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

… href="https://asia.nikkei.com/business/markets/ipo/unitree-shares-soar-in-humanoid-maker-s-landmark-shanghai-ipo" target="_blank" rel="noopener">high-flying Chinese startup Unitree Robotics</a>.
"Some of the startups depend on these less expensive products coming from other countries to build their own products," the association's Burnstein said.
"It's going to be harder for them to grow because the costs are going to go up," he said. "The availability might go down."
In the longer term, the fresh curbs could also backfire in creating an unstable regulatory environment that deters investment, with many companies expecting the FCC to add more products to the Covered List.
"When you have so much that's done through an executive order or you have these other processes that lack durability that can be undone by the next administration, it makes investment in the US kind of difficult," the Global Electronics Association's O'Hanlon said.
This article first appeared on <a href="https://asia.nikkei.com/business/technology/us-robot-curbs-lay-bare-industry-s-dependence-on-chinese-parts">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>.

US robot curbs lay bare industry's dependence on Chinese parts

China

Robotics

… core algorithm researchers from Z.ai’s research institute, and still failed.”
One reason is compensation.
As Z.ai’s market capitalization rose, stock options held by some key researchers came to be worth more than RMB 1 billion (USD 148.5 million), according to people familiar with the company.
But employees and recruiters said another factor was the network of academic and personal relationships connecting Z.ai with its core talent. Those ties are difficult for competitors to replicate.
Much of the company, from its founder to employees across key roles, has roots at Tsinghua University. Many are connected through shared academic advisers, classmates, or teacher-student relationships.
“It’s a combination of a company, a university, and a research institute,” one investor said.
For an organization centered on model training, one person is especially important: founder and chief scientist Tang Jie.
A professor in Tsinghua University’s Department of Computer Science and Technology, Tang was among the earlier scholars in China to work on machine learning and natural language processing.
Several sources told 36Kr that when they were looking in late 2022 for Chinese companies that could become equivalents of OpenAI, the first Tsinghua academics who came to mind were Tang Jie, Sun Maosong, and Huang Minlie.
Many Z.ai employees describe Tang as enthusiastic, persuasive, and execution-oriented. They also said he sets demanding standards for both technology and business.
One algorithm researcher at Z.ai’s R&amp;D institute said that when training results for its agent product AutoGLM 2.0 were unsatisfactory, Tang postponed its launch.
The company’s core team is also connected through Tsinghua.
CEO Zhang Peng, who runs commercialization, earned his bachelor’s, master’s, and doctoral degrees at the university and, like Tang, comes from its computer science department.
Z.ai president Wang Shaolan, who is responsible for government relations, and chairman Liu Debing also graduated from Tsinghua University. Both previously worked with members of the founding team at the university’s Big Data Research Center.
More than 70% of the hundreds of researchers at Z.ai are current or former undergraduate, master’s, or doctoral students from Tsinghua’s computer science department, according to 36Kr.
That concentration of teachers, graduates, and students has created a relatively stable algorithm organization shaped by adviser-student relationships and a strong academic culture.
It is visible in the way the R&amp;D institute works. The institute has a flat hierarchy and operates more like a laboratory than a conventional corporate organization, according to employees. Important directions often begin with a technical question or a discussion about an experimental result. Researchers from different backgrounds join the discussion, while younger researchers have considerable room to challenge assumptions and test ideas.
Many students told 36Kr they would be reluctant to leave because of the respect and personal attachment they feel toward their teachers and senior schoolmates.
“Some students who want to pursue master’s or doctoral degrees under Tang will choose to work at Z.ai first,” a former Z.ai algorithm researcher said.
A steady supply of new talent, combined with relatively low turnover, has become one of Z.ai’s organizational advantages.
According to 36Kr, after GLM 5 launched, one Z.ai algorithm intern received nearly ten calls from headhunters in a single day.
ByteDance recruiters have gone through author lists on GLM research papers and contacted researchers individually.
<a href="/company/tencent">Tencent's</a> Hunyuan team was willing to offer one Z.ai master’s graduate twice the base salary to move, according to 36Kr.
On August 19, Tang Jie posted on X to explain his view of the scaling law. Model capabilities, he argued, are no longer determined simply by parameter count. Instead, model development increasingly resembles a controlled experiment in which the variables include how much data is used, where the compute budget is spent, and who ultimately runs the model and under what conditions.
That argument helps explain the work behind GLM 5.3.
Compared with its predecessor, the model’s gains came primarily from post-training rather than simply increasing its pretraining parameter count, according to 36Kr.
Other leading foundation model companies, meanwhile, have continued increasing model size. Over the past month, Moonshot AI, DeepSeek, and <a href="/company/alibaba">Alibaba Group</a> have all released new flagship models, with parameter counts ranging from 160–280 million, two to four times that of GLM 5.3.
Among them, Kimi K3, released by Moonshot AI on July 17, displaced GLM 5.2 after the latter had held the open-source SOTA position for about a month.
Just over 20 days later, Z.ai reclaimed the top position with GLM 5.3.
The broader race remains unsettled. Model direction, training routes, data quality, and algorithm talent all continue to shape how foundation model companies compete.
For Z.ai, coding was the capability that aligned those pieces at the right time.
KrASIA features translated and adapted content that was originally published by 36Kr. This <a href="https://36kr.com/p/3947451908357257">article</a> was written by Zhou Xinyu and Zhang Yuxin for 36Kr.
Note: HKD, RMB figures are converted to USD at rates of HKD 7.84 = USD 1 and RMB 6.74 = USD 1 based on estimates as of September 1, 2026, unless otherwise stated. USD conversions are presented for ease of reference and may not fully match prevailing exchange rates.

Inside Z.ai’s turnaround after falling behind in enterprise AI

East Asia

Z.ai

Artificial Intelligence

Deep Tech

Foundation and Large Models

… href="https://36kr.com/p/3957651741949056">36Kr</a></li>
<li aria-level="1">Xingyi Space, a commercial space telemetry, tracking, and control company, has raised a nine-figure RMB sum in a Series A round. Investors include Suzhou Capital Group, Lihe Qingtong, Septwolves Holdings, Zhichen Investment, Lanxi Capital, and Haisi Venture Capital. The company provides in-orbit tracking and control services for satellites and launch vehicles. It will use the funding primarily to expand its overseas ground station network. The round follows a nearly RMB 300 million (USD 44.5 million) pre-Series A round led by SCGC in late 2025. —<a href="https://36kr.com/p/3956027533507970">36Kr</a></li>
<li aria-level="1">Lingluo Technology, a Shenzhen-based sleep technology company, has raised an eight-figure RMB sum in an angel round led by Shenzhen HTI Group, with existing investor Heding Co Investment also participating. Founded in 2024, the company develops AI-enabled wearable hardware for sleep intervention, combining neural signal collection with AI-based decision models. Its first product, an ankle-worn sleep device, went on sale in China in July. Lingluo will use the funding for technology development, supporting hardware and software systems, commercialization, market expansion, and scaled production and delivery. —<a href="https://36kr.com/p/3954623000952197">36Kr</a></li>
<li aria-level="1">LingYu Robotics, an embodied AI infrastructure provider, has raised a nine-figure RMB sum in a pre-Series A round from Zhuzhou Industrial Development Investment Holding Group, Flyfot Ventures, Future Margin Ventures, KHK Fund, and other investors. Founded in 2025, the company develops robot hardware, real-world data collection systems, and the Nexus cloud operations platform. Its robots have been deployed at scale in 7Fresh supermarkets. LingYu will use the funding to expand production capacity, upgrade its real-world data collection pipeline, and continue developing its cloud operations platform. —<a href="https://36kr.com/p/3948044825279621">36Kr</a></li>
<li aria-level="1">Synapath AI, a Shenzhen-based embodied AI infrastructure company, has completed two consecutive angel plus funding rounds. The first raised an eight-figure RMB sum from KylinHall Partners, Shenzhen HTI Group, and an unnamed investor. The second raised nearly RMB 100 million from Shenzhen Co-Stone Asset Management, Nanshan SEI Investment, Suzhou Jushi Venture Capital, COE Capital, Hangzhou Industrial Investment Group, and XHVC. Founded in 2024, Synapath converts human behavior in real-world work settings into training data and physical representations that robots can learn from, supporting embodied AI model training and deployment. —<a href="https://36kr.com/p/3954612909227138">36Kr</a></li>
</ul>
whereRelatedArticle
<h3>Buddy Bites, Living Roots, Craif, and others made recent headlines:</h3>
<ul>
<li aria-level="1">Buddy Bites, a Hong Kong-founded pet food company, raised USD 4.2 million in a Series A round led by Digitalis Ventures, with participation from Hong Kong venture investor Adrian Lai.</li>
<li aria-level="1">Living Roots, a Thailand-based agriculture technology company developing biological alternatives to synthetic fertilizer, raised seed funding from a group of returning and new angel investors under Epic Angels. The amount was not disclosed.</li>
<li aria-level="1">Craif, a Tokyo-based biotechnology company developing urine-based tests for early disease detection, closed a roughly USD 33 million Series D round. The financing was co-led by Granite-Integral and Tauns. Unreasonable and existing investor X&amp;KSK also participated, alongside institutional and individual investors from Singapore, the US, and Japan.</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>.
Note: RMB figures are converted to USD at rates of RMB 6.74 = USD 1 based on estimates as of September 1, 2026, unless otherwise stated. USD conversions are presented for ease of reference and may not fully match prevailing exchange rates.

Deals in brief: Bioactivx raises pre-Series A funding, SMBC and Singtel Innov8 back fileAI, latest China investments, and more

Venture Capital

Private Equity

Investment

Merger &amp; Acquisition

Singapore

Indonesia

Philippines

Japan

Southeast Asia

Chinese jewelry brand Laopu Gold downplayed headwinds on August 26 after its earnings missed estimates, stressing its continued growth through a rough patch for the industry.
The previous day, the Beijing-headquartered Laopu reported a 60.3% increase in first-half revenue to RMB 19.8 billion (USD 2.9 billion), while net income rose 88.2% to RMB 4.27 billion (USD 633.9 million). While those growth rates would be enviable for many companies, they were far below the figures Laopu reported for the same period last year, and fell short of analysts' estimates as a pullback in gold prices weighed on Chinese jewelry demand.
"Faced with a complex macroeconomic environment and gold-price cycles in the first half of 2026, Laopu Gold has further widened its lead over both international and domestic jewelry brands," Feng Jianjun, executive director and vice general manager, told an earnings conference on Wednesday, citing average sales of more than RMB 500 million (USD 74.2 million) per store during the six-month period.
Feng added that as of June 2026, despite a 41% decline in China's gold demand in the first half of the year, membership of Laopu's loyalty program climbed to 730,000, a 20% increase from 2025.
Dubbed the "Hermes of gold" by some, the brand became a <a title="" href="https://kr-asia.com/chinas-hermes-of-gold-courts-turbulent-world-after-dazzling-at-home" target="_blank" rel="noopener">darling of investors and shoppers</a> last year, attracting long lines at its boutiques while its share price soared. Its stock performance has been far less impressive this year. When Laopu published preliminary first-half results in late July, which were weaker than expected, its share price fell 24% the next day. The company's shares in Hong Kong are down about 35% since the start of 2026, although they were up about 2.6% in trading on August 26.
The company did not give breakdowns of its quarterly earnings. But based on its first-quarter sales forecast—revenue of approximately RMB 16.5–17.5 billion (USD 2.4–2.6 billion), with net profit of around RMB 3.6–3.8 billion (USD 534.5–564.1 million)—its second-quarter revenue is estimated to have been RMB 2.31–3.31 billion (USD 342.9–491.4 million), with net profit of about RMB 470–670 million (USD 69.8–99.5 million). That would mark a sharp sequential decline.
In a July note, HSBC said that Laopu's slowdown in the second quarter was caused by the correction in spot gold prices and the widening of its product price premium over spot gold.
Contrary to standard industry practice, Laopu does not price its products by weight or adjust them in line with global gold price fluctuations. Instead, it typically raises listed prices two to three times a year. Laopu insists it is resilient against gold price fluctuations because of its brand premium.
As sales came under pressure, Laopu ramped up spending on sales and distribution by 50% to RMB 2.2 billion (USD 326.6 million) in the first half. Even so, its gross profit margin improved to 41.3%, from 38.1% a year earlier, as the company had built up a cushion of relatively low-cost inventory heading into the year, and it raised prices at the end of February.
Since July, the company has rolled out a series of measures to spur sales, including bundled gifts, enhanced member perks, steeper discounts and new products priced more closely to prevailing gold prices. The company has also completed refurbishments and upgrades across eight of its stores in mainland China since the start of the year.
While the slowdown highlights Laopu's sensitivity to swings in gold's value, Morningstar said in an earlier note that consumers tend to favor weight-based products over fixed-price jewelry when prices are falling. The financial services firm expects gold prices to recover more moderately amid geopolitical uncertainty.
Laopu struck a positive tone on its outlook.
"The pullback in the second quarter was a rare and unusual phenomenon," founder and Chairman Xu Gaoming said at the briefing. "After the test of the second quarter, we are confident that Laopu has already established strong brand pricing power."
Xu forecast that gold prices will continue to climb from the third quarter through next year.
"Gold is back," he added. "And in a very strong way."
The overlap rate between the homegrown brand's consumers and those of leading international luxury houses, such as Louis Vuitton, Hermes and Cartier, rose from 77.3% in July 2025 to 84.6% in August 2026, according to the company.
Laopu Gold will focus on expansion in Southeast Asia for the rest of the year, while planning "aggressive" inroads in overseas markets.
Xu said the company is currently in discussions with partners in Singapore, Malaysia, and Thailand. It also secured an outlet in Tokyo's Ginza district, he said, but concluded that "it is not the best time" to enter the Japanese market.
In addition, Laopu has held talks with a luxury shopping mall in Los Angeles and is looking to open stores in the US, Canada, France, and the UK next year.
This article first appeared on <a href="https://asia.nikkei.com/business/retail/laopu-gold-stresses-bright-outlook-and-global-plans-as-growth-slows">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>.
Note: RMB figures are converted to USD at rates of RMB 6.74 = USD 1 based on estimates as of August 31, 2026, unless otherwise stated. USD conversions are presented for ease of reference and may not fully match prevailing exchange rates.

Laopu Gold stresses bright outlook and global plans as growth slows

Retail

Consumer Economy

Hong Kong-listed UBTech Robotics released its 2026 interim results on August 28, with its full-size humanoid robot business emerging as its largest source of revenue.
In the first half, total revenue rose 104.2% year-on-year to RMB 1.27 billion (USD 188.5 million). Revenue from full-size embodied humanoid robot products and services surged 1,445% to RMB 590.3 million (USD 87.6 million) from RMB 38.2 million (USD 5.7 million) a year earlier, accounting for 46.5% of group revenue. Sales volume in the category reached 921 units, up 1,946.7%.
The shift in revenue mix also lifted profitability at the gross level. Gross profit increased 160.9% to RMB 566.9 million (USD 84.2 million), while gross margin rose 9.7 percentage points to 44.7%. UBTech attributed the margin improvement mainly to the higher contribution from full-size humanoid robot products and services, which carry higher gross margins.
The company nevertheless remained loss-making. Its net loss narrowed 23.0% year-on-year to RMB 338.8 million (USD 50.3 million), while its adjusted EBITDA loss narrowed to RMB 174.1 million (USD 25.8 million) from RMB 321.8 million (USD 47.8 million) a year earlier.
UBTech also continued to spend heavily on product development. R&amp;D expenses rose 38.9% to RMB 303.1 million (USD 45 million) in the first half, equivalent to 23.9% of revenue. The company said the increase was mainly driven by continued investment in full-size humanoid robots.
Much of that investment is directed toward moving humanoid robots beyond demonstrations and into industrial use.
UBTech said its industrial deployments focus on tasks including material handling, loading and unloading, sorting, palletizing, and depalletizing. During the first half, it validated solutions for several of these applications while combining technologies including teleoperation and vision-language-action (VLA) models to address different industrial tasks.
The company has also been expanding deliveries of its Walker S2 industrial humanoid robot and deployments of its wheeled Cruzr Y1. Its industrial robots have been tested or deployed in sectors including automotive manufacturing, logistics, and aviation. Airbus, for example, acquired a Walker S2 earlier this year as part of a partnership exploring the use of humanoid robots in aircraft manufacturing.
UBTech is broadening its lineup at the same time. In the first half, it launched the Cruzr Y1 for manufacturing and warehousing, the bipedal Walker C1 for commercial services, education, and research, and the <a href="https://kr-asia.com/ubtechs-uworld-u1-tests-demand-for-humanoid-robots-at-home">U1 series</a> for companionship applications.
whereRelatedArticle
The three product families reflect UBTech’s attempt to build across industrial, commercial, and eventually household applications rather than rely on a single market.
Industrial use remains the most developed of the three. UBTech said it is working to move from individual task validation toward larger deployments in which multiple robots can perform longer sequences of work and collaborate on production lines.
Supporting that effort is the company’s <a href="/industry/ai">artificial intelligence</a> stack.
UBTech released Thinker 1.0, its foundation model for embodied intelligence, in the first half. The company said the model ranked first on nine embodied intelligence leaderboards. It also introduced Thinker-WM, a world model designed to help robots model physical environments and anticipate the outcomes of actions. According to UBTech, Thinker-WM ranked first on the Libero embodied intelligence benchmark during the period.
Its Thinker-VLA model, meanwhile, combines visual perception, language-based instructions, and action generation. UBTech has been applying VLA technology to industrial handling and loading tasks, where robots need to adjust their actions as objects and surroundings change.
The company is also developing BrainNet 2.0 and Co-Agent, technologies intended to coordinate multiple robots. UBTech said the system can divide tasks, schedule operations, share information, and coordinate work among humanoid robots, extending its approach from single-robot autonomy toward multi-robot collaboration.
Data is another part of the strategy. UBTech is building humanoid robot data collection and testing centers and a platform designed to connect data collection, annotation, model training, and deployment. It is also developing a simulation platform that can reconstruct scenarios, generate assets and synthetic data, and test models before deployment on physical robots.
As of June 30, UBTech held 3,112 granted patents, including 530 overseas patents, up 4.2% from the end of 2025.
UBTech is also expanding beyond robot hardware itself.
In the first half, the company completed its acquisition of a 43.01% stake in Shenzhen-listed Zhejiang Fenglong Electric, becoming its controlling shareholder. Fenglong’s financial results have been consolidated into UBTech’s accounts since April, contributing RMB 139.2 million (USD 20.7 million) in revenue from garden machinery, automotive, and hydraulic components during the period.
Separately, UBTech and Chinese GPU developer MetaX established a joint venture to develop and produce chips for embodied intelligence, extending the company’s involvement further into the hardware stack.
UBTech is also working with Siemens on digital tools for robot design and manufacturing. The partnership is intended to support UBTech’s effort to scale production. UBTech has set an annual production target of 10,000 full-size humanoid robots for 2026.
For the second half, UBTech plans to continue expanding Walker S industrial models while developing its Walker C and U1 series for commercial, educational, and household applications. It also plans to release an upgraded Thinker-WM 2.0 world model and continue improving its VLA and multi-robot coordination technologies.
This <a href="https://xueqiu.com/2854806290/407260459">article</a> was adapted based on a feature originally written by Stone Jin and published on IPO Zaozhidao. KrASIA is authorized to translate, adapt, and publish its contents.
Note: RMB figures are converted to USD at rates of RMB 6.74 = USD 1 based on estimates as of August 31, 2026, unless otherwise stated. USD conversions are presented for ease of reference and may not fully match prevailing exchange rates.

UBTech’s full-size humanoid robot revenue jumps 1,445% in H1 2026

Embodied Intelligence

Hong Kong

ChaPanda released its interim results on August 28. Revenue for the first half rose 6.2% year-on-year to RMB 2.66 billion (USD 394.9 million), while profit for the period increased 3.5% to RMB 345 million (USD 51.2 million).
At its earnings briefing, management said the company was focusing less on expanding store count for its own sake and more on improving store-level operations through product development and supply chain efficiency.
ChaPanda stepped up product R&amp;D in the first half, building out a portfolio centered on ingredients including fresh fruit, fresh milk, and coffee.
Its “Litchi Iced Milk” sold more than 17 million cups in just over 50 days, according to the company. ChaPanda subsequently introduced other iced milk drinks featuring flavors including guava, grape, and lemon, expanding the category into a more distinct product line.
The company has also been broadening the occasions its products are designed to serve. Since the start of 2026, ChaPanda has introduced items positioned for light meals, including a cassava-based dessert bowl, as well as drinks marketed for everyday hydration, such as a bergamot-and-lemon beverage.
The additions extend ChaPanda’s menu beyond the traditional afternoon tea occasion and into areas including breakfast, light meals, and everyday refreshment.
A broader product range is also placing greater demands on its supply chain, particularly for fresh ingredients.
During the reporting period, ChaPanda increased the number of its warehousing and distribution centers nationwide to 27, while the share of fruit handled through its centralized distribution system rose to 91%. About 95% of its stores can now receive deliveries at least twice a week.
The company said its cold-chain network, which connects production areas with stores, helps maintain product quality and improve distribution efficiency.
ChaPanda has continued expanding in lower-tier cities while maintaining its presence in first- and second-tier markets. By the end of the reporting period, its store count had increased to 8,863. The number of stores in China’s third-tier and lower cities rose 7.2% year over year.
The company is also using its store network to expand its coffee business alongside tea. In the first half, freshly ground coffee machines had been installed in more than 2,700 ChaPanda stores, ahead of the company’s expectations at the beginning of the year.
Meanwhile, ChaPanda’s registered membership surpassed 200 million, giving the company a large base of customers it can reach directly through its own channels.
Looking ahead, the company’s growth will depend in part on whether its coffee business can develop into a meaningful profit contributor and whether it can convert its growing membership base into stronger repeat purchases.
For now, ChaPanda’s first-half results suggest that its focus is shifting from simply adding stores toward getting more from the network it already has, through a broader menu and a more efficient supply chain.
KrASIA features translated and adapted content that was originally published by 36Kr. This <a href="https://36kr.com/p/3960415196052873">article</a> was written by Yang Shuo for 36Kr.
Note: RMB figures are converted to USD at rates of RMB 6.74 = USD 1 based on estimates as of August 31, 2026, unless otherwise stated. USD conversions are presented for ease of reference and may not fully match prevailing exchange rates.

ChaPanda boosts H1 2026 performance with new products, supply chain efficiency

CaoCao Mobility released its interim results for the six months ended June 30, 2026, on August 27. Its ride-hailing business continued to grow in the first half, while the company stepped up investment in <a href="/industry/autonomous-vehicles">autonomous driving</a> and <a href="/industry/ai">artificial intelligence</a> through its RoboX strategy.
Mobility services revenue rose 13.9% year-on-year to RMB 9.8 billion (USD 1.46 billion), while total revenue increased 9% to RMB 10.3 billion (USD 1.53 billion). As of June 30, CaoCao operated in 215 cities across China, including 20 added during the first half.
Average monthly active users reached 44.6 million, up 17.1% year-on-year, while average monthly active drivers increased 36.8% to 758,000.
Profitability also improved. Gross margin rose to 9.0% from 8.7% a year earlier. The company attributed the increase partly to greater operating scale and said it continued to use AI to improve the algorithms behind CaoCao Brain, its transaction and dispatch system.
In June, CaoCao unveiled its RoboX strategy and announced a broader push into AI-powered mobility. The company said it plans to combine purpose-built vehicles, autonomous driving technology, and fleet operations as it develops services including robotaxis and autonomous delivery vans.
CaoCao is building on capabilities it has developed in vehicle management, ride dispatch, safety operations, and customer service, with plans to apply them to a wider range of autonomous mobility services.
As of June 30, the company had deployed 140 second-generation robotaxis. It plans to expand its robotaxi fleet in China and overseas in the second half while also exploring mobility services that combine autonomous vehicles with low-altitude aircraft.
CaoCao is also working with <a href="/company/geely">Geely</a> and other partners on Eva Cab, its third-generation vehicle designed specifically for robotaxi services. The vehicle was unveiled in the first half and is scheduled to enter mass production in 2027.
The company is also experimenting with AI-based customer interfaces. In June, CaoCao began a limited pilot of an AI ride-hailing service developed with Doubao after joining the AI platform’s mobility ecosystem.
Overseas expansion is another part of the strategy, with Hong Kong and the UAE among CaoCao’s initial target markets.
In Hong Kong, the company plans to establish a joint venture with Octopus to develop technology-enabled mobility services.
In June, CaoCao also signed a cooperation agreement with Abu Dhabi-based K2 to advance the deployment and operation of robotaxi projects in the UAE.
This <a href="https://xueqiu.com/2854806290/407091520">article</a> was adapted based on a feature originally written by Stone Jin and published on IPO Zaozhidao. KrASIA is authorized to translate, adapt, and publish its contents.
Note: RMB figures are converted to USD at rates of RMB 6.74 = USD 1 based on estimates as of August 31, 2026, unless otherwise stated. USD conversions are presented for ease of reference and may not fully match prevailing exchange rates.

CaoCao Mobility looks to robotaxis for growth after H1 2026 revenue clears RMB 10 billion

Transportation

Autonomous Vehicles

Mobility

On August 27, Hong Kong-listed Manycore Tech released its interim results for 2026.
During the reporting period, the spatial intelligence company generated RMB 405 million (USD 60.1 million) in revenue, while its overall gross margin rose to a record 83%. Adjusted net profit reached RMB 55.42 million (USD 8.2 million), up 211% year-on-year and approaching the level recorded for all of last year.
The same day, Manycore launched Lux3D, a 3D generation model, and began internal testing of a new-generation world model product. The technology announcements lifted investor sentiment, with the company’s shares closing 6.97% higher.
Beyond the earnings release, Manycore could also become eligible for Southbound Stock Connect in early September, potentially broadening its investor base and improving the stock’s liquidity and market visibility. The company was recently selected as a constituent of the Hang Seng Composite Index, with the change set to take effect on September 7.
More notable than the improvement in profitability is the way spatial intelligence is moving from a technology foundation toward commercial monetization.
Three details stand out:
<ol>
<li aria-level="1">First, revenue from <a href="/industry/ai">artificial intelligence</a> applications surged 177% year-on-year, suggesting the company’s new AI products and applications have begun establishing a growth trajectory and that its efforts to commercialize spatial intelligence are gaining traction.</li>
<li aria-level="1">Second, capital expenditure rose 124% year-on-year, reflecting the company’s commitment to full-stack investment. By prioritizing compute capacity, Manycore is laying the groundwork for the next phase of business growth.</li>
<li aria-level="1">Third, the company disclosed its average daily token consumption, signaling that usage of its underlying models is scaling rapidly and that its strategy of becoming a midstream infrastructure provider is beginning to take shape.</li>
</ol>
More specifically, LuxReal, an AI video creation platform launched in the first half of the year, combines Manycore’s proprietary spatial generation model SpatialGen with 3D technology. It is designed primarily to address spatial consistency in AI-generated video and can rapidly create long-form narrative videos with a high degree of spatiotemporal consistency.
In June, LuxReal’s monthly registered users increased by more than 270% from the previous month, while monthly revenue rose by more than 470%. Both user numbers and revenue have climbed rapidly since the product launched.
In physical AI, SpatialVerse, which provides virtual training environments for AI agents, secured RMB 6.8 million (USD 1 million) in orders during the first half of the year, already exceeding its total for all of last year.
The product can generate high-fidelity, physically accurate synthetic data to meet training data requirements in areas including world models and embodied intelligence. SpatialVerse has established partnerships with Chinese physical AI companies including Agibot, Galbot, Noematrix, AI² Robotics, Songying Technology, and Hesai Technology. It also maintains close industry and academic collaborations with several major overseas technology companies.
As physical AI adoption accelerates, the business could continue expanding its addressable market in China and overseas, becoming an important growth engine in Manycore’s push to build infrastructure for physical AI.
In addition, leveraging its proprietary spatial intelligence foundation models and high-quality 3D data, Manycore has integrated capabilities including spatial reconstruction, generation, editing, and understanding into its Aholo spatial intelligence platform, which it has opened to external users.
The platform serves industries including entertainment and media, AI agent training, industrial digital twins, gaming, and spatial design. Its core objective is to make Manycore Tech an infrastructure provider for spatial intelligence.
The company has established partnerships with Arashi Vision, Pico, Hesai, and Huace Film &amp; TV, among others, as it continues to broaden the range of applications for spatial intelligence.
To support this infrastructure strategy, Manycore is also continuing to develop an integrated full-stack technology system spanning compute, models, and applications.
At the compute layer, the company’s self-built GPU cluster processed an average of about 12.9 million computing tasks per day during the first half of the year.
At the model layer, in addition to the newly launched Lux3D model and the world model product currently undergoing internal testing, Manycore Tech has previously open-sourced SpatialLM, a spatial understanding model, and SpatialGen, a world generation model.
For world model development, the company is pursuing an explicit 3D approach, with a focus on improving core capabilities including dynamic interaction, stable consistency, and editability. These models are intended to provide underlying support for applications including AI video, gaming, embodied intelligence training, and digital twins.
whereRelatedArticle
As one of the relatively few Hong Kong-listed companies developing world models, Manycore Tech is extending the scope of Hong Kong’s AI sector from language-based foundation models toward applications involving the physical world.
At the application layer, the token-based business model currently appears capable of becoming a new source of revenue growth.
Under this system, deployed applications generate large volumes of 3D data. That data feeds back into model training, and improvements in model capabilities can then further enhance the company’s products, creating a reinforcing flywheel.
As Manycore continues iterating its world models, accumulating 3D digital assets, and deploying its technology across more industries, its strategy of becoming an infrastructure provider for spatial intelligence could move more quickly toward commercial realization.
Note: RMB figures are converted to USD at rates of RMB 6.74 = USD 1 based on estimates as of August 31, 2026, unless otherwise stated. USD conversions are presented for ease of reference and may not fully match prevailing exchange rates.

Manycore steps up spatial intelligence push as AI product revenue jumps 177%

<a href="/company/anta">Anta Sports</a> delivered results that were largely well received by the market. In the first half of 2026, the group’s revenue rose 12.9% year-on-year (YoY) to RMB 43.5 billion (USD 6.5 billion), marking the first time its first-half revenue exceeded RMB 40 billion (USD 5.9 billion). Overall gross margin increased 0.5 percentage points to 63.9%, while operating profit climbed 16.1% to RMB 11.8 billion (USD 1.8 billion). Its operating margin rose to 27%, the highest level for a first half in nearly seven years.
Anta Group co-CEO Lai Shixian said at the earnings briefing that performance was broad-based across its brands. Growth at the Anta brand and its other brands exceeded the full-year guidance set at the beginning of the year, while Fila China reached the upper end of its guidance range. Lai said the results reflected the advantages of the group’s multibrand portfolio.
CFO Bi Mingwei added that the group maintained steady profitability even as revenue exceeded RMB 40 billion. Excluding the one-off gain from the placement of Amer Sports shares, profit attributable to shareholders was about RMB 7.9 billion (USD 1.2 billion), up 12.9% YoY.
The market reacted positively. On the day the results were released, Anta Sports shares closed at HKD 78.75 (USD 10), up 9.53%, lifting its market capitalization back above HKD 220 billion (USD 28.1 billion).
A breakdown by brand shows a more nuanced picture.
Revenue at the core Anta brand rose 4.8% YoY to RMB 17.8 billion (USD 2.6 billion). Operating profit increased 1.2% to RMB 4.0 billion (USD 593.8 million), while operating margin slipped to 22.5% from 23.3% a year earlier.
During the reporting period, the brand recorded growth in several key product categories. Sales of four running shoe lines, PG7, C Family, Mach, and Zone 2, exceeded 5.6 million pairs in the first half, up more than 30% YoY. In apparel, sales of products including Storm Mecha surpassed two million units, an increase of nearly 35%.
Lai said the Anta brand is becoming more focused in its product strategy, with a greater emphasis on professional sports products. At the same time, a higher mix of <a href="/industry/e-commerce">e-commerce</a> and directly operated retail weighed on margins.
In July, former Anta brand CEO Tsui Yeung left the company for family reasons. Seven store formats he had championed, including Anta Super Store, Anta Sneakerverse, and Anta Arena, are still being tested.
As of June 30, the core Anta brand, excluding Anta Kids, had 7,122 stores, down 81 from the end of 2025. Lai said at the earnings briefing that the ultimate measure of channel reform was not how many stores were opened, but whether products could reach consumers more effectively through the right channels.
The comment suggests the company is placing greater emphasis on channel effectiveness after two years of retail experimentation.
Overseas markets, meanwhile, have become an increasingly important growth area for the core brand.
Anta’s overseas revenue rose 35% YoY in the first half. According to the company, the group is advancing a plan to operate as many as 1,000 stores in Southeast Asia. In North America, Anta has also opened its first flagship store in Los Angeles and is working with retailers including Foot Locker, DSG, and JD Sports.
The overseas growth, however, comes from a relatively small base. Expanding toward 1,000 stores in Southeast Asia will also require sustained investment, leaving uncertainty around both the pace of Anta’s international expansion and its returns.
whereRelatedArticle
Fila presented a different picture. In the first half, revenue rose 6.1% YoY to RMB 15.1 billion (USD 2.2 billion). Operating profit increased 9.7% to RMB 4.3 billion (USD 638.4 million), while operating margin rose one percentage point from a year earlier to 28.7%.
Over the past two years, Fila has centered its business on the “One Fila” strategy and expanded its presence in sports including tennis and golf. The simultaneous increase in revenue and margin suggests the strategy is contributing to improved profitability.
Within Fila, apparel has again become a growth driver. Women’s apparel increased its share by 2 percentage points to more than 40%.
Fila has long been associated with premium sports fashion. Footwear, however, remains an important category for sportswear brands seeking to demonstrate product technology, and Fila has made fewer visible gains in this area.
Anta’s “other brands” segment, led by Descente and Kolon Sport, continued to record strong growth. Revenue rose 44.2% year over year to RMB 10.7 billion (USD 1.6 billion), surpassing RMB 10 billion (USD 1.5 billion) in a half-year reporting period for the first time. Operating profit reached RMB 3.5 billion (USD 519.6 million), up 43.9%.
Beyond its multibrand strategy, Anta Sports’ investment in <a href="/industry/ai">artificial intelligence</a> emerged as another focus of its results.
Lai said the group is advancing an “AI 365” strategy. Its self-developed Linglong design model can reportedly turn a sketch into a design rendering in 15 seconds, while its Lingxi styling model has been deployed in retail stores to recommend outfits for different scenarios and provide virtual try-ons.
Anta said it is also using AI in corporate management to analyze social media and consumer trends in support of market forecasting and inventory management. During the reporting period, the group’s R&amp;D spending amounted to 2.5% of revenue.
The company is expanding the use of AI across areas ranging from design to retail and customer service, although it remains too early to determine how much those tools will improve operating efficiency.
Lai said the initiative was still at an early stage and that its value would ultimately need to be measured through better products, faster responses, healthier inventory, and an improved consumer experience.
Those are also areas that will remain important as growth in the broader sportswear market becomes more difficult to sustain.
KrASIA features translated and adapted content that was originally published by 36Kr. This <a href="https://36kr.com/p/3957560779292037">article</a> was written by Xie Yunzi for 36Kr.
Note: HKD, RMB figures are converted to USD at rates of HKD 7.84 = USD 1 and RMB 6.74 = USD 1 based on estimates as of August 31, 2026, unless otherwise stated. USD conversions are presented for ease of reference and may not fully match prevailing exchange rates.

Anta’s multibrand strategy continues to deliver in H1 2026 as its AI ambitions take shape