Alibaba

Baidu

ByteDance

Meituan

Pinduoduo

Tencent

Most Followed

KrASIA Connection

China's Auto Sector

Earnings Reports

Inside China's Startups

Venture Roundup

Our Events

View All

Collections

News

Features

Insights

Ecosystem

Press Releases

Anker

DeepSeek

Li Auto

MiniMax

Xiaomi

Xpeng

On the Radar

Events

Latest

About

Consumer

Robotics

Fintech

… fashion.
Instashopping reflects a shift in consumer behavior. At an industry event in October 2024, Wang Puzhong, head of local commerce at Meituan, said quick commerce is no longer just for emergencies. Instead, it signals a psychological shift. “It represents high certainty in life, whereby there’s no planning, no stockpiling, with everything delivered on demand,” Wang said.
<figure id="attachment_147626" aria-describedby="caption-attachment-147626" style="width: 1080px" class="wp-caption alignnone"><img loading="lazy" decoding="async" class="wp-image-147626 size-full" src="https://console.kr-asia.com/wp-content/uploads/2024/02/Meituan-Wang-Puzhong-World-Artificial-Intelligence-Conference-2023.jpg" alt="Photo of Wang Puzhong, head of Meituan’s local commerce business." width="1080" height="720" srcset="https://console.kr-asia.com/wp-content/uploads/2024/02/Meituan-Wang-Puzhong-World-Artificial-Intelligence-Conference-2023.jpg 1080w, https://console.kr-asia.com/wp-content/uploads/2024/02/Meituan-Wang-Puzhong-World-Artificial-Intelligence-Conference-2023-300x200.jpg 300w, https://console.kr-asia.com/wp-content/uploads/2024/02/Meituan-Wang-Puzhong-World-Artificial-Intelligence-Conference-2023-1024x683.jpg 1024w, https://console.kr-asia.com/wp-content/uploads/2024/02/Meituan-Wang-Puzhong-World-Artificial-Intelligence-Conference-2023-768x512.jpg 768w" sizes="(max-width: 1080px) 100vw, 1080px" /><figcaption id="caption-attachment-147626" class="wp-caption-text">Photo of Wang Puzhong, head of Meituan’s local commerce business. Photo from KrASIA’s archive.</figcaption></figure>
In April, the local commerce head reiterated on social media that the goal is to replace traditional e-commerce logistics with a “30-minute-to-home” model. Instashopping is expected to push deeper into categories like appliances and fashion, long dominated by JD and Tmall.
That growth requires spending. Investment banks estimate that Q2 order volume will rise 27% year-on-year, but the platform is expected to post a loss due to promotions for the 618 shopping festival. On May 27, Meituan confirmed its participation. A 36Kr price check found a 500 ml bottle of Wuliangye (8th edition, 52%) priced at RMB 850–860 (USD 119–120.4), RMB 100–150 (USD 14–21) cheaper than on Tmall or JD, and lower than Sam’s Club.
Between food delivery and Instashopping, Meituan’s local commerce segment is bracing for slower revenue growth and a sharp drop in profit. But the ultimate stance is clear: Meituan intends to win.
whereRelatedArticle
<h3>Ambitions abroad, and a bet on AI</h3>
From Hong Kong to Saudi Arabia, <a href="https://kr-asia.com/meituans-keeta-launches-in-saudi-arabia-igniting-competition-and-price-hikes">Keeta’s expansion</a> has fueled optimism around Meituan’s international ambitions, though not without cost.
In Q1, new initiatives revenue rose 19.2% YoY to RMB 22.2 billion (USD 3.1 billion), led by platforms like Kuailv, Youxuan, Xiaoxiang, and international operations. Still, quarterly losses widened from RMB 2.2 billion (USD 308 million) to RMB 2.3 billion (USD 322 million), driven by global expansion spending.
Hong Kong, Keeta’s first overseas market, was a clear early success. Within a year, it overtook Deliveroo in market share. Deliveroo cited intense competition in its 2024 annual report and exited in March. Keeta grew by paying riders more, charging merchants less, and heavily subsidizing consumers. “A year after Keeta entered, it handled 85% of our takeout orders, while Deliveroo was in the single digits,” one restaurant told local media.
But growth through subsidies comes with friction. In May, riders protested new pay structures and changes to order dispatch. Public goodwill, it turns out, is as important as operational efficiency.
In Saudi Arabia, Keeta now operates in nine cities, covering all major urban centers. A staff member in Riyadh said it has become the third largest player and is pushing for the top spot. Its strategy of using deep discounts to pressure incumbents remains intact.
On May 22, Meituan also launched Keemart in Riyadh. The warehouse-style retail business offers groceries and household essentials via the Keeta app, now serving the Al Yasmin and Granada districts. Full city coverage is planned by year’s end.
The latest international move: a USD 1 billion investment in Brazil over five years. It marks Meituan’s formal entry into Latin America, where it will compete with iFood and Didi. Local commerce head Wang has pushed back on views that these markets are small. He sees Saudi Arabia and Brazil as future growth engines, and says Meituan will proceed cautiously, evaluating each move.
Morgan Stanley expects Keeta’s losses could reach RMB 6–7 billion (USD 840–980 million) this year.
AI is Meituan’s other long-term bet. On the earnings call, the company outlined a three-pronged strategy: building large language models (LLMs), deploying AI tools internally, and integrating AI into customer-facing products.
A key launch is coming in June: Kangaroo Advice, an AI-driven business assistant to help merchants with location scouting, cuisine selection, menu design, and operations.
Internally, Meituan has already rolled out NoCode, an AI programming platform. The company claimed that 52% of Meituan’s new code is now AI-generated, and some teams see over 90% adoption. “Our goal is 100% adoption among engineers,” he said.
Meituan has fought and won many competitive battles, from group-buying to food delivery. It has also weathered doubt when previous bets, like Youxuan, failed to meet expectations. Now, it finds itself fighting on familiar ground while reaching for new frontiers. The challenge: staying focused on the current fight without losing sight of the future.
KrASIA Connection features translated and adapted content that was originally published by 36Kr. This <a href="https://36kr.com/p/3314475857471239">article</a> was written by Ren Cairu for 36Kr.

As rivals pile on, Meituan braces for a brutal fight in delivery and commerce

Xiaoxiang Supermarket

Keeta

East Asia

China

On-demand Services

E-commerce

Logistics

Food Delivery

Grocery Delivery

Consumer Economy

… high-end models add LiDAR (light detection and ranging) into the mix.
<figure id="attachment_154262" aria-describedby="caption-attachment-154262" style="width: 1920px" class="wp-caption alignnone"><img loading="lazy" decoding="async" class="size-full wp-image-154262" src="https://console.kr-asia.com/wp-content/uploads/2025/06/Eufy-E15-E18.jpg" alt="Image of Eufy’s E15 (left) and E18 robotic lawn mowers." width="1920" height="1080" srcset="https://console.kr-asia.com/wp-content/uploads/2025/06/Eufy-E15-E18.jpg 1920w, https://console.kr-asia.com/wp-content/uploads/2025/06/Eufy-E15-E18-720x405.jpg 720w, https://console.kr-asia.com/wp-content/uploads/2025/06/Eufy-E15-E18-1024x576.jpg 1024w, https://console.kr-asia.com/wp-content/uploads/2025/06/Eufy-E15-E18-768x432.jpg 768w, https://console.kr-asia.com/wp-content/uploads/2025/06/Eufy-E15-E18-1536x864.jpg 1536w, https://console.kr-asia.com/wp-content/uploads/2025/06/Eufy-E15-E18-480x270.jpg 480w" sizes="(max-width: 1920px) 100vw, 1920px" /><figcaption id="caption-attachment-154262" class="wp-caption-text">Image of Eufy’s E15 (left) and E18 robotic lawn mowers. Image source: Eufy.</figcaption></figure>
Eufy’s design leans on vision-only obstacle avoidance, a lower-cost but technically demanding route. These systems must operate reliably in dynamic outdoor settings where sunlight, shadows, animals, and garden furniture can create unpredictable challenges.
“Pure vision-based mowers are quite complex,” a robotics investor told 36Kr. “One company’s model once mistook a hedgehog for grass and ran it over, causing public outrage. Lawns are messy environments. They might have animals, droppings, patio furniture. Training for such edge cases takes time and a lot of trial and error.”
Hardware is just one part of the problem.
Anker aims to sell its mowers in both North America and Europe, but lawn care norms vary widely. European lawns tend to be fine-textured and low-maintenance. North American grass is tougher, grows faster, and often requires cutting multiple times per month—conditions that typically favor gas-powered machines.
As of now, boundary-free robotic mowers account for just 2% of the US market. Traditional mowers still dominate.
A more immediate hurdle is distribution.
“Robotic mowers are rugged, high-maintenance machines,” another investor said. “They need servicing and repairs. Installation is tricky, and users often need in-person training. Everyone wants to sell them online, but the reality is these are offline products.”
That creates tension with Anker’s strengths. The company generates around 71.2% of its revenue through online sales, according to its latest annual report. That imbalance could prove problematic as it tries to scale a product that demands local servicing and retail presence.
Still, Anker is unlikely to walk away. As a listed company, it needs to demonstrate new growth avenues, and few categories are expanding faster than consumer robotics.
KrASIA Connection features translated and adapted content that was originally published by 36Kr. This <a href="https://36kr.com/p/3302574385748741">article</a> was written by Leslie Zhang for 36Kr.

Eufy and EufyMake mark Anker’s second act in lawn mowers and UV printing

Hardware

<a href="/company/xiaomi">Xiaomi</a> has made a key hire to deepen its push into assisted driving. Multiple sources confirmed to 36Kr that Chen Guang, formerly CTO at FAW’s Nanjing research institute, has joined Xiaomi’s <a href="/industry/ev">electric vehicle</a> division to lead the development of perception capabilities for assisted driving. He now reports to Ye Hangjun, Xiaomi’s head of assisted driving. Chen replaces Rui Cai, who has since moved to Xiaomi’s <a href="/industry/robotics">robotics</a> unit.
Before joining Xiaomi, Chen was CTO at FAW’s Nanjing research institute, a subsidiary of the state-owned FAW Group. There, he oversaw the technical architecture and development of the company’s autonomous driving business and led the creation of FAW Hongqi’s third-generation Level 4 robotaxi. In recent months, FAW’s research team in Nanjing has reportedly disbanded.
Chen holds a PhD in electrical and computer engineering from the University of Missouri and has extensive experience in artificial intelligence and computer vision. He has published several papers at top-tier conferences such as the Conference on Computer Vision and Pattern Recognition (CVPR). Before FAW, he spent years at Baidu Apollo’s R&amp;D center, where he led the development of perception systems.
According to 36Kr, Xiaomi’s assisted driving division is now split into two main groups: one focused on end-to-end (E2E) functionality and another on exploratory research. Chen’s role centers on accelerating the implementation of E2E driving features.
Unlike traditional rule-based models, E2E systems integrate perception, prediction, planning, and control into a unified deep learning model. To support this structure, carmakers typically divide the work between dedicated teams.
In February, Xiaomi rolled out its full-scenario E2E assisted driving system to all users. This version was trained on three million video clips. The company plans to launch a more advanced model trained on ten million clips.
On the research side, Xiaomi is exploring vision-language-action (VLA) models, a rising frontier in autonomous tech. This effort is led by Long Chen, a former principal scientist at UK-based autonomous driving startup Wayve, who was personally recruited by Xiaomi founder Lei Jun. Insiders told 36Kr that Xiaomi began working on VLA tech relatively early and is roughly on par with Li Auto’s progress.
As a newcomer to EV manufacturing that only launched its first model about a year ago, Xiaomi faces a steeper climb than rivals and must invest heavily to meet industry benchmarks in assisted driving.
Still, it has steadily built up its bench of high-level technical talent. Alongside recent hires like <a href="https://kr-asia.com/xiaomi-adds-former-tusimple-cto-to-its-ranks-deepens-intelligent-driving-initiative">Wang Naiyan, former CTO of TuSimple</a>, and Long Chen from Wayve, the addition of Chen adds further depth to Xiaomi’s autonomous driving ranks.
As of now, Xiaomi’s assisted driving team has grown to around 1,200 people, according to 36Kr.
Yet hiring talent is just one piece of the puzzle. Engineering for mass production and ensuring system stability remains a time-intensive process that can’t be fast-tracked.
Li Auto’s experience highlights this challenge. Since 2023, the company has repeatedly <a href="https://kr-asia.com/li-auto-leans-into-end-to-end-autonomy-with-smart-driving-overhaul">changed its strategic direction and restructured departments</a>, betting heavily on both E2E models and VLA before reaching its current level of progress. CEO Li Xiang once said that “skipping straight to the tenth bun is impossible,” a metaphor emphasizing the need to progress step by step, from rule-based algorithms to E2E systems, before advancing to VLA.
Xiaomi is facing similar growing pains.
Three years into its automotive venture, the company began pushing a rule-based navigate-on-autopilot (NOA) system to users in June 2024. But as the industry evolves, these systems are proving inadequate for complex urban scenarios. With limited performance ceilings, automakers are increasingly following Tesla’s lead into E2E territory.
In September 2024, Xiaomi formally established its E2E division. After six months of development, the company launched its first version this February. Despite the milestone, user feedback suggests Xiaomi still trails behind top industry players in real-world performance.
Sources told 36Kr that Xiaomi is now racing to develop its next-generation system, based on VLA models, with a possible release by year’s end.
In essence, Xiaomi is trying to compress a multiyear development curve into a much shorter timeline, aiming to leapfrog the competition the way Li Auto has in recent quarters.
But it's doing so under intense public scrutiny. In March, a Xiaomi SU7 equipped with entry-level assisted driving was involved in a <a href="https://kr-asia.com/three-dead-in-su7-crash-fatal-accident-renews-scrutiny-of-smart-driving-safety">fatal highway accident</a> in Tongling, Anhui, resulting in three deaths. The tragedy triggered a wave of backlash and thrust Xiaomi into a crisis of confidence over its autonomous driving capabilities.
Founder Lei later described it as the most difficult period since founding Xiaomi.
whereRelatedArticle
Now, both Xiaomi and its assisted driving division are seeking a decisive comeback.
That burden partly falls on the company’s second vehicle, <a href="https://kr-asia.com/after-su7-setback-xiaomi-reboots-ev-push-with-safer-smarter-yu7">the Xiaomi YU7</a>. At its debut event on May 22, Lei sought to reassure users by highlighting that all YU7 models come standard with full-stack assisted driving hardware.
Lei sought to reassure users by highlighting that all YU7 models come standard with full-stack assisted driving hardware.
Each unit is equipped with Nvidia’s four-nanometer Thor in-vehicle chip, which delivers 700 TOPS (tera operations per second) of computing power, along with one LiDAR sensor, one 4D millimeter-wave radar, 11 high-definition cameras, and 12 ultrasonic sensors.
Still, with flagship hardware and a heavyweight technical team in place, the real question is whether Xiaomi can finally demonstrate that its assisted driving program is ready to lead the pack.
KrASIA Connection features translated and adapted content that was originally published by 36Kr. This <a href="https://36kr.com/p/3310623111847431">article</a> was written by Li Anqi for 36Kr.

Xiaomi taps FAW veteran to steer its next phase in assisted driving

Transportation

Autonomous Vehicles

Electric Vehicles

Mobility

As home automation reshapes how consumers manage their outdoor spaces, Aiper is anchoring its global operations in Singapore. Best known for its cordless robotic pool cleaners, the company has relocated its operational headquarters to the city-state as it ramps up international expansion and moves into smart yard technology.
From its new base, Aiper will coordinate global operations using a model that combines Singapore’s regulatory stability and infrastructure with scalable operations in neighboring markets. The company already manages a regional ecosystem in the Asia Pacific, covering supply chain, branding, and customer service.
The shift coincides with a significant funding milestone. Aiper has raised USD 100 million from Fluidra, a global pool and wellness equipment manufacturer. Announced at a signing ceremony in Singapore, the investment is Aiper’s largest to date and, according to the company, the biggest single deal in the robotic pool cleaner segment.
<figure id="attachment_154281" aria-describedby="caption-attachment-154281" style="width: 480px" class="wp-caption aligncenter"><img loading="lazy" decoding="async" class="wp-image-154281" src="https://console.kr-asia.com/wp-content/uploads/2025/06/Aiper-Fluidra.jpg" alt="Photo of Aiper and Fluidra signing their investment agreement during a ceremony on May 29 in Singapore." width="480" height="360" srcset="https://console.kr-asia.com/wp-content/uploads/2025/06/Aiper-Fluidra.jpg 1280w, https://console.kr-asia.com/wp-content/uploads/2025/06/Aiper-Fluidra-720x540.jpg 720w, https://console.kr-asia.com/wp-content/uploads/2025/06/Aiper-Fluidra-1024x768.jpg 1024w, https://console.kr-asia.com/wp-content/uploads/2025/06/Aiper-Fluidra-768x576.jpg 768w" sizes="(max-width: 480px) 100vw, 480px" /><figcaption id="caption-attachment-154281" class="wp-caption-text">Photo of Aiper and Fluidra signing their investment agreement during a ceremony on May 29 in Singapore. Photo and header image source: Aiper.</figcaption></figure>
While Fluidra provides capital and international reach, Aiper stressed that it will continue to operate independently. The two companies plan to collaborate on technology, branding, and customer service, with the funding supporting the development of a broader smart backyard ecosystem. This includes pool cleaners, smart irrigation systems, and connected yard tools aimed at simplifying outdoor maintenance.
Aiper positions itself around automation and ease of use. Its messaging targets homeowners seeking efficient, low-maintenance products for pool and lawn care. The company reports that its devices are now available in more than 7,000 retail stores across 45 countries, with around two million users globally. Aiper’s design-led products have also received industry recognition through awards from 2023–2025.
At the signing event, founder and CEO Richard Wang described the partnership as a collaborative milestone in Aiper’s expansion. “We will continue to drive progress in product innovation, brand influence, and global service capabilities,” he said. “Our goal is to solve real-world cleaning challenges and empower customers to enjoy a truly smart, worry-free, and effortless outdoor living experience.”
The relocation of strategic functions to Singapore reflects more than a geographic shift. It appears intended to improve flexibility, reduce supply chain risks, and enable faster execution in the region. In a category where margins are tight and friction is high, operational agility is just as critical as product quality.
Aiper’s future growth will depend on its ability to convert its engineering and design strength into lasting customer relationships, especially as it enters high-potential but underpenetrated markets. But with funding secured, infrastructure in place, and a clear focus, Aiper seems ready to evolve beyond poolside convenience and rethink the possibilities of automated outdoor living at a global level.

Aiper relocates headquarters to Singapore as it expands from pools to smart yards

Singapore

Southeast Asia

… skeptics, and that’s fine by me.
<h3>AN: Lastly, recommend three books you love.</h3>
XY: First, “Ludwig Wittgenstein: The Duty of Genius.” That book helped me get through an existential crisis, and in some ways, probably saved me.
Second, “Deng Xiaoping and the Transformation of China” by Ezra Vogel. I think Deng was the greatest entrepreneur and CEO of modern history, bar none.
The third will probably be “Consciousness-Only: A New Science of Consciousness” by Yang Dingyi.
<a href="/collections/100-ai-creators">100 AI Creators</a> is a collaborative project between AI Now! and KrASIA, highlighting trailblazers in AI. Know an AI talent we should feature? <a href="mailto:ainow@kr-asia.com">Reach out to us.</a>

Think AI can’t tell your fortune? This startup tapping Eastern metaphysics says otherwise

Artificial Intelligence

Entertainment

Deep Tech

Creative

… quarters of 2024. Total tourist spending (excluding gambling) rose 8% to MOP 56.21 billion (USD 6.9 billion). Yet per-capita spending dropped 17% to MOP 2,168 (USD 268), with Chinese mainland tourists spending 21.7% less on average.
“The golden age of luxury may never return,” said Li Jiaxiang, a retail property investor. “Forget about 2013 or 2014 levels. We’ll be lucky if we get back to where we were in 2018 or 2019 in the next four or five years.”
This isn’t just a story of consumers trading down. The deeper truth may be that European luxury brands never really cracked the cultural code in China.
A decade ago, GQ ran a video series documenting China’s luxury boom. In one early episode, a French yacht broker stood along the lit-up Bund in Shanghai and said, with a sincerity hard to judge: “We need to teach the Chinese how to enjoy themselves.”
That era coincided with China’s transition from an emerging to a mature luxury market. The country’s newly wealthy elite, shaped by economic liberalization and armed with global aspirations, sought to project taste through high-end goods. “Taste used to signify class. Now it signifies brand,” someone once remarked.
But that era is over. LVMH’s most recent quarterly report showed a 2% drop in revenue and a 5% fall in fashion and leather goods sales, missing expectations. Sales in China remain weak. Japan also showed a YoY decline, an alarming sign that the brands may have finally lost their edge in exploiting regional price gaps.
Curiously, Chinese consumers haven’t stopped spending. They are just spending elsewhere. The runaway success of homegrown brands like Laopu Gold has sent ripples through the industry. After decades of trying to shape—or tame—the Chinese taste for luxury, the West’s grip may finally be slipping.
KrASIA Connection features translated and adapted content that was originally published by 36Kr. This <a href="https://36kr.com/p/3259314973831431">article</a> was written by He Zhexin for 36Kr.

As China’s luxury market cools, brands turn to food and wine to stir demand

Retail

Washington was "clearly wrong" to assume China cannot make <a href="/industry/ai">artificial intelligence</a> chips, <a href="/company/nvidia">Nvidia</a> CEO Jensen Huang said Wednesday, adding it is critical for US chips to be exported to the Chinese market and have Nvidia chips used to run open-source models like <a href="/company/deepseek">DeepSeek</a> and <a href="/company/alibaba">Alibaba's</a> Qwen.
"The US has based its policy on the assumption that China cannot make AI chips. That assumption was always questionable, and now it's clearly wrong," said Huang during a first-quarter earnings call.
Huang spoke after the Financial Times reported the Trump administration has told US chip design tool makers to stop selling to China.
The CEO said whoever leads the AI market in China is "positioned to lead globally," but that USD 50 billion market has been "effectively closed" to Nvidia and other US chip companies because of export controls.
"China's AI moves on, with or without US chips… The question is not whether China will have AI, it already does. The question is whether one of the world's largest markets will run on American platforms," Huang added.
Nvidia launched the H20 chip for the Chinese market after the previous Biden administration tightened restrictions on AI chip exports to China early last year. However, the US government started requiring Nvidia to obtain licenses for exporting the H20 on April 9.
It said Washington's latest chip export controls have led to an additional USD 4.5 billion charge in the three months ended in April, and it expected "meaningful decrease" in China data center revenue for the ongoing May-July quarter.
Nvidia sold USD 4.6 billion in H20 chips in China during the quarter before the new export licensing requirements, according to an earnings release for the first quarter.
China's leading internet companies like <a href="/company/bytedance">ByteDance</a>, Alibaba, and Tencent <a title="" href="https://asia.nikkei.com/Business/Tech/Semiconductors/ByteDance-Alibaba-and-Tencent-stockpile-billions-worth-of-Nvidia-chips" target="_blank" rel="noopener">stockpiled billions of dollars in Nvidia H20 chips</a> this year ahead of the April ban amid mounting geopolitical uncertainties.
"The new export controls on H20 did not provide a grace period to allow us to sell through our inventory," said Nvidia CFO Colette Kress at the earnings call, adding the company was unable to ship USD 2.5 billion of H20 chips during the quarter.
While the company managed to soften the blow by reusing certain materials, Washington's latest H20 curb proved costly.
"We are still evaluating our limited options to supply data center compute products [in China] compliant with the US government's revised export control rules," Kress said.
Nikkei Asia previously reported the US AI chip leader is preparing <a title="" href="https://asia.nikkei.com/Business/Technology/Nvidia-to-ship-new-AI-chip-for-China-next-quarter-modified-Blackwell-later" target="_blank" rel="noopener">new chips for the China market</a> including a Blackwell chip that will come with GDDR7 memory, not HBM, to comply with export restrictions.
"The key is to understand the limits and see what we can come up with, interesting products that could continue to serve the Chinese market," Huang said.
Meanwhile, the emergence of cost-effective, Chinese-made DeepSeek AI has raised questions around the billions of dollars poured into data centers and other AI computing infrastructure that drove sales.
Huang said the rise of reasoning AI models like DeepSeek have led to more computing needs, as they require longer thinking and more token generation, which will drive up demand for Nvidia chips.
The Nvidia CEO called DeepSeek and Alibaba's Qwen model "among the best open-source AI models released freely," which have gained traction across the US, Europe, and elsewhere.
"DeepSeek also underscores the strategic value of open source AI," Huang said. "US platforms must remain the preferred platform for open-source AI, that means supporting collaboration with top developers globally, including in China."
"America wins when models like DeepSeek and Qwen run best on American infrastructure," he added.
For the three months ended April 27, Nvidia reported a 69% year-on-year revenue jump to USD 44.1 billion. Its main revenue driver, data centers, jumped 73% on the year to USD 39.1 billion.
For the ongoing quarter, company-wide revenue is expected to be USD 45 billion, plus or minus 2%, which reflects a loss of around USD 8 billion in H20 revenue.
This article first appeared on <a href="https://asia.nikkei.com/Business/Tech/Semiconductors/US-wins-if-DeepSeek-runs-on-American-AI-chips-Nvidia-CEO-says">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 wins if DeepSeek runs on American AI chips, Nvidia CEO says

Nvidia

Semiconductors

<a href="https://lumosrobotics.com/">Lumos Robotics</a> has completed a second follow-on round of angel funding, with backing from Damon Technology, Fosun RZ Capital, and Wuzhong Financial Holdings. The new capital will support product refinement, deeper integration with industrial partners, and the advancement of its core solutions toward commercialization.
This marks Lumos Robotics’ second funding round in just two months and its third within six months. The company has now raised nearly RMB 200 million (USD 28 million) cumulatively through angel-stage investments. Earlier backers include Innoangel Fund, SenseCapital, Plum Ventures, Puhua Capital, and Gao Bingqiang.
Founded in 2024, Lumos Robotics focuses on embodied intelligence for household applications. Its business spans R&amp;D, sales of <a href="/industry/robotics">humanoid robots</a>, <a href="/industry/ai">artificial intelligence</a> algorithms, and software systems. Core product lines include the LUS and MOS humanoid robot series, as well as components such as robotic joint modules and visuotactile sensor units.
The company was founded by Yu Chao, a Tsinghua University graduate with nearly a decade of experience in embodied robotics R&amp;D and commercialization. He previously led the humanoid robotics division at Dreame Technology, where he assembled the teams behind high-profile projects like Xiaomi’s CyberDog and Dreame’s quadruped robot. Yu is also credited with creating the world’s first backflipping embodied robot powered solely by electric drive.
Lumos’s core team includes alumni from institutions such as Tsinghua University, Shanghai Jiao Tong University, and the Chinese University of Hong Kong. Several founding members previously worked at Dreame’s MagicLab and are recognized experts in large embodied models. Collectively, the team brings a decade of experience in embodied cognition, motion control, and component design, with capabilities spanning perception, decision-making, and execution, alongside a track record in scaling consumer-grade robotics for commercialization.
Earlier this year, Lumos Robotics unveiled its full-scale humanoid robot, LUS, built entirely with in-house components. The model has entered pre-production and is expected to begin volume shipments later this year.
On the hardware side, the company has developed key components in-house, including high-performance integrated joints, precision encoders, and visuotactile sensors. These components are designed to ensure high performance and cost efficiency as the company prepares for mass production. Multiple robots optimized for industrial scenarios are also in development.
At the software level, Lumos has built an “embodied brain” system based on a proprietary, differentiable end-to-end architecture for large embodied models. The system integrates multimodal data (including vision and haptics) with task-specific algorithms to interpret and execute complex activities.
Most recently, Lumos launched its first visuotactile hardware product line: the LUX series. These modules use end-to-end imitation learning to tightly couple haptic sensing, visual recognition, and motion control. Among them, the LUX-G gripper is reportedly nearly as sensitive as a human fingertip, capable of detecting minute changes in force and torque.
Commenting on future plans for its large embodied models, Yu emphasized the importance of tactile sensing in enabling dexterous manipulation. Lumos has already integrated tactile sensing into its vision-language-action (VLA) framework and is preparing to launch its first VTLA (VLA plus tactile) model. This foundational system is aimed at both robotics and multimodal applications. The company's proprietary visuotactile sensors provide the essential data infrastructure, with potential applications extending to manufacturing, healthcare, consumer electronics, and scientific research.
According to 36Kr, Lumos Robotics has formed partnerships with Damon Technology and COSCO Shipping. These collaborations aim to deploy embodied AI in logistics and smart manufacturing, co-develop key components, explore new use cases, and link robotic innovation with the broader industrial supply chain.
Looking ahead, Lumos plans to expand its product portfolio through ongoing technical innovation, ensuring alignment with diverse industrial needs. The company also aims to deepen collaborations across sectors such as logistics, healthcare, industrial automation, and services, accelerating the integration of robotics into real-world industry.
Commenting on the investment, Xing Lizhe, partner at Fosun RZ Capital, said that the next phase of embodied artificial intelligence hinges on real-world deployment. “Lumos Robotics owns the entire stack ranging from the VLA framework to physical hardware. This gives it strong potential for commercialization,” Xing said.
“Humanoid robots are poised for real-world applications, and logistics is one of the most promising sectors. Lumos has deep experience in this field and a clear alignment with logistics use cases,” said Zhuo Xu, chairman of Damon Technology. “We believe the company is on track for rapid growth and will support its efforts in rolling out its technology across industrial scenarios.”
KrASIA Connection features translated and adapted content that was originally published by 36Kr. This <a href="https://36kr.com/p/3262489896665216">article</a> was written by Huang Nan for 36Kr.

Touch-enabled robots inch closer to reality as Lumos Robotics secures new funding

<a href="https://en.deepthink.ai/">Deepthink’s</a> advanced imaging engine, designed to process complex lighting environments using <a href="/industry/ai">artificial intelligence</a>, has made its commercial debut through the launch of <a href="/company/gac">GAC’s</a> Hyptec HL model. The company’s technology is now on track to be integrated into more than a dozen models under the GAC brand.
Founded in 2017, Deepthink has built its technical moat around an AI-powered image signal processing (ISP) system. This system allows standard light sensors to deliver full-color, high-definition, real-time imaging even in low-light conditions.
The company now operates across three product lines: drone payloads, handheld imaging devices, and imaging modules. Its first product reached the market in July 2022. Since then, Deepthink has reported over 200% year-on-year revenue growth for three consecutive years.
Previously, the company focused on night vision products tailored for industrial drones, including payload cameras, handheld devices, and scopes. But its entry into the automotive sector has emerged as its most significant growth opportunity. Deepthink is already collaborating with several of China’s leading carmakers.
<figure id="attachment_154218" aria-describedby="caption-attachment-154218" style="width: 480px" class="wp-caption aligncenter"><img loading="lazy" decoding="async" class="wp-image-154218" src="https://console.kr-asia.com/wp-content/uploads/2025/06/Deepthink-S8.jpg" alt="Image of Deepthink’s S8 night vision gimbal camera." width="480" height="480" srcset="https://console.kr-asia.com/wp-content/uploads/2025/06/Deepthink-S8.jpg 1080w, https://console.kr-asia.com/wp-content/uploads/2025/06/Deepthink-S8-720x720.jpg 720w, https://console.kr-asia.com/wp-content/uploads/2025/06/Deepthink-S8-1024x1024.jpg 1024w, https://console.kr-asia.com/wp-content/uploads/2025/06/Deepthink-S8-360x360.jpg 360w, https://console.kr-asia.com/wp-content/uploads/2025/06/Deepthink-S8-768x768.jpg 768w, https://console.kr-asia.com/wp-content/uploads/2025/06/Deepthink-S8-600x600.jpg 600w, https://console.kr-asia.com/wp-content/uploads/2025/06/Deepthink-S8-300x300.jpg 300w" sizes="(max-width: 480px) 100vw, 480px" /><figcaption id="caption-attachment-154218" class="wp-caption-text">Image of Deepthink’s S8 night vision gimbal camera. Image and header image source: Deepthink.</figcaption></figure>
“2025 will be the Cambrian explosion of smart driving. We’re going to see an eruption of vehicle-focused AI technologies,” said Zhang Qining, founder of Deepthink, in an interview with 36Kr. “Smart vehicles are poised to become the biggest platform for AI algorithms. The entry barrier is sky-high, but once you're in, growth potential outpaces nearly every other field.”
Following the launch with GAC, Zhang said other automakers have reached out to explore partnerships.
He also emphasized that perception, decision-making, and control are the three core components of any intelligent system. With decision-making and control systems reaching maturity, perception has become the primary bottleneck. AI-powered imaging engines capable of operating in varied lighting conditions are expected to gain traction quickly. These engines will be tightly integrated with domain controllers and evolve alongside driving algorithms via real-time over-the-air updates.
“Autonomous driving systems may have reached Level 3, but the onboard cameras are still stuck at Level 1,” Zhang said. As the system’s “eyes,” vehicle cameras are still priced under RMB 200 (USD 28), and their performance lags far behind the human eye, particularly in low light or poor weather. This limitation makes smart driving systems vulnerable during nighttime operations.
Zhang noted that traditional ISP relies on hand-crafted rules, which often struggle in complex lighting scenarios. In contrast, the new generation of AI-based computational imaging mimics human visual neural networks and improves continuously through training to better support intelligent driving algorithms.
Although the entry barrier to the ISP field is relatively low, Zhang emphasized that executing it well is another matter entirely. Deepthink is currently the only company worldwide to achieve commercial, automotive-grade performance using an end-to-end neural AI imaging engine.
Technically, Zhang said the underlying neural architecture is still evolving, and current algorithms aren’t yet mature enough for chip-level integration. Prematurely locking in the technology could limit its future development.
“Thanks to the rapid growth of our automotive business, we expect to become the first AI imaging company to turn a profit in the ISP space this year,” Zhang said.
KrASIA Connection features translated and adapted content that was originally published by 36Kr. This <a href="https://36kr.com/p/3312985596650246">article</a> was written by Ou Xue for 36Kr.

Deepthink’s AI imaging tech makes automotive debut with GAC, eyes broader rollout