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

… expand its partnership with China in renewable energy to attract greater investment and accelerate its energy transition," he said. "We should capitalize on China's leadership in clean energy by deepening bilateral cooperation."
This article first appeared on <a href="https://asia.nikkei.com/business/energy/renewables-firm-up-china-s-status-as-bangladesh-s-top-power-investor">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>.

Renewable energy firms up China's status as Bangladesh's top power investor

Deep Tech

Energy

China

… now?</h3>
AZ: I do not think the AI-related bubble is actually very severe.
Beyond evidence such as the budgets of major companies, the China Development Forum officially disclosed in March that domestic token consumption had increased 1,800-fold in two years.
Today’s compute infrastructure is already substantial, but the rate of growth is nowhere close to 1,800-fold.
That is one reason I remain optimistic as a whole.
From a long-term macro perspective, one factor behind the internet bubble in 2000 was that fiber optic networks had been built too far ahead of demand.
Today, many people use that analogy when they describe their worry about AI. Looking back, however, much of what happened then was simply market panic.
Even if this AI wave turns out to have moved one or two years ahead of fundamentals in the short term, I do not think that would be a serious problem.
<h3>IPO Zaozhidao: How well do you personally fit with this “young” era?</h3>
AZ: I really like learning from history.
I have seen too many top investors from the previous generation in China and the US leave the stage with regret when new industrial waves arrived.
I certainly hope my own prime investing years can last longer. The opportunities in this cycle are simply too exciting.
Interestingly, after AI emerged, much of the sense of distance I used to feel because of age faded away. Interacting and communicating with other people has become much easier, and I also feel closer to younger people.
Perhaps young people today are mentally more mature. When I talk with people around 20 years old, I hardly feel any obvious generation gap.
At the root of that is the fact that access to information has become much more equal.
This <a href="https://xueqiu.com/2854806290/404610447">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 6.75 = USD 1 based on estimates as of August 18, 2026, unless otherwise stated. USD conversions are presented for ease of reference and may not fully match prevailing exchange rates.

Qiming’s Alex Zhou on staying half a step ahead of the AI market

Venture Capital

Artificial Intelligence

Investment

… originally published by 36Kr. This <a href="https://36kr.com/p/3934484753398918">article</a> was written by Deng Yongyi for 36Kr.
Note: RMB figures are converted to USD at rates of RMB 6.75 = USD 1 based on estimates as of August 18, 2026, unless otherwise stated. USD conversions are presented for ease of reference and may not fully match prevailing exchange rates.

LatentVerse takes a third path beyond VLA and world models for embodied intelligence

Robotics

Embodied Intelligence

… data-tmfx-note="1">Note: MYR, RMB figures are converted to USD at rates of MYR 4.06 = USD 1 and RMB 6.75 = USD 1 based on estimates as of August 18, 2026, unless otherwise stated. USD conversions are presented for ease of reference and may not fully match prevailing exchange rates.

Deals in brief: Vietnam’s N2TP secures funding, VentureTech invests in three Malaysian companies, 12 China investments, and more

Private Equity

Merger &amp; Acquisition

Singapore

Malaysia

Vietnam

Southeast Asia

… href="https://asia.nikkei.com/business/markets/trading-asia/china-s-national-team-battles-ai-stock-rollercoaster">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 6.75 = USD 1 based on estimates as of August 17, 2026, unless otherwise stated. USD conversions are presented for ease of reference and may not fully match prevailing exchange rates.

China's "national team" battles AI stock rollercoaster

Barely a year and a half ago, TikTok briefly went dark in the US, capping a yearslong saga that threatened to remove one of the world’s most popular social media apps from one of its largest markets.
<a href="/company/bytedance">ByteDance</a> eventually found a way through. In January, it completed a deal that transferred control of TikTok’s US data, algorithms, and app security to a majority US-owned joint venture. ByteDance retained a 19.9% stake, while continuing to oversee revenue-generating businesses, including <a href="/industry/e-commerce">e-commerce</a> and advertising, through another subsidiary.
The latest numbers from TikTok Shop offer another measure of what was commercially at stake.
TikTok Shop generated an estimated USD 11.8 billion in gross merchandise value (GMV) in the US in the first half of 2026, more than double the same period last year, according to a <a href="https://momentum.asia/insights/detail/tiktok-shop-in-the-us-h1-2026">new report</a> by Singapore-based research firm Momentum Works and e-commerce analytics provider Tabcut. The 103% year-on-year increase made the US TikTok Shop’s largest individual market, overtaking Indonesia.
That growth is increasingly visible beyond TikTok’s own ecosystem.
Transaction data from <a href="https://www.consumeredge.com/resources/insights/tiktok-shop-is-reaching-a-meaningful-share-of-u-s-retail-shoppers/">Consumer Edge</a> found that TikTok Shop accounted for around 1.5% of total US online retail spending in the fourth quarter of 2025, gaining about 50 basis points over the previous year. The research firm said the platform’s share had risen steadily since its US launch in late 2023.
The figure remains small relative to the overall US e-commerce market. But the growth is notable for a shopping platform that did not operate in the country less than three years ago.
Consumer Edge had already identified TikTok Shop’s growing footprint while a ban still appeared possible. In January 2025, it estimated that TikTok Shop represented nearly 0.5% of tagged US online spending in 2024. At the time, it said a TikTok exit could create an opportunity for competitors to recapture spending that had shifted to the platform.
That exit never happened. Instead, TikTok Shop kept expanding.
<h3>From videos to a marketplace</h3>
Momentum Works estimates TikTok Shop generated USD 50.3 billion in global GMV in the first half of 2026, up 92% year-on-year. The US accounted for roughly 23% of that total.
The research firm expects global GMV to reach USD 123.5 billion for the full year, although that figure remains a projection. Southeast Asia is still the largest contributor collectively, but the US has overtaken each of the region’s individual markets.
There are also signs that TikTok Shop’s US business is changing as it grows.
TikTok built its commerce proposition around an advantage conventional marketplaces lacked: a vast stream of content that could create buying intent before users actively searched for a product. That formula has been particularly effective in Southeast Asia, where TikTok Shop’s mix of short videos, live streaming, and commerce helped it enter a market long dominated by platforms such as Shopee and Lazada.
But the model developing in the US looks different.
In the first half of 2025, videos accounted for 50% of TikTok Shop’s attributed US GMV, compared with 36% generated through Shop, its marketplace interface, and 14% through live streaming. One year later, Shop had become the largest channel at 51.4%. Video fell to 40.4%, while live declined to 8.2%.
The shift does not necessarily mean TikTok’s content advantage is weakening. Momentum Works said videos continue to drive product discovery, while users are increasingly completing purchases through Shop after searching for and comparing products.
In other words, TikTok Shop is starting to operate more like a marketplace in its own right, rather than simply a checkout function attached to viral videos.
Live streaming shows how far the US model has diverged from TikTok Shop’s more mature Asian markets. Nearly 69% of US influencers analyzed by Momentum Works used video exclusively to promote products in the first half of the year, while only 6.5% relied solely on live streaming. In China and Southeast Asia, live commerce is more often built around professional hosts and multichannel networks.
Live streaming remains effective at the top end in the US. Eight of the ten highest-grossing influencers generated most of their TikTok Shop GMV through live sessions. But that concentration has not made live streaming the dominant shopping format across the wider market.
<h3>More sellers, more competition</h3>
Seller growth has been just as swift.
The number of TikTok Shop stores in the US nearly tripled from 475,000 in the first half of 2025 to 1.35 million a year later, according to Momentum Works. More than three-quarters recorded sales.
Growth was especially strong among larger sellers. Around 5,700 stores generated more than USD 1 million in GMV, up from 752 a year earlier. Of those, 506 exceeded USD 10 million, compared with just 36 in the first half of 2025.
The ten largest stores generated a combined USD 748 million in GMV, up 177% year-on-year. Medicube and Dr.Melaxin each exceeded USD 100 million, while merchants ranging from QVC to Crocs ranked among the ten largest sellers.
Consumer Edge’s data also points to a widening customer base. In the fourth quarter of 2025, 21% of Shein shoppers also made purchases on TikTok Shop. Cross-shopping among customers of Hollister, Victoria’s Secret, Nike, Ulta, H&amp;M, and Zara ranged 10–14%, suggesting TikTok Shop is increasingly overlapping with established retail audiences rather than operating within a separate social commerce niche.
That scale, however, is also making TikTok Shop more competitive for merchants. Momentum Works found that average prices fell in 21 of 27 product categories, while eight of the ten largest influencers by GMV were new to the ranking in the first half of 2026.
Even established sellers could grow without maintaining their position. Tarte Cosmetics increased its GMV by 73% but fell from first to fifth among stores, while Halara doubled its GMV and slipped from second to sixth.
Momentum Works described the US market as increasingly easy to enter, but harder to win.
TikTok Shop remains a relatively small part of US e-commerce, and its rapid growth does not guarantee that it can challenge more established marketplaces. While the US is now its largest individual market, further expansion may depend on whether it can keep growing as competition among sellers intensifies and it takes on entrenched players such as Amazon.

TikTok fought to stay in the US. Its shopping business shows what was at stake

On-demand Services

E-commerce

Consumer Economy

Chinese <a href="/industry/ai">artificial intelligence</a> company <a href="/company/z-ai">Z.ai</a> has released GLM-5.3, a new flagship model that it said can match or edge some leading US models in identifying software vulnerabilities while narrowing the gap with Anthropic on long-running coding tasks.
That comparison has an important limit. On CyberGym, which tests whether a model can inspect source code, find vulnerabilities, and verify them, Z.ai said GLM-5.3 scored 84.5%. That was slightly ahead of Anthropic's Mythos 5 at 83.8% and OpenAI's GPT-5.6 Sol at 83.6%. But Mythos 5 remains well ahead when the task shifts from finding a flaw to building a working exploit.
Z.ai said GLM-5.3 scored 54.4% on ExploitBench, compared with 78.0% for Mythos 5. In another timed test, GLM-5.3 completed 105 attack development tasks in two hours and 130 in six hours. Mythos 5 completed 181 and 247, respectively.
GLM-5.3 uses the same base model as its predecessor, GLM-5.2, with Z.ai attributing the improvement entirely to post-training, or the work performed after a foundation model completes its main pretraining run.
That makes GLM-5.3 less a jump to a larger foundation model and more a test of how far Z.ai can improve an existing one by training it in longer and more realistic software engineering environments.
There is also early independent evidence that the coding gains extend beyond Z.ai's own tests. FrontierSWE, which tests models on difficult software engineering projects, currently ranks GLM-5.3 second in its tested Claude Code configuration, behind Claude Fable 5. GLM-5.2 ranks fifth. GLM-5.3 recorded an average rank of 4.50 and a 78% dominance score, compared with 2.88 and 88% for Fable 5.
The result does not make GLM-5.3 a stronger model than Fable 5 across the board. Rather, it suggests Z.ai is becoming more competitive with proprietary frontier systems in software engineering, long-horizon agent tasks, and some security-related coding work.
Z.ai's own description makes that focus clear. In its launch blog, it emphasized GLM-5.3 as a model for "frontier coding with emergent cyber capabilities,” while co-founder Tang Jie described it on X as "built to code” and “ready for cyber defense.” Tang said the gains came from post-training on the same 743 billion-parameter base.
The company said GLM-5.3's training goes beyond isolated programming exercises. Its training environments cover workflows that require the model to identify a problem, consider possible solutions, make changes, check the results, and deliver finished work. Some tasks, according to Z.ai, are designed to resemble several days of work by a senior engineer and require access to computing clusters, storage systems, internal documents, and code libraries.
That focus builds on GLM-5.2, which Z.ai released in June for long-running tasks. GLM-5.2 expanded the model's context window to one million tokens and was designed to keep track of goals and engineering requirements over lengthy projects. GLM-5.3 keeps that context window and supports a maximum output of 128,000 tokens while remaining a text-focused model.
The gains are clear in Z.ai's benchmark results. Its Terminal-Bench 3.0 score rose from 4.6 for GLM-5.2 to 28.3 for GLM-5.3. DeepSWE v1.1 increased from 46.2 to 66.9, while Agents' Last Exam improved from 23.8 to 28.5. On Z.ai's own Code Bench, the company claims a 50% improvement in coding performance over GLM-5.2.
The improvement was even larger in cybersecurity. GLM-5.3's ExploitBench score more than doubled from GLM-5.2's 24.4% to 54.4%, even though the base model did not change. Z.ai acknowledged that GLM-5.3 is currently better at the earlier stages of finding vulnerabilities, including code review, discovery, and verification, than at deeper exploitation or complete offensive and defensive operations.
That may explain why the company describes its cyber capabilities as "emergent." GLM-5.3 was not trained as a dedicated cybersecurity model. Z.ai said these capabilities appeared as it expanded reinforcement learning and trained the model in longer and more varied software development environments.
Cybersecurity therefore serves as both a new selling point and a test of Z.ai's broader idea: if a model is trained to work more like an autonomous software engineer, the same skills it uses to understand, debug, and change complex code may also help it find weaknesses in that code.
Z.ai has not yet released the model weights that would allow developers to download and run GLM-5.3 independently. The company said it would delay the public release for about two weeks while carrying out more security checks and strengthening safeguards. Its most sensitive cybersecurity functions will initially be limited to selected partners and verified users under a "trusted access" program.
Investors did not respond to the launch with another immediate rally. Z.ai's Hong Kong-listed shares closed down 3.6% on August 14, the day GLM-5.3 was announced. That contrasted with the market enthusiasm surrounding GLM-5.2 earlier in the summer.
By late June, Z.ai's shares had risen more than 2,000% from its January listing as GLM-5.2 drew attention for coming closer to leading US models in coding and agent tasks.
GLM-5.3 marks progress in a narrower set of capabilities. It does not show that Z.ai has overtaken Anthropic or OpenAI across frontier AI more broadly. Instead, it suggests the gap is narrowing in areas such as coding and cybersecurity, and that meaningful gains may not always require another large pretraining run.

Z.ai says GLM-5.3 nears Anthropic in cyber defense as coding gains accelerate

East Asia

Z.ai

Foundation and Large Models

<a href="/company/jd">JD.com</a> returned to profit growth in the second quarter of 2026 even as revenue declined, helped by stronger margins in its core retail business and a sharp reduction in losses from food delivery.
Revenue fell 2.9% year-on-year to RMB 346.4 billion (USD 51.3 billion) for the three months ended June 30, about RMB 10 billion (USD 1.5 billion) less than a year earlier. Even so, net income attributable to ordinary shareholders rose to RMB 7.1 billion (USD 1.1 billion) from RMB 6.2 billion (USD 918.7 million), while non-GAAP net income increased to RMB 8.9 billion (USD 1.3 billion) from RMB 7.4 billion (USD 1.1 billion). On an adjusted basis, JD.com earned about 21% more despite generating less revenue.
The improvement was more pronounced at the operating level. JD.com recorded RMB 4.5 billion (USD 666.8 million) in operating income, reversing a RMB 0.9 billion (USD 133.4 million) loss a year earlier. Its non-GAAP net margin rose to 2.6% from 2.1%, meaning it retained more profit from each yuan of revenue.
The main drag on sales came from electronics and home appliances, JD.com’s largest product category. Revenue from the segment fell 11.8% to RMB 157.9 billion (USD 23.4 billion), roughly RMB 21 billion (USD 3.1 billion) less than a year earlier. The company attributed the decline to a difficult comparison with the same quarter last year, when China’s trade-in subsidies boosted demand, as well as higher costs that raised prices for products including smartphones and PCs.
Other parts of the business performed better. General merchandise revenue grew 5.6%, while service revenue rose 6.8%, supported by marketplace, advertising, and logistics income. The shift in revenue mix also supported profitability, as those services generally carry higher margins than JD.com’s first-party retail business.
Management said electronics and home appliance sales began to improve in June. At JD Retail, gross margin rose to 18.5%, while operating margin edged up to 4.6% from 4.5%. JD.com said this was a record operating margin for a quarter that includes the heavily promotional 618 shopping festival. Management attributed the improvement to supply chain efficiencies and a larger contribution from higher-margin marketplace and marketing services.
Lower spending on newer businesses also helped. Marketing expenses fell by about a quarter from a year earlier to RMB 20.3 billion (USD 3 billion), mainly because JD.com spent less on promotions for new initiatives.
Food delivery was another contributor to the improvement. JD Food Delivery cut its losses by more than half from a year earlier as the company reduced subsidies per order, improved delivery efficiency, and generated more revenue from commissions and advertising. The broader new businesses segment still lost RMB 9.9 billion (USD 1.5 billion) during the quarter, but that was about RMB 5 billion (USD 740.9 million) less than the RMB 14.8 billion (USD 2.2 billion) loss recorded a year earlier.
CEO Sandy Xu described the quarter as a "definitive turning point" for JD.com's profitability trajectory.
The results also exceeded market expectations. Analysts surveyed by LSEG had expected revenue of RMB 344.6 billion (USD 51.1 billion), according to Reuters, putting JD.com’s reported figure about RMB 1.8 billion (USD 266.7 million) above consensus. Separately, FactSet estimates pointed to non-GAAP net income of RMB 7.8 billion (USD 1.2 billion), meaning JD.com reported roughly RMB 1.1 billion (USD 163 million) more in adjusted profit than analysts had expected.
Management had warned after the first quarter that electronics and home appliances would remain under pressure because of a difficult comparison with last year’s trade-in-driven sales and higher product prices. The subsequent decline in the category was therefore broadly consistent with management’s earlier guidance.
Looking ahead, JD.com expects pressure on retail sales to ease. Management said JD Retail should return to revenue growth in the third quarter as the comparison base becomes less demanding, with electronics and home appliances expected to improve meaningfully in the second half. Higher product prices, however, could continue to weigh on consumer demand.
Food delivery could provide another lift to earnings if its losses continue to narrow. JD.com expects losses from the business to decline through the rest of the year as subsidies fall and efficiency improves. Advertising revenue could also benefit from a recovery in retail sales.
The company is continuing to invest heavily in technology. R&amp;D expenses rose to RMB 7.3 billion (USD 1.1 billion) in the quarter, about 38% more than a year earlier, as JD.com increased spending on <a href="/industry/ai">artificial intelligence</a> and automation across advertising, customer service, sourcing, and logistics. The higher spending adds to near-term costs, while management is positioning the investments as a way to improve efficiency over time.
Note: RMB figures are converted to USD at rates of RMB 6.75 = USD 1 based on estimates as of August 17, 2026, unless otherwise stated. USD conversions are presented for ease of reference and may not fully match prevailing exchange rates.

JD.com profit rises despite lower second-quarter revenue

JD.com

Semiconductor Manufacturing International Corporation (SMIC), China's top contract chipmaker, said the <a href="/industry/ai">artificial intelligence</a> boom is driving massive demand for peripheral AI chips, and that there is "no chance" the company will lower its prices despite weaknesses in the smartphone and automotive markets.
On August 14, co-CEO Zhao Haijun said the company's most substantial growth came from the China market, "mainly driven by robust demand for AI peripheral chips and continued strengthening of localized manufacturing."
"We see the massive demand for peripheral chips to support the AI infrastructure build-out," Zhao said. "Looking ahead to the second half of this year, the industrial momentum and spillover effects generated by AI will persist."
SMIC's AI peripheral chip segment grew by around 40% in the April-June period, he said, adding that the company's products and technologies were gaining market recognition.
"For example, if an AI server has 72 graphics processing units (GPUs), it will need 16,000 power-related chips. That illustrates just how massive the demand has become," the CEO said.
Zhao said the chipmaker did not raise prices for chips used in smartphone, automotive, and industrial applications, given that these markets have been lackluster and overall demand is declining this year.
Some customers, however, have begun building up inventories of display driver ICs and other chips used in smartphones and TVs despite weak consumer demand, he said. "People are worried that they might not get enough chip supplies next year and would face higher prices, so they are building some inventory now for next year."
"We don't see any chance of the price [of our chipmaking services] going down this year," Zhao said.
For the tightest supply segments, such as power management and optical communication chips, SMIC will continue to discuss price increases with clients, he said. "We will never be the first one to raise the price and we will not be the one that raises the most. Our prices are discussed with clients and are reasonable."
SMIC's revenue contribution from its home market reached a high of 90.2% in the April-to-June quarter, compared with 84.1% the same time last year, while sales from the US market dropped to 8.2% from last year's 12.9%.
Zhao said rising supply chain costs have been passed through to the manufacturing segment and that SMIC will "proactively address these challenges and mitigate headwinds to minimize the overall impact."
SMIC's revenue rose 36.1% on the year to USD 3 billion last quarter, with its net income surging nearly 400% on the year to USD 733.21 million. Gross margin increased to 25.3% from 20.4% a year ago. SMIC attributed the revenue growth to increased wafer shipments, rising average selling prices, and a better product portfolio. Its production utilization rate also remains at the high level of 93.7% last quarter and is expected to further grow to about 95% in the current quarter, according to the CEO.
The chipmaker said the outlook is positive for the second half of 2026 as industrial momentum and spillover effects generated by AI are expected to continue. Revenue is estimated to grow 2–4% compared to last quarter, while gross margin is expected to improve to 26–28%. The company attributed the improving profit margin to rising average selling prices and a favorable product mix. Multiple chipmakers and electronics component makers have raised their own selling prices this year.
In the first half of this year, SMIC spent USD 3.4 billion on capital expenditure to expand production capacity.
This article first appeared on <a href="https://asia.nikkei.com/business/tech/semiconductors/china-s-smic-says-ai-spillover-effects-boosting-peripheral-chip-prices">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>.

China's SMIC says AI "spillover effects" boosting peripheral chip prices

SMIC

Semiconductors