we already make will naturally find their way to the world.”
KrASIA Connection features translated and adapted content that was originally published by 36Kr. This <a href="https://36kr.com/p/3414071119498889">article</a> was written by Ou Xue for 36Kr.

Chongqing’s machinery makers eye global markets, push for higher-value exports

Hardware

… players can’t afford to tackle everything at once—tech giants can.
For those still in the race, the next round of eliminations may only be just beginning.
KrASIA Connection features translated and adapted content that was originally published by 36Kr. This <a href="https://36kr.com/p/3411113687944579">article</a> was written by Zhou Xinyu for 36Kr.

China’s “AI tigers” return to the ring as the foundation model race reignites

European leaders can complain to Beijing all they want about cheap Chinese <a href="/industry/ev">electric vehicles</a> flooding the continent, but their resistance may be futile.
This summer, more and more transporters have been spotted on Germany's autobahns ferrying loads of <a href="/company/byd">BYD</a>, <a href="/company/leapmotor">Leapmotor</a>, and Great Wall Motor cars from ports to dealerships. On Swedish and British roads, <a href="/company/xpeng">Xpeng</a> and Polestar cars are also becoming more common.
Industry researcher Jato Dynamics published analysis recently showing that 5.1% of new vehicle registrations in the first half of this year were for Chinese brands in 28 European countries (including the UK), a record high and a near doubling from a year ago. To put that figure in context, new registrations for Mercedes—once a darling of luxury car owners—reached 5.2% in the first half.
BYD is a big winner, registering 70,500 electric and hybrid vehicles in that period, a whopping 311% year-on-year jump. In June, BYD was the tenth most registered new EV brand, ahead even of Ford. Industry experts across the continent told Nikkei Asia that there has been a dramatic shift in consumer perception.
"Whereas ten years ago, Chinese cars were often characterized as 'lethal' or 'dangerous' over failed crash tests, major Swedish auto magazines and newspapers are now giving Chinese EVs top marks," said Frederic Cho, China adviser and vice chair of the Sweden-China Trade Council. "Xpeng, <a href="/company/nio">Nio</a>, and <a href="/company/geely">Geely's</a> Zeekr and Polestar have established prominent concept or flagship stores in central Stockholm."
Cho expects Chinese brands to have gained 10–15% of Sweden's EV market over the next five years, despite "some lingering hesitation [that] exists among consumers due to a general negative perception of China for political reasons and concerns about personal information being sent to Beijing."
Chinese automakers may also have won some hearts in Sweden because Geely turned Volvo's fortunes around after it acquired the ailing iconic Swedish carmaker from Ford in 2010. Another factor, Cho said, is that Sweden's population is concentrated in urban areas covered by the Chinese companies' dealerships and workshops.
"This concentration means shorter distances to dealerships, unlike in Germany where a major share of the population lives in small towns that are covered by European, Korean, and Japanese brands but not by the Chinese," Cho said.
In neighboring Denmark, Jorn Gronkjaer, chair of the Danish EV Association, has also noted a progressively more positive attitude toward Chinese EV brands. The market share of Chinese cars in Denmark reached 5.5% in June, up from 4.4% at the end of 2024, according to Danish online car marketplace Bilbasen.
"Danes have traditionally shown strong solidarity towards European cars like Volvo and Volkswagen, but the availability of lower-priced EVs, notably from brands like BYD, Xpeng, and MG, has been crucial for making electric cars accessible to every family in Denmark," Gronkjaer said.
He projected that the market share for Chinese brands in Denmark could reach 20% in the next five years, particularly with innovation, quality improvements and potential acquisitions of other brands.
Even in the Swiss market that is still dominated by German brands, BYD has broken through.
"The BYD Dolphin Surf is the first Chinese car to make a splash in the Swiss media and generate interest, especially among young people who cannot afford more expensive models like the Audi e-tron and have grown up with high-quality Chinese electronics and Chinese e-commerce websites like Temu and Shein," said Holger Wahl, chair of the Electric Vehicle Club of Switzerland.
The organization estimates that the Chinese EV market share is still under 1%, but expects it to grow to 3% by the end of the decade. The Swiss, with higher disposable incomes, tend to go for Audi and VW cars.
But Wahl said that young Swiss people buy cars via apps like Autoscout 24, which levels the playing field for Chinese brands as they do not have the extensive dealership presence that European carmakers have.
With Chinese cars now increasingly a common sight on the continent and in the UK, where does this leave European carmakers? Several experts said they had confidence that European brands will hold their own over the long term.
"I'm quite trustful that some of the European brands will safeguard their large market shares," said Gronkjaer. "If you are driving on Danish roads, you will see Volkswagen's IDs, BMW i, Renault E-Tech models, and Ford Explorer much more than in many other places."
Wahl believes the luxury segment will still be dominated by European brands. "I see the Chinese potential more in the lower price range, that is, growing unit numbers with low revenue," he said.
Indeed, BYD has mass production in its sights to serve the European market. The company had set up operations in Hungary's Szeged city to avoid European Union (EU) tariffs on Chinese-made EV imports.
But Reuters reported recently that BYD was delaying mass production at Szeged to 2026 and was shifting much of the planned output to Turkey, which isn't in the EU but has a tariff-free trade agreement with the bloc.
This article first appeared on <a href="https://asia.nikkei.com/business/automobiles/electric-vehicles/cheap-and-sleek-chinese-evs-turn-european-heads">Nikkei Asia</a>. It has been republished here as part of 36Kr’s ongoing partnership with <a href="https://www.kr-asia.com/nikkei-teams-up-with-36kr-in-asia-tech-news-coverage/">Nikkei</a>.

Cheap and sleek Chinese EVs turn European heads

Transportation

Electric Vehicles

Mobility

Europe

… acquisition, and inventory management.
whereAdBanner.Uptake
<h3>Recent deals completed in China:</h3>
<ul>
<li aria-level="1">Zhendian Zhigan, a sensor tech company, has raised an eight-figure RMB sum in an angel financing round led by KylinHall Partners, with participation from Shuimu Tsinghua Alumni Seed Fund. The startup, founded this year, is developing new sensor technologies and energy harvesting platforms. Proceeds will go toward R&amp;D, test line construction, commercialization, and team growth. —<a href="https://36kr.com/p/3420889721834889">36Kr</a></li>
<li aria-level="1">BCIFlex, a life sciences company, secured over RMB 300 million (USD 42 million) in a Series funding A round co-led by a Zhongguancun fund (managed by Legend Capital Management) and IDG Capital. Other participants include Yuanbio Venture Capital, China Life Capital, Lenovo Capital, Meituan’s DragonBall Capital, and Baidu Venture. —<a href="https://36kr.com/p/3418367276437123">36Kr</a></li>
<li aria-level="1">LindenBot, based in Hefei, has closed a nine-figure RMB angel funding round for its general-purpose embodied robotics. The round was led by Oriza Hua and joined by Xunfei Ventures, PreIPO Capital, Yingke Private Equity, Tianzhi Investment, and Hefei Hi-tech VC. —<a href="https://36kr.com/p/3417090204945801">36Kr</a></li>
<li aria-level="1"><a href="https://en.szporos.com/">Porosvoc</a>, an audio tech company, has raised an eight-figure RMB in Series A funding from Hanlian Semiconductor Industry Fund. Founded in 2018, Porosvoc develops tri-core audio AI chips and algorithms for use in smart devices and toys. —<a href="https://36kr.com/p/3417773665488263">36Kr</a></li>
<li aria-level="1">Xunlu Innovation Technology, an electric bike manufacturer, has raised an eight-figure RMB in a funding round led by HongShan’s seed fund, with continued support from Shunwei Capital and other existing backers. The funds will go toward the mass production of its debut model and R&amp;D for new products. —<a href="https://36kr.com/p/3413659660160640">36Kr</a></li>
<li aria-level="1">Artificial Productivity, an AI hardware startup founded in 2024 by former 4Paradigm chief scientist Tu Weiwei, has secured a seven-figure USD sum in seed funding. Backers include Ameba Capital, SenseCapital, and 4Paradigm, with Yuanhe acting as financial advisor. The company has delivered its first companion robot and plans to launch a mass-produced smart toy for children next month. —<a href="https://36kr.com/p/3412589113527684">36Kr</a></li>
</ul>
whereRelatedArticle
<h3>SG Enviro, TopView, Oncode, and more made recent headlines:</h3>
<ul>
<li aria-level="1"><a href="https://sg-enviro.com/">SG Enviro</a>, a Singapore-based wastewater engineering company, secured SGD 8 million (USD 6.2 million) in funding led by Emerald Technology Ventures.</li>
<li aria-level="1"><a href="https://www.topview.ai/">TopView</a>, a Singapore-based platform that enables brands to create product content using artificial intelligence, raised USD 8.5 million in a Series A funding round led exclusively by Kamet Capital.</li>
<li aria-level="1"><a href="https://www.oncodesc.com/">Oncode Scientific</a>, a Malaysia-based molecular diagnostics company, received an undisclosed investment from the Malaysian Technology Development Corporation (MTDC).</li>
</ul>
If there are any news or updates you'd like us to feature, get in touch with us at: <a href="mailto:hello@kr-asia.com">hello@kr-asia.com</a>.
<style>.addcaption{font-size:14px!important;line-height:20px!important;color:inherit;padding-top:8px;border-top:1px solid #e8e8e8}</style>

Deals in brief: Naluri and Prefer raise funding, Antler invests in seven AI startups, six China deals, and more

Venture Capital

Private Equity

Investment

Merger &amp; Acquisition

Singapore

Southeast Asia

… data.
<h3>36Kr: In your view, what still lacks consensus in robotics?</h3>
GY: Plenty. The significance of actuators, for instance. Or which use cases will land first. Even the VLA framework is still evolving rapidly. Its core architecture may be settling, but implementation details are changing all the time, and fast.
<h2>Are large-scale data collection factories worth it?</h2>
<h3>36Kr: What’s your take on robotics companies building large-scale data collection factories? Isn’t there a problem if the data can’t transfer across different robot platforms?</h3>
GY: I don’t think large-scale data collection makes much sense at this stage. Robot hardware is still evolving quickly, and changes in form factor significantly reduce the reusability of collected data. That undercuts its value.
Also, at least with our algorithm, massive data collection isn’t necessary. What matters more is high-quality pretraining. Right now, some companies have it backward. They are overinvesting in data collection and underinvesting in foundational model quality.
<h3>36Kr: Some companies use data factories as part of their business model.</h3>
GY: In the short term, sure, it can bring in revenue. In the US, labor is expensive, so AI companies often buy data instead of collecting it. But long term, it’s a shaky model, especially when cross-platform generalization hasn’t been solved yet.
whereRelatedArticle
<h3>36Kr: If someone buys data that doesn’t match their robot’s hardware, is it still useful?</h3>
GY: It’s still useful, but you’re getting diminished returns.
<h3>36Kr: Many demos now focus on robots folding clothes or opening appliances. Why those tasks?</h3>
GY: Folding clothes is one of the most difficult challenges in robotics. Clothing is deformable and behaves unpredictably, so it can’t be programmed in advance. That makes it a strong benchmark as it shows how robust the model really is.
Tasks like opening fridges or washing machines are familiar household interactions. They help people envision practical use cases for robots.
<h3>36Kr: What proportion of your training data comes from the internet, and what roles do different data types play?</h3>
GY: Over 95% of our training data is sourced from the internet. This kind of data provides wide-ranging context and variability, which is ideal for pretraining. The goal is generalization: can the robot handle unfamiliar situations?
Teleoperation data connects that generalization with precise physical control. Watching isn’t enough; robots also need to practice. Teleoperation enables the kind of fine motor training they require.
<h3>36Kr: How does generalization show up in practice?</h3>
GY: Let’s say a robot trained only on iPhones encounters a foldable phone for the first time. If it can still identify the object’s shape, understand its physical properties, and manipulate it correctly, that’s generalization.
<h3>36Kr: Is the industry good at that yet?</h3>
GY: We’re still in the early days. But we’ve seen significant gains since incorporating internet video into training. Generalization rates on unfamiliar objects can improve by 60–80%. And when you layer in teleoperation data, those gains compound even further.
KrASIA Connection features translated and adapted content that was originally published by 36Kr. This <a href="https://36kr.com/p/3413531313409670">article</a> was written by Qiu Xiaofen for 36Kr.

Entrepreneurship is like “playing a game,” says Spirit AI co-founder Gao Yang

… for long-zoom, low-light concert recording.
And then there’s hospitality. Platforms like Trip.com and Klook now bundle concert tickets with flights and hotels. Outside stadium gates, Didi-labeled buses wait for fans headed home.
Taiyuan alone hosted 32 concerts in 2024, generating RMB 4.1 billion (USD 574 million) in consumption. Ticket sales made up just RMB 1.04 billion (USD 145.6 million), a quarter of the total. The rest came from everything else: the travel, food, shopping, and memories built around the event.
According to the China Association of Performing Arts, every RMB 1 (USD 0.14) spent on a ticket fuels an additional RMB 4.8 (USD 0.67) in related local spending.
Three golden hours onstage, and a whole economy is set in motion. Artists, managers, promoters, local governments, platforms, scalpers, food stalls, hotels—everyone gets a cut. Behind the scenes, it's a story of timing, culture, and people.
The roots of today’s concert boom lie in the legacy of Mandopop’s golden era. And few artists represent that legacy more clearly than Stefanie Sun. In 2025, she returned to the stage after a decade-long hiatus. She had changed. Her fans had changed. So had the entire industry.
Yifan and Tony attended her Shanghai concert. Yifan cried during the crowd’s singalong. Tony, who’d brought tissues just in case, ended up smiling the whole time. For both, the concert stirred memories from the past decade, ones they hadn’t realized they’d been holding on to.
KrASIA Connection features translated and adapted content that was originally published by 36Kr. This <a href="https://36kr.com/p/3398447154808962">article</a> was written by Wang Yuchan for 36Kr.

Three hours onstage, and an economy moves: Inside China’s concert boom

Entertainment

Music

Creative

… https://console.kr-asia.com/wp-content/uploads/2025/08/Spirit-AI-Moz1-World-Robot-Conference-2025-768x512.jpg 768w, https://console.kr-asia.com/wp-content/uploads/2025/08/Spirit-AI-Moz1-World-Robot-Conference-2025-1536x1024.jpg 1536w" sizes="(max-width: 1920px) 100vw, 1920px" /><figcaption id="caption-attachment-156503" class="wp-caption-text">Spirit AI’s Moz1 humanoid robot (pictured), demonstrated in various scenarios at WRC 2025. Photo source: 36Kr.</figcaption></figure>
Demonstrations, including walking, running, and acrobatics, as well as Unitree’s robot combat displays, were part of the program.
A key focus was potential real-world applications. Many booths presented deployment scenarios, with humanoid robots navigating the exhibition area or performing simulated operational tasks.
Industrial, service, logistics, and healthcare uses were prominent. Demonstrations included robots sorting parcels, moving crates, and providing retail or restaurant services, indicating adaptability and interactive capability in commercial settings.
In industrial automation, demand-side constraints are emerging despite mature technology. In his opening keynote, Takayuki Ito, president of the International Federation of Robotics, said 2024 has been a challenging year for industrial robotics. “Preliminary results show that global new installations fell 3% to 523,000 units. All three major regions reported a decline in investment, Asia down 2%, Europe down 6%, and the Americas down 9%,” he said. “Today’s humanoid robots still cannot work in collaborative modes. There remain many problems to solve before they can be commercially viable in such scenarios,” Ito added.
For embodied intelligence to gain broader adoption, manufacturers will need to improve collaborative functions, develop sector-specific solutions, and reduce costs.
The shift toward mass production is already reshaping pricing and distribution. At this year’s WRC, several new products showed lower price points. The Unitree R1 “Intelligent Companion,” starting at RMB 39,900 (USD 5,600), was the least expensive humanoid robot on display. Weighing about 25 kilograms, it supports custom development and integrates a multimodal model for voice and vision. Other consumer-oriented robots, including quadruped models, powered exoskeletons, and educational or companion units, were offered at even lower prices.
Distribution channels are also expanding. <a href="/company/jd">JD.com</a>, the WRC’s exclusive global strategic partner, launched a plan to accelerate industry development, committing more than RMB 10 billion (USD 1.4 billion) to support 100 robotics brands in reaching annual sales of RMB 1 billion (USD 140 million) each within three years, and to integrate robots into over one million user scenarios.
On August 9, Robot Mall, described as the world’s first embodied intelligence 4S store, opened in Yizhuang Industrial Park. The 4,000-square-meter venue displays more than 50 robot models from over 40 companies, including Skywork, UBTech’s Walker S, and Unitree’s G1 humanoid robot.
For industrial robot makers, overseas markets are well established. In the first half of 2025, China exported 94,200 industrial robots worth USD 746 million, a year-on-year increase of 59.74%.
Humanoid robot companies are following a similar approach. Several WRC exhibitors reported that a significant portion of their sales come from abroad, with clients in Europe, North America, Japan, South Korea, and the Middle East. Customers include multinational corporations, major technology companies, and academic institutions. Common market entry methods include attending international expos, responding to inbound inquiries, and working with distributors.
Robot Era, an embodied intelligence company incubated at Tsinghua University, receives more than half of its orders from overseas. Nine of the world’s top ten technology companies are among its customers.
<figure id="attachment_156505" aria-describedby="caption-attachment-156505" style="width: 1920px" class="wp-caption alignnone"><img loading="lazy" decoding="async" class="size-full wp-image-156505" src="https://console.kr-asia.com/wp-content/uploads/2025/08/Robot-Era-L7-World-Robot-Conference-2025.jpg" alt="Photo of Robot Era’s L7 bipedal humanoid robot on display." width="1920" height="1280" srcset="https://console.kr-asia.com/wp-content/uploads/2025/08/Robot-Era-L7-World-Robot-Conference-2025.jpg 1920w, https://console.kr-asia.com/wp-content/uploads/2025/08/Robot-Era-L7-World-Robot-Conference-2025-720x480.jpg 720w, https://console.kr-asia.com/wp-content/uploads/2025/08/Robot-Era-L7-World-Robot-Conference-2025-1024x683.jpg 1024w, https://console.kr-asia.com/wp-content/uploads/2025/08/Robot-Era-L7-World-Robot-Conference-2025-768x512.jpg 768w, https://console.kr-asia.com/wp-content/uploads/2025/08/Robot-Era-L7-World-Robot-Conference-2025-1536x1024.jpg 1536w" sizes="(max-width: 1920px) 100vw, 1920px" /><figcaption id="caption-attachment-156505" class="wp-caption-text">Photo of Robot Era’s L7 bipedal humanoid robot on display. Photo source: 36Kr.</figcaption></figure>
At a WRC forum, Rev Lebaredian, vice president of Omniverse and simulation at <a href="/company/nvidia">Nvidia</a>, said China’s advantages in robotics include <a href="/industry/ai">AI</a> R&amp;D capacity, a large talent pool, expertise in electronics and electrical engineering, and the ability to use manufacturing resources for mass production, deployment, and iteration.
Although technical barriers remain, Chinese companies are now significant participants in the global robotics sector. For many of the new developers, targeting overseas markets from the outset has become part of their operating strategy.

Taking the pulse of China’s humanoid robot push at WRC 2025

… contrast, have more flexibility. Agility has become their biggest advantage.
Generative AI researcher Zeng Yiqin summed it up well on X. “Manus is a good product,” Zeng <a href="https://x.com/qinzytech/status/1897505798974267730">wrote</a>. But he followed on by claiming that Manus does not represent a “technology breakthrough” because it mostly follows a predefined flow to execute tasks, and works in a “relatively constrained environment.” That’s emblematic of today’s leading AI agents: responsive builds layered on third-party models, tailored for product-market fit.
<h3>Speed and specificity</h3>
The Manus team had just 50 people and brought the product to market in three months. Lovart’s entire development cycle was even shorter: less than two months.
The window of opportunity is narrow. Among the wave of Manus-inspired clones claiming to be pioneers in agentic AI, only Lovart has managed to sustain real momentum.
But speed alone isn’t enough. Founders need to clearly define use cases, refine product structure, and improve user experience through thoughtful engineering.
Even when built on external models, well-designed wrappers still matter. “The shell has its value,” said Monica founder Xiao Hong—a phrase that has since become common shorthand in AI circles.
In a blog post published on July 18, Manus co-founder and chief scientist Ji Yichao (also known as Peak Ji) revealed that the team had already restructured its agent framework four times to enhance performance and scalability.
<h3>Is the real future in vertical use cases?</h3>
But how useful are general-purpose agents, really?
Even for relatively simple tasks like website generation or equity report parsing, users have reported that Manus still struggles with hallucinations, poor intent recognition, and inconsistent outputs. The polished demos often fail to reflect day-to-day performance.
Usage trends support that skepticism. Traffic for both Manus and Genspark has declined since their respective launches, suggesting that neither has yet achieved a stable product–market fit.
“As a paying user, I’d rather just use Claude 4,” one developer said. “The quality is better.”
This highlights a key vulnerability for startups in the general-purpose agent space: their heavy reliance on external models.
“Eventually, OpenAI or Anthropic will own general-purpose agents,” one investor told 36Kr. “For companies that control the best models, making AI agents is the easy part.”
Case in point: OpenAI’s own agent, “Deep Research,” is already available to ChatGPT Pro subscribers. Around 20% of users have upgraded to the USD 200 monthly plan just to access it.
As for the many Manus-inspired spinouts, the same investor was blunt: “They will pivot to vertical use cases.”
That shift is already underway. Several companies originally focused on specific verticals are now integrating agent-like tools into their offerings.
LiblibAI, which began as an image model community in 2023, now operates Lovart, a design-focused agent. Cai Haoyu, a co-founder of Mihoyo, recently launched <a href="https://kr-asia.com/mihoyos-ceo-is-back-with-a-new-game-and-its-nothing-like-genshin-impact">Whispers from the Star</a>, an AI-powered game where each character functions as an agent capable of dynamic storytelling through user interaction.
Model performance still falls short of supporting truly universal AI agents. But in vertical domains, where the context is more constrained and the needs are clearer, agents are starting to deliver meaningful value.
In these settings, domain knowledge and networked resources form the most defensible moats. They shape product design and determine how well agentic AI can be embedded into professional workflows.
For startups, this approach also provides a strategic buffer, offering a more sustainable path forward as big tech continues its advance.
KrASIA Connection features translated and adapted content that was originally published by 36Kr. This <a href="https://36kr.com/p/3392778755754121">article</a> was written by Zhou Xinyu for 36Kr.

Chinese AI agents made a splash in 2025, but can they go the distance?