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

Yusheng Holdings, the parent company of secondhand car platform Taocheche, filed its prospectus with the Hong Kong Stock Exchange on May 6, seeking a main board listing. Citi is acting as the sole sponsor.
Founded in 2018 and known internationally as Taocars, Taocheche has built a cooperative network to address persistent issues in China’s used car market, including fragmentation, inconsistent transaction processes, and uneven customer service. Its proprietary framework, referred to as TCN, integrates online and offline operations through a data-driven system. The company said this approach standardizes transactions and supports scalable operations across the buying journey, offering customers a more consistent way to access used vehicles.
According to Frost &amp; Sullivan, Taocheche ranked first among China’s used car trading platforms by gross merchandise value in 2025, with a 3.8% market share. The platform recorded RMB 15.5 billion (USD 2.3 billion) in GMV that year. From inception through December 31, 2025, cumulative GMV reached about RMB 70 billion (USD 10.3 billion), reflecting a compound annual growth rate of roughly 50% over seven years, according to the prospectus.
Transaction volume rose steadily over the past three years, reaching 105,937 in 2023, 139,504 in 2024, and 191,487 in 2025.
On the ground, Taocheche reported operating 62 self-run sales centers across China as of December 31, 2025, which it described as the largest offline network among peers. It also said it could deliver vehicles to customers in every city nationwide through its last-mile logistics system. In parallel, the company maintained 9,294 designated display spaces across the country.
Online, Taocheche has expanded through multiple digital channels. In 2025, it reported more than 13 million monthly active users across its ecosystem. Live streaming has become a key sales driver, supported by a team of over 190 staff stationed at sales centers. The company said transactions generated through live streaming grew about 156% between 2024 and 2025.
Taocheche counts Yixin Group and <a href="/company/tencent">Tencent</a> among its key backers. Yixin provides automotive financing solutions that integrate with Taocheche’s platform, while Tencent has supported the company with cloud infrastructure and artificial intelligence tools. Other investors include Sino-Italy Ningbo Ecological Park Holding Group and <a href="/company/jd">JD.com</a>.
Financially, revenue increased from RMB 4.4 billion (USD 644.7 million) in 2023 to RMB 5.5 billion (USD 805.9 million) in 2024 and RMB 6.7 billion (USD 981.7 million) in 2025. Gross profit rose from RMB 417 million (USD 61.1 million) to RMB 588 million (USD 86.2 million) and RMB 679 million (USD 99.5 million) over the same period, with margins of 9.4%, 10.8%, and 10.2%, respectively.
Adjusted EBITDA, a non-IFRS measure, turned positive in 2024 at RMB 38 million (USD 5.6 million), before declining to RMB 9.2 million (USD 1.3 million) in 2025. The drop suggests that while the company has reached operating breakeven on an adjusted basis, profitability remains fragile.
Proceeds from the IPO will be used to expand sales centers in core domestic and overseas markets, strengthen marketing and brand capabilities, and invest in R&amp;D, particularly in data analytics and AI. The company also plans to pursue acquisitions and allocate part of the funds to working capital and general corporate purposes.
This <a href="https://xueqiu.com/2854806290/387347234">article</a> was adapted based on a feature originally written by 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 6.82 = USD 1 based on estimates as of May 7, 2026, unless otherwise stated. USD conversions are presented for ease of reference and may not fully match prevailing exchange rates.

Taocheche files for Hong Kong IPO as China’s leading used car trading platform

Hong Kong

China

Doubao, the <a href="/industry/ai">artificial intelligence</a> assistant developed by <a href="/company/bytedance">ByteDance</a>, is preparing to introduce a paid tier with additional services layered on top of its free offering, according to a service notice on its app store page cited by 36Kr.
The notice outlines three subscription tiers:
<ul>
<li aria-level="1">Standard, priced at RMB 68 (USD 10) per month, or RMB 688 (USD 100.8) per year.</li>
<li aria-level="1">Enhanced, priced at RMB 200 (USD 29.3) per month, or RMB 2,048 (USD 300.1) per year.</li>
<li aria-level="1">Professional, priced at RMB 500 (USD 73.3) per month, or RMB 5,088 (USD 745.5) per year.</li>
</ul>
These paid features have not yet been rolled out within the Doubao product.
In an official statement, Doubao said it has consistently provided free services but is exploring value-added offerings to address differing user needs. It added that the plans remain in a testing phase, and full details will be released through official channels once finalized.
A source familiar with the matter told 36Kr that the paid features are expected to focus on complex, productivity-driven tasks, including slide deck generation, data analysis, and film and video production. As the underlying model improves, the product is being positioned to handle more advanced, higher-value use cases. These tasks are likely to require greater computing resources and longer inference times.
KrASIA features translated and adapted content that was originally published by 36Kr. This <a href="https://36kr.com/p/3794799114476809">article</a> was written by Li Xiaoxia for 36Kr.
Note: RMB figures are converted to USD at rates of RMB 6.82 = USD 1 based on estimates as of May 7, 2026, unless otherwise stated. USD conversions are presented for ease of reference and may not fully match prevailing exchange rates.

ByteDance explores AI monetization for Doubao with paid subscription tiers

East Asia

Artificial Intelligence

Deep Tech

… could shape the limits of embodied intelligence
SynReal mainly consists of three parts:
<ol>
<li aria-level="1">SynReal Sim, a high-fidelity simulation engine.</li>
<li aria-level="1">SynReal Arena, an embodied intelligence training platform.</li>
<li aria-level="1">SynReal Core, a training model based on large-scale synthetic interaction data.</li>
</ol>
Put simply, SynReal Sim creates a “virtual world” governed by physical rules. SynReal Arena gives robots a “virtual training ground.” SynReal Core helps robots learn from those simulated interactions.
Together, these components allow robots to train in virtual environments, potentially running millions, or even tens of millions, of exercises per minute as they learn how to interact with the physical world.
Another common approach to embodied intelligence training is to collect data manually through human demonstrations. That method is simpler to implement from a technical standpoint. Because the data comes from the real physical world, it also captures details and noise that can help robots adapt to real environments.
Its drawbacks are clear: it is expensive, slow, labor-intensive, and often difficult to generalize.
Human-supported data collection is often measured in months or years. Gathering high-quality demonstration data for a single task can require a professional team working for several months, with costs ranging from six- to seven-figure RMB sums.
A simulation platform, by contrast, can generate millions of data trajectories covering many variations within a few hours.
Compared with other physical AI simulation platforms, SynReal emphasizes three areas: accuracy, speed, and stability:
<ol>
<li aria-level="1">In accuracy, Style3D said SynReal is supported by large reserves of rigid body and deformable 3D simulation data, allowing it to provide robots with training environments that are more realistic, more complex, and closer to the real world. The company added that SynReal can perform more accurate static and dynamic mechanics calculations, reducing errors by nearly 20% compared with the industry benchmark Isaac Sim.</li>
<li aria-level="1">In speed, simulation throughput determines whether data generation can scale. Style3D’s technical team is said to have rebuilt the 3D deformable object simulation workflow around a GPU-based parallel computing architecture. The company said this makes SynReal five to ten times faster than Nvidia’s Isaac Sim, improving the efficiency of robot training.</li>
<li aria-level="1">In stability, the challenge is contact. When two or more deformable objects touch, the difficulty of modeling their interaction rises sharply. In 3D games, deformable objects such as hair and clothing are often prone to “clipping,” where objects visually pass through each other. To address this, Style3D’s technical team introduced a path-based computational method based on incremental potential contact (IPC), an approach presented at SIGGRAPH 2020. According to the company, this enables SynReal to maintain simulation stability even when multipoint, multilayer, and self-contact interactions change frequently, giving robots more stable learning results in complex scenarios.</li>
</ol>
<figure id="attachment_161766" aria-describedby="caption-attachment-161766" style="width: 640px" class="wp-caption aligncenter"><img loading="lazy" decoding="async" class="wp-image-161766" src="https://console.0722vps.aieehr01.top/wp-content/uploads/2026/05/Style3D-Simulation.jpg" alt="" width="640" height="360" srcset="https://console.0722vps.aieehr01.top/wp-content/uploads/2026/05/Style3D-Simulation.jpg 1920w, https://console.0722vps.aieehr01.top/wp-content/uploads/2026/05/Style3D-Simulation-720x405.jpg 720w, https://console.0722vps.aieehr01.top/wp-content/uploads/2026/05/Style3D-Simulation-1024x576.jpg 1024w, https://console.0722vps.aieehr01.top/wp-content/uploads/2026/05/Style3D-Simulation-768x432.jpg 768w, https://console.0722vps.aieehr01.top/wp-content/uploads/2026/05/Style3D-Simulation-1536x864.jpg 1536w, https://console.0722vps.aieehr01.top/wp-content/uploads/2026/05/Style3D-Simulation-480x270.jpg 480w" sizes="(max-width: 640px) 100vw, 640px" /><figcaption id="caption-attachment-161766" class="wp-caption-text">Image source: Style3D.</figcaption></figure>
As labor costs rise and demand for robot training data grows, massive volumes of lower-cost, high-quality simulation data are becoming increasingly important.
Only after repeated training in virtual environments supported by systems such as SynReal are robots likely to have a better chance of reducing costs and entering real-world scenarios, whether they are folding laundry, cooking, providing eldercare, or offering companionship.
Historical turning points are often driven by crossovers.
Nvidia began with graphics cards for gaming. Xiaomi started with smartphones for enthusiasts. Many technology companies have entered new markets by applying hard-won capabilities from one field to unresolved problems in another.
These crossovers bring different industry DNA. They use deep accumulation in one domain to solve persistent pain points in another. They are not always bound by established assumptions, and they can bring new tools and perspectives to old problems.
Deformable 3D simulation is becoming an important dividing line in embodied intelligence. Style3D, with its accumulated deformable 3D model data and simulation expertise, could be one of the more credible crossovers into this field.
KrASIA features translated and adapted content that was originally published by 36Kr. This <a href="https://36kr.com/p/3788704783801607">article</a> was written by Xiao Xi for 36Kr.

Style3D’s fashion tech offers unexpected answer to a tough robotics problem

Fashion

Robotics

Creative

Embodied Intelligence

Global pharmaceutical executives are hungry for more deals with Chinese biotechnology companies after a record year, even as a stock market boom shows signs of cooling down.
2025 was a breakout year for China's biopharma industry. Out-licensing deals, in which a Chinese biotech company grants another drugmaker the rights to develop, produce, and sell its drug candidates, hit a record USD 135.7 billion, according to data provider Pharmcube. That was nearly triple the value in 2024.
"It's hard to be at this moment and not think about extending resources on the ground here," Stacy Feld, global head of external scientific innovation at Johnson &amp; Johnson, said at the ChinaBio Partnering Forum in Shanghai this week.
Matthias Muellenbeck, head of global business development and alliance management at Merck, said that on top of licensing drugs from Chinese biotech firms, the company is looking to invest in new companies formed outside of China to develop drug candidates in what is known as a "newco" model. "We want to have skin in the game," he said.
China has emerged as a go-to destination for global pharmaceutical companies hunting for new blockbuster drugs, thanks to biotech companies' rapid pace of development achieved at low cost. For Chinese biopharma, out-licensing deals help raise cash quickly and bolster their reputation by announcing tie-ups with global giants.
In January, UK-based AstraZeneca agreed to pay USD 1.2 billion upfront to CSPC Pharmaceutical Group for rights to develop a portfolio of obesity and diabetes drugs. CSPC said it is eligible for up to USD 17.3 billion in payments tied to additional R&amp;D and sales milestones. Sanofi licensed the global rights to Sino Biopharmaceutical's oral drug rovadicitinib in a deal worth up to USD 1.5 billion.
Chuan Sun, managing partner of the Shanghai and Hong Kong offices of law firm Morrison Foerster, said the growing value of the average licensing deal partly reflects a shift in the balance of power between Chinese licensors and the global licensees. "Quite a number of Chinese biotechs are probably no longer short of cash, meaning they are in a better position to negotiate," he told Nikkei Asia on the sidelines of the Shanghai conference.
Sun recalled his own experience representing a Chinese biotech company for a deal back in 2021 and again last year. "We were able to get way better terms" in last year's transaction, he said.
Investors said key trends in China's biotech sector include companies developing platform technologies for drug discovery, rather than a small number of drug candidates, and a deeper integration of <a href="/industry/ai">artificial intelligence</a>. "Previously, we [thought] AI is a very good, supportive tool," said Kevin Chen, founding partner of venture capital firm BioTrack Capital. "Today, we started to realize AI is going to play a more important role based on our interactions with the new generation of AI entrepreneurs".
In recent years, Hong Kong's stock market has emerged as a key funding channel for cash-starved biotechs. The Hong Kong stock exchange, or HKEX, introduced Chapter 18A listings in 2018 to enable pre-revenue biotech companies to be listed. As of the end of January, 84 companies have listed under Chapter 18A amid a major jump of such listings in 2025.
There are signs that investor appetite is cooling for biotech companies yet to turn a profit, however. In several joint statements since late 2025, the HKEX and Hong Kong's Securities and Futures Commission have vowed to uphold the quality of stock listings. During the period, some biotech companies' shares flopped on their trading debuts.
More recently, shares of Mabwell Bioscience, a Shanghai-based developer of treatments for cancer and age-related diseases, were priced at the bottom of their offering range ahead of their trading debut in Hong Kong on April 28. On April 29, shares dropped below the offering price.
Four biotech firms including Mabwell have listed under Chapter 18A so far this year, according to HKEX data, and three of them are currently trading below their offered prices.
"It makes sense for the sponsors to become a bit more selective, rather than rushing to get companies listed," said Sun.
"You obviously wanted to ask hard questions that they probably should have asked a few years ago," he said, "Whether or not the science and technology are genuinely differentiated compared to their competitors, whether or not there is a strong experienced management team that can execute corporate strategies, whether or not the development plan is actually credible."
This article first appeared on <a href="https://asia.nikkei.com/business/health-care/global-drugmakers-eye-more-china-biotech-deals-after-record-year">Nikkei Asia</a>. It has been republished here as part of 36Kr’s ongoing partnership with <a href="https://www.0722vps.aieehr01.top/nikkei-teams-up-with-36kr-in-asia-tech-news-coverage/">Nikkei</a>.

Global drugmakers eye more China biotech deals after record year

Biotechnology

… detection range reportedly reaches 400 meters. Dreame said it can identify small targets such as water-filled traffic barriers within 300 meters and small animals within 280 meters, and recognize traffic lights, lane markings, and small objects with precision.
In chassis control, Dreame offers a system that combines fully steer-by-wire technology with <a href="/industry/ai">artificial intelligence</a>-powered coordination. The company said this setup breaks through the limits of traditional mechanical control, allowing the rocket car to balance extreme acceleration with driving stability. In power and range, Dreame said its all-solid-state battery, paired with integrated packaging technology, addresses current energy pain points in NEVs.
<h3>The third leap for Chinese new energy vehicles</h3>
After more than a decade of rapid development, China’s NEV sector has gone through two major rounds of iteration among its new automakers.
The first generation of new automakers was represented by Tesla, <a href="/compnay/nio">Nio</a>, <a href="/company/li-auto">Li Auto</a>, and <a href="/company/xpeng">Xpeng</a>. These companies helped move NEVs from concept to scaled mass production.
The second generation is represented by the Harmony Intelligent Mobility Alliance (HIMA). Its goal is not to build cars directly or invest heavily in full-vehicle assets, but to rely on technology R&amp;D strengths and support automakers in areas such as intelligent cockpits, driving assistance, and in-vehicle ecosystems. In doing so, it has helped NEVs move from affordability toward the high end.
The debut of Dreame’s Nebula Next program and car may point to the emergence of a third generation of new automakers.
After years of development, China has formed one of the world’s most complete NEV industrial systems. Upstream core components, midstream vehicle manufacturing, and downstream ecosystem support have all reached a high degree of domestic control. These capabilities provide the basic conditions needed to challenge the global high-end segment. But entering that market also requires breakthroughs in original technology, brand building, and global operations.
Dreame is trying to carry forward the advantages of the first two generations in deep technology development, while avoiding both asset-driven competition and the limits of software-based empowerment. Drawing on years of experience in consumer electronics and smart hardware, it has built capabilities across electric drive systems, perception, algorithms, and multi-category hardware integration. It has also been developing what it describes as an integrated human-vehicle-home technology ecosystem.
From this perspective, the new car is less a sudden turn toward carmaking than a milestone in Dreame’s public display of its technological capabilities. The message is straightforward: Dreame is not starting from scratch. It has a foundation built over years, even if that expertise was not accumulated directly in the automotive sector.
KrASIA features translated and adapted content that was originally published by 36Kr. This <a href="https://36kr.com/p/3790587098832137">article</a> was written by 36Kr Auto.
Note: RMB figures are converted to USD at rates of RMB 6.79 = USD 1 based on estimates as of June 5, 2026, unless otherwise stated. USD conversions are presented for ease of reference and may not fully match prevailing exchange rates.

Making sense of Dreame’s rocket-boosted car and automotive ambitions

Dreame

Transportation

Electric Vehicles

Mobility

… comparable to commercial cinemas.”
AI is also reshaping the form and experience of projection products. The Xiaolong series is already equipped with an intelligent color calibration system, which replaces the traditional model of engineers manually calibrating color coordinates. The system uses AI to learn from samples and automatically complete standardized and personalized color optimization.
“In the past, color calibration depended on engineers manually fine-tuning frame by frame. Each person had a different visual standard, so calibration results varied. Now, AI can independently learn from massive samples. The more data it accumulates, the more precise color calibration becomes.”
Sonnoc has also included robotic visual interaction in its future plans. Its Japanese company is developing robotic visual capture capabilities. Cameras can capture images, but Zhu said robots still lack image output capabilities. “We want to be the robot’s ‘second eye,’ filling the image output gap and completing the entire visual interaction loop.”
Sonnoc’s long-term R&amp;D goal is medium-free 3D spatial imaging, which would require no screen, water curtain, scrim, or other carrier. Instead, it would present three-dimensional images using optical technology alone. Zhu said that if the technology is realized, it could reshape the display industry.
Zhu has designated 2026 as Sonnoc’s “first year overseas,” with an annual overseas sales target of RMB 50 million (USD 7.3 million). Its priority markets are Southeast Asia, Central Asia along the Belt and Road Initiative, and South Korea.
According to Zhu, South Korea has almost no domestic projector manufacturers and has previously relied on Japanese and Taiwanese suppliers. Now that Japanese products have lost some of their price advantage, he said, Sonnoc is more competitive than Taiwanese brands in both technology and cost. Combined with the accelerating domestic substitution of core components, Zhu believes the timing for overseas expansion is now mature.
Having watched single-color lasers rapidly replace traditional bulb light sources, Zhu has a clear forecast for the adoption pace of triple-color laser projection. Over the next three years, he expects the penetration rate of triple-color laser products to exceed 50–60% in high-end scenarios such as premium home studios, cultural heritage and museum exhibitions, and cultural tourism lighting. He estimates this corresponds to a high-end professional projection market of about RMB 3–4 billion (USD 438.6–584.8 million).
“As the domestic supply chain matures and drives costs down, market demand for high color gamut and high brightness will continue to rise. Triple-color laser replacement is already an irreversible industry trend.”
On the funding front, Sonnoc plans to launch a new funding round this year. It will seek additional investment from state-owned capital while bringing in high-quality, market-oriented industrial capital to continue supporting technology R&amp;D and inventory investment in core components.
KrASIA features translated and adapted content that was originally published by 36Kr. This <a href="https://36kr.com/p/3785172264197384">article</a> was written by Li Xiaoxia for 36Kr.
Note: RMB figures are converted to USD at rates of RMB 6.84 = USD 1 based on estimates as of May 6, 2026, unless otherwise stated. USD conversions are presented for ease of reference and may not fully match prevailing exchange rates.

Triple-color laser opens a new window for Sonnoc’s projector ambitions

Hardware

… href="https://36kr.com/p/3740907027054600">36Kr</a></li>
</ul>
whereRelatedArticle
<h3>Cube, PvX Partners, Pudu Robotics, and others made recent headlines:</h3>
<ul>
<li aria-level="1">Cube, an <a href="/industry/e-commerce">e-commerce</a> data intelligence startup, raised USD 3.7 million in a Series A round led by Betatron Venture Group, with participation from M Venture Partners and angel investors.</li>
<li aria-level="1">PvX Partners, a financial services platform focused on cohort financing for consumer applications, raised USD 10.5 million in a Series A round led by T-Accelerate Capital, with participation from Z Venture Capital and Drive by DraftKings. Existing investors, including Play Ventures and General Catalyst, also joined the round.</li>
<li aria-level="1"><a href="https://www.pudurobotics.com/en">Pudu Robotics</a>, an embodied intelligence company, raised nearly USD 150 million in a new round, lifting its valuation above USD 1.5 billion. Investor details were not disclosed.</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@0722vps.aieehr01.top">hello@0722vps.aieehr01.top</a>.
Note: RMB figures are converted to USD at rates of RMB 6.84 = USD 1 based on estimates as of May 5, 2026, unless otherwise stated. USD conversions are presented for ease of reference and may not fully match prevailing exchange rates.

Deals in brief: Vertex Ventures backs ReN3 and Formas.ai; Featherless and SiamDL raise Series A funding; latest China investments, and more

Venture Capital

Private Equity

Investment

Merger &amp; Acquisition

Singapore

Thailand

Southeast Asia

… robot vacuums, wet-dry vacuums, vacuum cleaners, hair dryers, air conditioners, refrigerators, and washing machines, Dreame is also developing products still in the prototype stage, including card-shaped recorders and AI pendants. The number of categories is extensive.
The approach resembles applying internet-style experimentation to the hardware industry: test broadly, combine features, learn from mistakes, and iterate quickly. “<a href="/company/bytedance">ByteDance</a> built an app factory. In the <a href="/industry/robotis">robotics</a> era, I will build 1,000 products,” Yu once said in an interview.
whereRelatedArticle
<h3>Expansion through reusable capabilities</h3>
The question is what capability allows a company to expand on this scale.
Although Dreame operates across many categories, its expansion playbook has stayed largely consistent. Dreame started with hair dryers, adding features such as faster motor speeds and AI-powered functions. That approach reflects a broader pattern: improve a product category around one functional advantage, then scale from there.
After the company shifted toward overseas markets and a premium strategy, this product innovation cycle accelerated.
Jiemian News previously reported that Dreame has structured itself around innovation. At the level of individual product units, it cross-combines 250 product leads with 250 sales leads, creating more than 60,000 cells in the matrix. Dreame does not expect every cell to succeed. But when one works, it quickly commits resources, pushes execution, and aims for commercial success.
The technical foundation of this strategy is a large set of reusable underlying capabilities.
One example is the robotic arm. At the launch event, several products were fitted with robotic arms. Robot vacuums and lawn mowers had bionic dual robotic arms. The Z1 robot, designed for laundry care, is equipped with a multi-jointed robotic arm and a multimodal perception system, enabling it to autonomously open washing machine doors and identify and pick up clothing.
Tracks follow the same principle. Dreame has added tracks not only to robotic lawn mowers, but also to robot vacuums. After identifying user demand for robot vacuums that could enter rooms and move up and down stairs, the company adapted the feature for the category.
Seen through this playbook, the dozens of products on display at Dreame’s event looked less like a scattered product lineup than a set of test cases. The robot vacuum is the proven model. Phones and smart rings are new categories under validation. New categories such as automotives are more extreme experiments meant to test the ceiling of the company’s technology.
In product development, hardware is usually considered more time-consuming than software because it requires tooling, supply chains, and production costs. Dreame, however, appears to take a different view. When entering a new category, it uses relatively little capital for early validation, ranging from five-figure RMB sums to larger amounts. Yu put it this way:
<blockquote>“Building cars is not as expensive as people imagine. What is expensive is building the wrong car. Phones are the same. Making a phone is not what costs money. Making the wrong phone is.”</blockquote>
Silicon Valley in spring 2026 may offer another way to understand Dreame.
During the same week as Dreame’s launch event, Silicon Valley layoffs in the first four months of 2026 exceeded 73,000. Anthropic’s CEO also predicted that software engineering would eventually be disrupted by AI. Together, those signals point to a broader concern: some existing business models may struggle to survive a shift in technological paradigms.
No one knows what the industry will look like 18 months from now. In that sense, Dreame’s aggression and Silicon Valley’s anxiety mirror each other. Both are using one extreme to hedge against another. In an era of slow overall growth, the risk of pushing to the limit may be smaller than the risk of standing still.
KrASIA features translated and adapted content that was originally published by 36Kr. This <a href="https://36kr.com/p/3789057332043016">article</a> was written by Deng Yongyi for 36Kr.
Note: RMB figures are converted to USD at rates of RMB 6.84 = USD 1 based on estimates as of May 5, 2026, unless otherwise stated. USD conversions are presented for ease of reference and may not fully match prevailing exchange rates.

Dreame’s next act stretches from robot vacuums to rocket-boosted cars

Consumer Economy

… investment opportunities. V4’s success has shown the industry that large models can also reach leading performance on domestic Chinese computing infrastructure. This is expected to create clear demand for domestic Chinese computing resources, driving investment across the industry chain, from chip design and servers to cloud services.
Our judgment is that this year’s domestic Chinese computing infrastructure is comparable to last year’s overseas computing infrastructure. Its industry trend and reflection in capital markets will be especially strong.
We will concentrate resources on projects that can commercialize quickly, enter industries, and form product moats, while maintaining long-term investment in underlying architecture and computing infrastructure.
KrASIA features translated and adapted content that was originally published by 36Kr. This <a href="https://36kr.com/p/3788151000751364">article</a> was written by Zhou Xinyu and Wang Yuchan for 36Kr.

How do industry professionals in China view DeepSeek V4’s strengths and limits?

Foundation and Large Models