Anthropic Publicly Files IPO Application, Listing as Early as Mid-November

Top News Briefs

1. ChangXin refuses overseas orders from Apple and others, prioritizing domestic demand for memory chips

2. Seedance core R&D head Zhou Chang reportedly resigns? A person close to ByteDance denies it

3. Li Nan on the iPhone Duo screen cracking issue: the essence is that Apple and the supply chain have grown unfamiliar with each other

4. An apology that also went wrong? XPeng manager Etta's apology video is suspected to be AI-generated

5. Audi China's four-ring logo and letter logo officially split

6. Members experienced rate limiting and unnecessary quota consumption when using K3.1 Pro; Kimi apologizes and announces compensation

7. Breaking! Anthropic publicly files for IPO: net loss of $80.7 billion in half a year, listing as early as mid-November

8. Starting price 5,499 yuan, Apple releases new iPad mini

Today's Headlines

ChangXin refuses overseas orders from Apple and others, prioritizing domestic demand for memory chips

October 26 news, according to media reports, ChangXin recently declined to conclude deals with foreign customers, aiming to achieve self-sufficiency in China's memory industry. ChangXin not only refused to supply memory to Apple, but a server manufacturer in Taiwan, China was also turned down. The reason is mainly that domestic customer orders are currently dense, and existing capacity is going all out to meet the huge demand of the local market.

Industry chain sources pointed out that Apple had previously negotiated with ChangXin Memory hoping to obtain a memory supply quota with a more favorable price than Samsung, SK Hynix, and Micron, but ChangXin Memory did not accept the price-cutting conditions. In addition, another server manufacturer in Taiwan, China that planned to purchase ChangXin Memory's DDR5 memory was also unable to reach a contract with ChangXin.

Industry insiders believe that the refusal of these large orders is because ChangXin Memory is currently meeting the needs of Chinese customers. ChangXin Memory has supplied LPDDR6 DRAM to local partners such as Xiaomi, and the bottleneck it currently encounters is meeting demand. A semiconductor industry observer said that ChangXin Memory already has a mountain to climb in meeting the needs of Chinese customers. In fact, the goal of meeting domestic DRAM demand may last ten years, but companies such as Huawei and Xiaomi will reduce their dependence on memory makers such as Samsung or SK Hynix. (SEMI)

Domestic News

Seedance core R&D head Zhou Chang reportedly resigns? A person close to ByteDance denies it

October 26 news, according to media reports, rumors recently circulated that Zhou Chang, one of the core heads of ByteDance Seed AI R&D, has submitted his resignation. Sources said Zhou Chang will start a business, possibly related to embodied intelligence, and has already been actively approached by various well-known funds, strategic investments from major companies, and leading large-model unicorns. A person close to ByteDance responded on the 26th that Zhou Chang has not resigned.

Public information shows that Zhou Chang graduated with a bachelor's degree from Fudan University's Department of Computer Science and Technology in 2012, and received a doctorate from Peking University's Computer Software and Theory program in 2017. He then joined Alibaba, where his alias was "Zhonghuang," and served as the technical lead of the Tongyi Qianwen large model, having launched a series of language models and multimodal models with his team.

In August 2024, Zhou Chang joined ByteDance to work on AI large models, and currently serves as the head of the visual multimodal generation direction for ByteDance's Doubao large model. (The Paper)

Li Nan on the iPhone Duo screen cracking issue: the essence is that Apple and the supply chain have grown unfamiliar with each other

October 26 news, after Apple's first foldable phone, the iPhone Duo, went on sale, some users reported that the new device had screen bright lines, cracking, and other issues, drawing attention. Recently, Angry Miao founder Li Nan posted an article expressing his views on the screen problems of the iPhone Duo. Li Nan believes that the many display brightness and screen cracking problems with Apple's foldable phone essentially stem from Apple and the supply chain having grown unfamiliar with each other. He said that unlike in previous years, Apple's designers and engineers lived and ate on China's supply chain.

Li Nan pointed out that industrial design may seem unrestrained, but ultimately mass production must be achieved, and supply chain capability is the gravity that a product cannot escape. In his view, Chinese phone brands have already launched multiple generations of foldable products and have accumulated rich experience regarding the risks that design adjustments may bring, the compromises that need to be made, and the limitations in craftsmanship. Li Nan believes that Apple has increasingly distanced itself from China's supply chain in recent years, and that process, quality, and structural engineers on the production line may no longer have sufficient opportunities to fully communicate with Apple's team.

He said bluntly that the foldable screen industry has developed for so many generations, and he does not believe that the screen cracking and bright line problems Apple encountered are something China's supply chain cannot solve. Li Nan further speculated that the relevant risks were likely anticipated long ago but ultimately ignored by Apple's senior management. It is understood that previously, multiple users on overseas forums and platforms such as Xiaohongshu had reported that the iPhone Duo had bright lines and cracking on the screen after unboxing. (Kuai Technology)

An apology that also went wrong? XPeng manager Etta's apology video is suspected to be AI-generated

October 26 news, recently, at the XPeng MONA L05 pre-sale launch event, post-2000s product manager Etta repeatedly called her boss "laodeng" (old geezer), saying bluntly, "How would laodeng understand what young people want?" After the launch event, terms such as "laodeng" and "dengwei" sparked controversy online.

On October 22, Etta posted an apology video on her social media account. However, some netizens noticed that the XPeng manager's apology video appeared to be AI-generated. Netizens found that when the hands were clasped together, there was still a ring. But when the hands were separated, the ring disappeared. At the same time, a notice reading "suspected use of AI generation technology" appeared on the Douyin platform. As of press time, XPeng official and Etta herself had not issued a public response. (Bianews)

Audi China's four-ring logo and letter logo officially split

On October 26, Audi signed strategic agreements with its partners respectively, aimed at strengthening the differentiated positioning of the Audi four-ring brand and the AUDI (Audi letter logo) brand and their respective strategic focuses, and further clarifying the "dual brands, dual partners" channel architecture.

Among them, FAW will comprehensively strengthen the Audi four-ring brand's market advantages in both the intelligent luxury new energy vehicle and fuel vehicle fields; SAIC Audi will focus on the AUDI (Audi letter logo) brand, concentrate on the new energy market, and launch next-generation intelligent connected vehicles that meet the needs of Chinese consumers.

SAIC Audi can still normally sell four-ring products until the end of 2027. After-sales service for SAIC Audi's four-ring brand products and the all-new AUDI (Audi letter logo) brand products will both maintain stable operations. (Yicai)

Meituan plans to raise US$1 billion through bond issuance

October 26 news, according to foreign media reports, people familiar with the matter revealed that Chinese food delivery giant Meituan plans to raise about US$1 billion (about 6.677 billion yuan) through offshore RMB bonds, with the issuance plan being Meituan offshore RMB Reg S dual-tranche senior unsecured bonds, making Meituan the latest company to join the growing ranks of dim sum bond issuers.

People familiar with the matter said that Meituan stated at an investor communication meeting on Monday that the company has obtained approval from domestic regulatory authorities and has a higher quota for overseas bond issuance, but decided to retain part of the quota for refinancing US dollar bonds when market conditions are appropriate; the quota will expire in September next year. This batch of dim sum bonds is expected to be priced on Tuesday. Sources said this is the second time in Meituan's history that it has issued dim sum bonds, and the proceeds will mainly be used to replace existing offshore RMB loans.

Offshore RMB financing continues to rise in appeal for domestic and foreign companies due to its lower cost. Earlier this month, Tencent Holdings issued 26 billion yuan of offshore RMB bonds, the first time in a single issuance that the RMB portion exceeded the US dollar tranche. Like other domestic tech peers, Meituan has also developed its own large language model and open-sourced it, hoping to join the industry's first tier. However, compared with rivals such as Alibaba and ByteDance, Meituan's AI-related product iteration speed is relatively slow. (IT Home)

DeepSeek model ranks first globally in weekly call volume

October 26 news, according to calculations based on the latest OpenRouter data, last week (October 19 to 25), the total global call volume of AI large models was 188 trillion tokens, up 14.63% month on month. Among the AI large models on the list, the weekly call volume of Chinese AI large models reached 93.54 trillion tokens, up 18.14% month on month; during the same period, the weekly call volume of US AI large models was 22.47 trillion tokens, up 29.44% month on month. China's large model weekly call volume has exceeded that of the United States for 26 consecutive weeks, firmly ranking first globally.

Observations show that last week, the top five in global call volume were all Chinese AI large models. Among them, DeepSeek V4.1-Flash-1017 ranked first, with weekly call volume reaching 33.6 trillion tokens, up 801% month on month. DeepSeek V4.1 Flash ranked second, with weekly call volume reaching 18 trillion tokens, down 45% month on month; MiniMax M3.1 Flash (free) surged onto the list, ranking third, with weekly call volume reaching 13.7 trillion tokens. (National Business Daily)

Kimi apologizes and announces compensation after members using K3.1 Pro experienced rate limiting and unnecessary quota consumption

On October 26, Kimi Moonshot posted that it apologizes to everyone for the recent rate limiting and unnecessary quota consumption experienced by Kimi members when using K3.1 Pro.

Kimi will perform a reset for subscribed members across all tiers: all members within the validity period will receive 25% of the monthly total as compensation, equivalent to one week's usage. If the current usage is less than 25% of the monthly total, it will be reset to 0. The usage progress bar will show the adjustment results before the end of the 26th. At the same time, configurations have been adjusted, and unnecessary sub-agent calls and output length are being reduced so that the same quota can complete more actual tasks. (Yicai)

Cambricon chairman refutes rumor that "chip shipments will exceed 1 million units next year"

October 26 news, at Cambricon's Q3 2026 earnings briefing, regarding market discussion that "the company's chip shipments next year will exceed 1 million units," Cambricon chairman and general manager Chen Tianshi responded that the company has never had such communication.

Chen Tianshi said that company announcements are an effective way for all investors to obtain company information on an equal basis. If major matters requiring disclosure under laws and regulations are involved, the company will fulfill its information disclosure obligations in accordance with the law, and reminded investors to refer to information disclosed on the company's official website of the Shanghai Stock Exchange or the company's official website. (Lanjing News)

DJI launches exclusive hosting service for consumer drones in Beijing

On October 26, according to the DJI Service official WeChat account, DJI has launched an exclusive hosting service for consumer drones registered to individuals in the Beijing area. The application channel is open from 14:00 on October 26 to 20:00 on November 14. Users must ship their devices by courier before 24:00 on November 14. Old models that have already had service discontinued are not covered by this hosting service because they cannot be inspected and repaired.

According to the introduction, the corresponding hosting service covers storage of drones and accessories, pickup and return shipping, battery recharging, device inspection, and routine after-sales repair. In terms of cost, if the device has completed information verification with the relevant authorities before the specified time and the physical SN code matches the registration, users can enjoy free hosting for up to 1 year, including one round-trip pickup service with shipping included. For machines that have not completed verification, a hosting fee of 199 yuan per unit will be charged for up to one year.

The service has strict restrictions on logistics and timing. DJI said that the destination for drone device pickup or return shipping is limited to non-Beijing areas of mainland China, meaning that after hosting, the drone cannot be shipped back to Beijing again. Second, once a user uses the pickup or return shipping service, the current hosting process is deemed to have "closed the loop." If the user wants to send the machine back to DJI's hosting point again after finishing their use, that is a completely new process and requires paying a renewal fee of 199 yuan per unit again. (IT Home)

Future vehicles will be required to install cellular vehicle-to-everything technology! Person familiar with the matter: relevant departments have launched mandatory work

On October 26, according to media reports, an anonymous practitioner in the C-V2X field revealed: "Relevant departments are already pushing forward the development of mandatory standards for vehicle installation of C-V2X (cellular vehicle-to-everything). A relevant working group has been established, and dozens of units have already participated."

It is understood that commercial vehicles have already been the first to be required to install C-V2X. In July this year, the Ministry of Transport publicly released a draft for comments on the relevant national standard. The document clearly states that 25 months after the standard takes effect, newly manufactured commercial passenger vehicles, trucks, and tractor-trailer combinations in China will all need to be equipped as standard with LTE-V2X direct communication terminals. This is also the first time C-V2X terminals have passed a mandatory national standard, requiring automakers to install them at the pre-installation stage.

Looking globally, C-V2X has become a direction pursued by many countries. V2X can enable information exchange between vehicles and between vehicles and road infrastructure, fill perception blind spots for vehicles, and is an important part of intelligent connected systems and active safety systems. It is understood that at present, more and more mainstream automakers such as FAW, SAIC, BYD, and NIO have launched mass-production models with C-V2X installed as standard at the factory. As the development of domestic mandatory standards accelerates, whether passenger vehicles will subsequently be included in the mandatory list will also become a focus of continued industry attention. (Kuai Technology, IT Home)

Volkswagen China obtains L3 autonomous driving road testing license in Hefei

On October 26, Volkswagen Group China officially obtained an L3 autonomous driving road testing license in Hefei and has already started real-road testing and data collection for L3 technology in Hefei. It is reported that on October 28, 2026, the first Volkswagen Group China Tech Day will be held in Hefei, showcasing the latest technological progress.

According to a report in April, Han Sanchu, executive vice president of Volkswagen Group and CEO of CARIAD China, said: Intelligence is a factor in user choice, and Volkswagen is rapidly building development and iteration capabilities around the core demands of Chinese users. Volkswagen will achieve L3 conditional autonomous driving in 2027 and L4 autonomous driving in 2029.

Horizon Robotics and Volkswagen Group announced in July this year that, through CARIZON, the joint venture between CARIAD and Horizon Robotics, Volkswagen Group will further deepen technical cooperation with Horizon Robotics in the field of AI foundation models. According to the plan, Volkswagen Group's L3 autonomous driving capability will begin delivery as early as the second half of 2027. (IT Home)

Lao Gan Ma denies "using AI to cut 600 workers to 48, with annual revenue as high as 5.4 billion yuan"

On October 26, recently, many self-media accounts posted online claiming that "Lao Gan Ma used AI to cut 600 workers to 48, while its annual revenue reached as high as 5.4 billion yuan." The related content sparked heated discussion, and some consumers even said they would no longer buy Lao Gan Ma.

On October 26, the topic "Lao Gan Ma AI layoff rumor sparks controversy" once again topped social platform trending searches. Media sought verification from Lao Gan Ma's official hotline, and the staff member answering the phone responded that the related AI layoff rumor is false information, that the company has never released such information, and asked consumers to view it rationally.

Affected by the rumor, Lao Gan Ma's recent online sales have fallen significantly. Taking a certain e-commerce platform as an example, third-party data platform shows that since October 22, Lao Gan Ma's sales for several consecutive days were 1,000-5,000 yuan, whereas before the rumor began, its daily sales were mostly stable at 5,000-10,000 yuan.

According to another report, Li Xin, executive vice president and chief quality officer of Lao Gan Ma, disclosed at the 2026 Guizhou Provincial Quality Development Conference that "through four years of quality consolidation and system optimization, in 2025 our revenue reached 5.4 billion yuan, the highest peak in history." (Lanjing News)

On October 26, it was reported that after ByteDance previously merged its formerly independent AI short drama platform Xiaoyunque into Jimeng, ByteDance's Jimeng AI announced the official launch of the Jimeng Short Drama Studio, offering the self-developed script model DreamScript, the image model Anycook, and the storyboard model DreamShot.

With the three model sets of the Jimeng Short Drama Studio, plus Jimeng's existing Seedance video generation capability, a complete AI short drama pipeline can be achieved, from script to character design and scenes, then to storyboard previews, finished videos, post-production, and localization. This means ByteDance is consolidating its AI short drama resources and no longer dispersing its efforts.

Previously, Xiaoyunque focused on short drama agents, while Jimeng focused on image and video generation, and the two systems overlapped. Now, these two systems have been merged into a single short drama studio, focusing on short dramas, a highly commercialized track. To avoid internal competition, the script, character design, storyboard, and video capabilities have been integrated into a closed-loop tool aimed at short drama companies, studios, and overseas expansion firms, not just ordinary consumer users. From another perspective, short dramas are already one of the most certain scenarios for AIGC monetization as judged by ByteDance. (Xinwan'er)

Mu Yan, co-founder of Baihe.com, "begs" on WeChat Moments: earns over 5,000 yuan a day

On October 26, it was reported that recently Mu Yan, co-founder of Baihe.com and former member of the BOSS team on "Only You," posted on WeChat Moments that his assets had been frozen without reason, leaving him in financial hardship, and he was forced to "solicit alms" on WeChat Moments to seek help.

English text is being prepared.

Mu Yan also said that an old friend from 30 years ago reconnected because of this alms-seeking, and the two arranged to meet in Beijing for tea and to catch up. In this regard, he lamented: "Originally a bad thing, thanks to the support of friendship, it gradually turned into a good thing."

Public information shows that Mu Yan graduated from the Department of Computer Science at Tsinghua University, co-founded Baihe.com in 2005, and later promoted the merger of Baihe.com and Jiayuan. He became widely known to the public for participating in programs such as "Only You," and in recent years attracted attention for publicly questioning management issues on the Douyin platform. According to media reports, in recent years Mu Yan has repeatedly had his equity frozen and been subject to court restrictions on high consumption. (Bianniu Shi)

International News

Breaking! Anthropic publicly files for IPO: net loss of 80.7 billion USD in half a year, listing as early as mid-November

On October 26 local time, large model maker Anthropic publicly submitted IPO registration documents to the U.S. Securities and Exchange Commission (SEC), planning to list on the Nasdaq Global Select Market under the ticker symbol "ANT." According to people familiar with the matter, Anthropic plans to list as early as November 16. Its valuation is expected to reach 1.5 trillion to 2 trillion USD, potentially making it the largest IPO in history.

However, investors are faced with a set of highly contrasting figures. The prospectus discloses that Anthropic's revenue in the first half of 2026 reached 16.29 billion USD, with a net loss of as high as 80.75 billion USD. Of that, about 72.4 billion USD came from changes in the fair value of financial instruments such as convertible bonds, mainly related to the company's rising valuation. Adjusted operating profit for the same period was about 1.6 billion USD.

Greater pressure comes from computing power investment. The prospectus shows that Anthropic expects to commit at least 518 billion USD in spending over the next decade on cloud services, computing power, and infrastructure, of which about 80% is non-cancelable or payable regardless of actual usage. (Yicai, APPSO)

Starting at 5,499 yuan, Apple releases new iPad mini

On October 26, Apple Inc. released the new iPad mini, equipped with an 8.5-inch OLED display and the A20 Pro chip, with a body thickness of only 5.4 mm, and support for IP67-level water and dust resistance. Apple said the new device's AI performance is up to 4.5 times that of the previous generation, graphics performance up to 2.7 times, and cellular network data transmission speed up to 50% faster.

The new iPad mini is available in Midnight and Starlight colors and in 128GB, 256GB, and 512GB capacities. Orders will be accepted starting October 28, and official sales will begin on November 6. The Wi-Fi model starts at 5,499 yuan, and the cellular model starts at 6,799 yuan; the education discount starts at 5,099 yuan.

According to another report, Apple is testing a larger foldable tablet device. The foldable tablet exposed this time is described as a "huge foldable tablet," and its design references Huawei's MateBook Fold launched in August. Previous predictions by well-known analyst Ming-Chi Kuo indicated that the foldable iPad will arrive in 2028, with a screen diagonal size of 18 to 20 inches when unfolded. Well-known tech blogger Mark Gurman described the device as "two iPad minis side by side," suggesting it adopts a book-style hinge design similar to the iPhone Duo. (Xinhua Net Finance, 21st Century Business Herald)

On October 26, it was reported that Samsung Electronics said it plans to use artificial intelligence technology to shorten the development time of new semiconductor materials by 3 years.

The company announced this goal at a seminar held at the Korea Industry Association forum, where the discussion centered on the application of artificial intelligence in industrial research and development (R&D). Choi Young-sang, head of the Samsung Advanced Institute of Technology, said at the forum: "We plan to rely on AI to compress the R&D cycle for new semiconductor materials from 5 years to 2 years, while increasing the number of promising candidate materials screened out by more than 5 times." (Sina Finance)

Luxeed is expected to have four more SUV models going overseas, and subsequent development will all follow global vehicle (standards)

English text is being prepared.

The blogger stated that, as can be seen from the Luxeed V9 and RX, the models Luxeed subsequently develops are all based on global vehicle (standards). He also revealed that the four SUVs to follow are expected to debut next year.

According to earlier reports, the global models of the Luxeed V9 and Luxeed RX were exposed earlier this month. Among them, the overseas version of the Luxeed V9 adopts a new exterior design, with the front face changed to a large vertical-slat grille, and the headlight assembly shape has also been adjusted to a vertical layout on both sides of the grille; the exterior of the Luxeed RX differs little from the domestic version, but it retains the LUXEED lettered badge design. (IT Home)

Panasonic employee improperly uploaded information on colleagues and family members to a GenAI service, affecting 334,000 people

On October 26, Panasonic Excellence Operations (PEX) recently issued a notice announcing and apologizing to affected parties for an incident of improper information upload in a human resources information system project jointly carried out with its subsidiary Panasonic Digital. According to the notice, an employee responsible for the human resources information system project violated established rules by uploading data including current employees, former employees, and employees' family members to a prohibited external generative artificial intelligence service platform.

The incident affected 75,000 current employees, 178,000 former employees, 55,000 family members of current employees, and 26,000 family members of former employees using the Panasonic Group's unified human resources information system, totaling approximately 334,000 people. The data involved included current/former employee numbers, addresses, phone numbers, email addresses, as well as the names, addresses, phone numbers, and email addresses of family members serving as emergency contacts.

After the incident, the employee who caused the accident deleted the generative AI chat records and the relevant data on the work computer. The GenAI service provider stated that these data were not included in the training database, and the data retained in accordance with compliance requirements will also be deleted on October 25. No leakage of this batch of data has occurred so far. (IT Home)