Digital Intelligence Weekly: Jensen Huang Resigns as NVIDIA CEO, Alibaba Plans to Spin Off PingTouGe for Hong Kong Listing, Anthropic Launches IPO Roadshow
On November 12, NVIDIA announced that Jensen Huang has resigned from the positions of President, CEO and Director; CFO Colette Kress will serve as interim CEO, and the board has initiated a formal search. During intraday trading, its market capitalization briefly fell below $4 trillion. Alibaba plans to spin off its chip subsidiary PingTouGe for a separate listing on the Main Board of the Hong Kong Stock Exchange via H shares. As of April 2026, cumulative shipments of its Zhenwu series AI chips reached 560,000 units. Anthropic has launched its IPO roadshow, with an offering price of $70-78, corresponding to a valuation of approximately $1 trillion, a notable decline from the $1.5 trillion at the time of filing.
Body · AI translation
[The Digital Intelligence Weekly will consolidate the most important frontier trends, major policies, and industry research reports in the enterprise services, cloud computing, and big data fields this week.]
Viewpoints
Panel prices are brewing an increase; TCL CSOT's Zhao Jun says the driving force comes from costs rather than demand
TCL CSOT's Zhao Jun predicts shipment area growth next year, but the driving force for price increases does not come from demand—it is cost-driven.
Goldman Sachs CEO: AI infrastructure financing demand remains strong, and the overall market environment is still favorable for an Anthropic IPO
Goldman Sachs CEO David Solomon said on Monday that although recent market volatility has led investors to reassess the risks hidden behind large-scale computing capacity expansion, financing demand in the AI infrastructure sector remains strong. Solomon said that in recent weeks Goldman Sachs has been "very active" in AI-related capital markets transactions, and he expects this trend to continue. He said that factors such as changes in the political environment will bring "repricing and market fluctuations, and we are seeing such volatility right now." "But I have always firmly believed that AI has strong potential and can, in the long run, drive productivity growth in the U.S. economy."
Domestic News
SMIC's Zhao Haijun: The growth rate of Chinese customer demand is higher than the company's overall demand growth rate
SMIC Co-CEO Zhao Haijun said at the third-quarter earnings call that, by region, China and the United States accounted for 90.8% and 7.8% of third-quarter sales revenue, respectively; in terms of amount, China sales revenue grew 7% quarter over quarter, mainly because the growth rate of Chinese customer demand was higher than the company's overall demand growth rate, and the company accordingly allocated capacity to meet customer demand. Wafer revenue by size showed stable proportions for 12-inch and 8-inch wafers, at 79% and 21%, respectively; in terms of amount, 12-inch and 8-inch wafer revenue grew 9% and 4% quarter over quarter, respectively, mainly because customer demand was strong and the company's capacity expansion was mainly in 12-inch production lines.
Jiangfeng Electronics files with the Hong Kong Stock Exchange, aiming for "A+H"
On November 14, according to disclosures by the Hong Kong Stock Exchange, only Ningbo Jiangfeng Electronics submitted a prospectus to the Hong Kong Stock Exchange this week. Jiangfeng Electronics was already listed on the ChiNext in 2017 and focuses on ultra-high-purity metal sputtering targets and semiconductor precision components. By 2025 revenue, it ranks second globally and first in China. In the first nine months of 2026, revenue was 4.465 billion yuan, up 35.67% year on year; net profit attributable to the parent was 811 million yuan, up 102.38% year on year.
Alibaba plans to spin off T-Head for an independent Hong Kong listing, with annualized revenue in the tens of billions of yuan
On the evening of November 10, Alibaba announced that its chip subsidiary T-Head has applied for an independent listing. T-Head submitted a listing application to the Hong Kong Stock Exchange on September 11, 2026, and plans to list independently on the Main Board through H shares; after listing, it will remain an Alibaba subsidiary. T-Head was founded in 2018 and integrated DAMO Academy's in-house chip business and C-SKY Microsystems. As of April 2026, cumulative shipments of the Zhenwu series AI chips reached 560,000 units, and more than 60% of T-Head chips in Alibaba Cloud scenarios serve external commercial customers. Institutions value T-Head at between $25 billion and $86 billion.
DeepSeek releases full-suite application API
On November 13, DeepSeek officially opened its full-suite application API. By enabling the corresponding parameters in the existing API, users can obtain conversational features such as deep-thinking search and file input; voice and related underlying capabilities have also been opened as independent interfaces (including OCR, ASR, TTS, etc.), supporting separate or combined calls. Developers can easily implement complex conversational features in their own applications, or even replicate the complete DeepSeek App 1:1.
Tencent Hunyuan releases new real-time voice model Hy Realtime
According to the introduction, the model performs excellently in basic conversational capabilities such as voice understanding, instruction following, and anthropomorphic expression, while adding Agentic tool calling and singing capabilities. Hy Realtime has already launched in products such as Yuanbao voice calls and Tencent customer service, and will soon launch on Tencent Cloud to provide external API services.
Baidu Famou 3.0 released, already covering more than 60% of China's major industrial categories
Currently, Baidu Famou's industrial customers already cover 25 of China's 41 major industrial categories, with coverage exceeding 60%. Famou 3.0 targets complex tasks such as production operations, business strategy, and industrial R&D, connects to enterprises' real business data, and continuously optimizes around final business metrics, enabling AI to move from completing tasks to continuously creating business value.
Zhang Yiming: ByteDance AI may catch up with OpenAI and Anthropic as early as 2029 or 2030
According to multiple informed sources, ByteDance founder Zhang Yiming recently discussed the AI business again in an internal sharing session. It is understood that at a recent all-hands meeting for the AI team, Zhang Yiming said that ByteDance is capable of competing with OpenAI and Anthropic, but truly catching up with frontier technology will "at the earliest be 2029 or 2030." ByteDance's large models should benchmark against the international first tier. Zhang Yiming acknowledged that compared with top overseas AI companies, the pre-training dataset scale of ByteDance's Seed series models is still smaller, which is the issue ByteDance AI most needs to focus on at present.
Volcano Engine releases Seedance 2.5 flash: generation speed doubled, lower price
On November 12, Volcano Engine officially launched the video generation model Seedance 2.5 flash. The company calls it the model whose results are currently closest to Seedance 2.5, with generation speed doubled and a lower price. According to official data, taking text-to-video and image-to-video at 720P resolution as examples, Seedance 2.5 flash costs about 0.99 yuan per second to generate, roughly 34% lower than Seedance 2.5. The model supports video generation of up to 30 seconds per clip, accepts up to 10 images, 5 videos and 5 audio clips per input, and offers 480P and 720P resolution options.
MiniMax H3-Image tops both the AA text-to-image and image-editing open-weight leaderboards
On November 12, the leaderboard of global independent AI evaluation platform Artificial Analysis (AA) showed that MiniMax's newly released image generation model H3-Image ranks first among open-weight models in both the Text to Image and Image Editing categories.
Ant Lingguang connects to six major connectors including WeChat ClawBot and QQ Bot
On November 12, the Lingguang open ecosystem expanded again, adding support for six major connectors including WeChat ClawBot, QQ Bot, Luckin Coffee, Youdao Cloud Notes, TickTick and Tencent News. This expansion extends Lingguang's services to high-frequency scenarios such as instant messaging, knowledge management and task collaboration, bringing the number of external connection services to 47. According to reports, after this ecosystem expansion, Lingguang has connected some WeChat and QQ functions through the WeChat ClawBot and QQ Bot connectors. Users can send tasks and receive replies in WeChat private chats, and can also query message records and sending status. The QQ Bot connector supports processing private-chat and group-chat messages, proactively sending notifications, and scheduled message pushes after confirmation.
Doubao Work officially integrates more models including Kimi 3.1 Pro
On November 12, Doubao Work officially integrated multiple third-party models including Kimi 3.1 Pro, GLM 5.3 and DeepSeek V4.1 Flash, and launched a limited-time promotional plan. Users can choose the appropriate model to complete tasks according to their actual work needs. In addition, Doubao Work has launched limited-time discounts for Doubao 2.1 Lite and Deepseek V4.1 Flash, lowering the quota consumption multipliers for these two models.
Tokens become an independent category on Tmall Double 11 for the first time, with major model vendors offering 50% off
On November 11, the Tmall Double 11 Token recharge event opened, with major model vendors including Zhipu, MiniMax, Qianwen Office, Alibaba Cloud and Qoder participating in 50%-off promotions. Specifically: all three MiniMax monthly cards are 50% off; two Alibaba Cloud Token Plan monthly cards are up to 50% off; Zhipu weekly cards are up to 50% off; the first month of the Qianwen Office basic membership monthly card is up to 70% off; the Qoder monthly card is 67% off, as low as 19.9 yuan. Notably, this year's Double 11 is the first time Tokens have appeared on Tmall as an independent category.
Tencent responds that "50% layoffs" is a misreading; management proposes "half army, half navy"
On November 12, a person close to Tencent responded that the claim of "Tencent laying off 50%" is completely untrue. During Tencent's 28th anniversary Linktime online livestream, management said this organizational change differs from the "cost reduction and efficiency improvement" of a few years ago; it is a proactive move for the AI era and leans more toward "efficiency improvement." The internal formulation is: 50% of people are like an army with responsibility for defending territory, doing core business well through efficiency gains; the other 50% are like a navy embracing the wave, exploring AI Native and expanding new space. Tencent PR Director Zhang Jun also posted on Weibo that it is "purely a misunderstanding and rumor."
ByteDance releases Doubao image model Seedream 5.5
On November 9, ByteDance released the Doubao image creation model Seedream 5.5, which is being integrated simultaneously into Doubao, Doubao Work and Jimeng, with the API to go live on Volcano Ark this week. The new model strengthens the stability of multi-round editing; in actual tests, after a hundred consecutive rounds of modification, non-edited areas can remain pixel-level consistent, and it natively supports RGBA transparent material generation and layered output. In terms of cost and performance, editing a single 2K square image costs about 0.28 yuan, less than half that of Seedream 5.0 Pro, and text Tokens are not billed. Official evaluations show that the new model improves usability rates in six types of scenarios including e-commerce and film and television, reaching more than double the previous generation in film and television and office scenarios, with simultaneous optimization in small-text rendering and realistic lighting performance.
Alibaba Cloud: Token Plan quotas raised by more than 70%
On November 9, Alibaba Cloud raised the subscription quotas for the personal edition of Token Plan. The monthly fees for the Lite, Essential and Standard tiers remain unchanged, while monthly quotas increase by about 74% to 78%. Specifically, the Lite plan's quota rises from 11,500 Credits to 20,000 Credits at a monthly fee of 39 yuan; the Essential plan's quota rises from 25,500 to 45,000 Credits at a monthly fee of 79 yuan; the Standard plan's quota rises from 45,000 to 80,000 Credits at a monthly fee of 139 yuan.
On November 9, based on the latest OpenRouter data, total global AI large-model calls last week (November 2 to 8) were 210 trillion tokens, up 8.25% week-on-week. Among the AI large models on the list, Chinese AI large models recorded 116.18 trillion tokens in weekly calls, up 16.71% week-on-week; in the same period, US AI large models recorded 13.98 trillion tokens weekly, down 32.4% week-on-week. Chinese large models' weekly call volume has exceeded that of the United States for 28 consecutive weeks, firmly ranking first globally. Last week, among the top five AI large models by global call volume, four were Chinese large models. Among them, DeepSeek V4.1 Flash 1017 ranked first for the third consecutive week, with weekly calls reaching 56 trillion tokens, up 17% week-on-week.
On the afternoon of November 11, the "Wave Makers" book series launch event and the PKU Entrepreneurship Lecture, jointly hosted by China Machine Press and the Peking University Entrepreneurship Training Camp, was successfully held in the tiered classroom of the Peking University Global Student Innovation and Entrepreneurship Center. At the event, "Wave Makers: A Practical Breakdown of AI Entrepreneurship" was officially released. This also marks the official launch of the "Wave Makers" book series to the whole of society.
International News
Jensen Huang resigns as Nvidia CEO, president and director; CFO Colette Kress becomes interim CEO
On local time November 12, Nvidia announced that Jensen Huang has resigned from the positions of president, chief executive officer and board member, effective immediately. Current executive vice president and chief financial officer Colette Kress has agreed to serve as interim CEO, and the board has launched a formal CEO search that will consider internal and external candidates. It was previously reported that Nvidia's board had approached Arm CEO Rene Haas about succession, and Nvidia responded that it had not "formally approached" any candidate. Affected by this, Nvidia's intraday market capitalization briefly fell below $4 trillion.
Nebius: AI demand is "real and enormous"; expects pre-tax break-even in Q4 2027
AI cloud infrastructure company Nebius Group held a conference call. CEO Arkady Volozh responded at the meeting: "Demand for AI is real and enormous. Everything else, such as building capacity and financing it, is just a task." He reiterated that the company expects revenue of $12 billion to $13 billion in 2027 and pre-tax break-even in the fourth quarter of 2027.
Anthropic launches IPO roadshow, offering price $70-78, valuation nearly halved from peak
On US Eastern Time November 10, Anthropic launched the IPO roadshow for its Class A common stock, with an offering price range of $70 to $78 per share, planning to issue 825 million shares, and applying to list on Nasdaq under the ticker "ANT." At the upper limit, it would raise up to about $64.4 billion. This IPO corresponds to a valuation of about $1 trillion, significantly reduced from the $1.5 trillion when it filed half a month earlier, and nearly halved from the earlier $2 trillion target. The company plans to complete pricing next week and aims to list before Thanksgiving on November 26.
Google releases Gemini 4 Hydrogen, focused on everyday coding and agentic workflows
On November 12, Google released Gemini 4 Hydrogen, positioned as a "fast execution engine" for everyday coding and agentic workflows. API pricing is $1.5 per million input tokens and $7.5 per million output tokens, covering a 1 million token context. On the Artificial Analysis intelligence index, the default high-thinking setting scores 51, exceeding Gemini 4 Argon's 50; Terminal-Bench 4.1 score rises from 18.2% for Gemini 3.8 Flash to 57%, and task completion speed increases 2.5 times. It simultaneously launched the Orchestrator intelligent scheduling feature, which can dynamically switch between Carbon and Hydrogen, reducing average quota consumption by 24% and increasing speed by 19%. Free users can also access Hydrogen capabilities through the Auto mode in the Gemini app.
Microsoft apologizes for 10-hour global service outage affecting Teams, Outlook and Copilot
On local time November 13, Microsoft apologized for a global service outage. The failure left Teams, Outlook and Copilot unusable in at least 70 countries for about ten hours. Microsoft said the failure was "not related to a cyberattack or AI/agents" and was triggered by an earlier "network capacity upgrade," in which multiple servers unexpectedly went offline and put pressure on the remaining infrastructure. The UK NHS, the US federal government and court systems all reported being affected.
SpaceX cuts about 60 employees from Grok Imagine video team, leaving only a few staff to maintain the product
On November 13, according to reports, SpaceX's artificial intelligence team last week laid off about 60 employees responsible for the Grok Imagine image and video generation tool, leaving only a few staff to maintain the product's operations. The tool is still open to users, but due to the lack of dedicated team support, its future development prospects are uncertain. People familiar with the matter said that Musk's xAI is currently shifting resources toward AI coding tools and the Grok Bot agent product.
AI chip company Cerebras surges nearly 20% in US premarket trading
On the news front, Cerebras' core revenue in the third quarter was $230 million, up 70% year over year and above the market expectation of $215 million; among this, cloud services revenue grew 272% year over year to $145 million. As of September 30, remaining performance obligations (RPO) reached $26.1 billion. The company raised its full-year 2026 core revenue guidance from $880 million to $890 million, up to $940 million to $960 million. The company also disclosed that it has contracted 700 megawatts of data center capacity, planned for delivery by the end of 2027.
AMD: Fifth-generation EPYC processors to provide host infrastructure for Google TPU 8tx
On November 9, AMD announced that its fifth-generation EPYC server processors will provide host infrastructure for Google Cloud TPU 8tx. TPU 8tx is the first cloud TPU system to support hardware-enforced confidential computing. The two parties will combine AMD EPYC processors, Infinity Guard security technology and Google TPU compute power to help enterprises run AI workloads such as inference, fine-tuning and reinforcement learning in the cloud while protecting sensitive data. AMD said its EPYC processors have become the host CPU choice for more than 10 major hyperscale cloud service providers and frontier AI labs worldwide.
Google launches two new TPU chips, stepping up competition with Nvidia
On November 9, Google will launch two new tensor processing unit (TPU) chips, TPU 8tx and 8i Max, to meet the needs of different AI applications. Among them, TPU 8tx targets confidential computing needs of government agencies and regulated industries, and works in concert with AMD chips; TPU 8i Max mainly targets AI applications requiring long context windows, boosting compute power and memory capacity by packaging two TPU chips together. Amin Vahdat, head of Google's chips, said the two new products are expected to launch next year. Industry research estimates that sales of Google TPU-related chips, software and hardware will exceed $135 billion by 2028, up from an estimated $12 billion this year.
Meta signs long-term procurement agreements with suppliers to promote low-carbon manufacturing in the data center supply chain
Meta announced that it has signed long-term procurement agreements with multiple suppliers covering low-carbon cement, near-zero-emission steel, electric truck transportation and recycled plastics, in order to promote the adoption of low-carbon manufacturing technologies by industries related to the data center supply chain. Among them, Meta signed a multi-year agreement with Sublime Systems to support the production of at least 180,000 tons of low-carbon cement; signed a multi-year agreement with Swedish steel company Stegra to support near-zero-emission primary steel production; and reached a long-term procurement agreement with MacroCycle Technologies to support its construction of a commercial plant in the southeastern United States, with a target of producing 5,000 tons of high-value recycled PET per year. Meta said these multi-year commitments are intended to provide suppliers with revenue certainty and promote the commercialization of low-carbon technologies at scale.
CoreWeave reportedly seeks to raise at least $4 billion through investment-grade loans
On November 11, according to Bloomberg citing people familiar with the matter, Morgan Stanley and Mitsubishi UFJ Financial Group (MUFG) are leading the potential financing of at least $4 billion, planning to syndicate it to investors in stages, possibly launching as soon as next week, but the plan has not been finalized. CoreWeave released its earnings on Tuesday, raising its full-year revenue guidance to $13 billion to $13.3 billion, above the previous expected range and also above the Wall Street average estimate of $12.9 billion. The company's stock price rose sharply on Wednesday, and its bond prices also climbed.
US AI buildout hits power bottleneck
The United States has seen a boom in AI infrastructure construction over the past two years, but power supply has replaced chips as a new bottleneck constraining the industry's development. Industry companies are taking various measures to quickly supply power to data centers, but the resulting increase in greenhouse gas emissions and other issues cannot be ignored. Industry data shows that there are already thousands of data centers in operation in the United States, but most existing data centers are small in scale, while data center projects recently promoted or being planned by hyperscale cloud service companies often require power supply capacity exceeding 1 gigawatt or even 10 gigawatts, which the existing local power grids cannot handle in a short time.
More than 40% of planned data center capacity in the US faces policy risks
Amazon Web Services announces three major capability upgrades in China
On November 11, Amazon Web Services announced that, building on its four core advantages of being globally leading, trusted, industry-savvy and localized, it will further upgrade its technical capabilities, security capabilities and localized support in China. This upgrade will further improve its full-stack Agentic AI technical capabilities, strengthen AI security protection capabilities, and enhance localized support through multi-level technical collaboration with customers, development methodology enablement and talent training.
Financing
Piotech's share issuance to purchase assets approved by the Shanghai Stock Exchange
On November 12, Piotech announced that it plans to purchase 82.97% of shares in Wuxi Shangji Semiconductor, 100% equity in Shanghai Taina Micro and 100% equity in Wuxi Kuanxing through the issuance of shares and cash payment, and to raise supporting funds. The Shanghai Stock Exchange has approved the transaction.
Lightelligence plans to launch A-share IPO coaching
On November 11, Hong Kong-listed company Lightelligence-P announced that its board of directors has approved signing a coaching agreement with Guotai Haitong Securities, planning to initiate work related to the domestic issuance of RMB ordinary shares and submit a filing application for A-share listing coaching.
Xixuan Chuangzhi completes 500 million yuan angel round financing at a post-investment valuation of 4 billion yuan
Xixuan Chuangzhi, an embodied intelligence edge-side chip company jointly initiated by UBTECH and MetaX, has completed a 500 million yuan angel round of financing, co-led by GGV Capital and Matrix Partners China, with participation from Huali Capital, CDH Longxin Venture Capital, SAIC Hengxu Capital and others. The company's first self-developed chip is positioned as a "big brain and small brain integrated" SoC for robots, with the goal of comprehensively surpassing NVIDIA's Thor chip. The funds will be used for R&D of core chips and computing platforms and for building a software toolchain, with plans to provide a development and verification platform in 2027 and launch its first SoC product in 2028.
Tongxin Intelligence completes seed and angel rounds totaling hundreds of millions of yuan
Tongxin Artificial Intelligence Technology (Beijing) Co., Ltd. has completed seed and angel rounds of financing totaling hundreds of millions of yuan, with investors including the Tsinghua Alumni Seed Fund, the Beijing Institute for General Artificial Intelligence, Soling Technology, Pingzhi Information, Dark Matter Intelligence and Yuzi Zhangquan. The funds will mainly be used for core architecture design and R&D before chip tape-out; the company's focus is native AGI architecture chips.
Agile Robots receives investment from SoftBank and NVIDIA, valuation reaches $8.5 billion
Germany-headquartered humanoid robot and robotic arm manufacturer Agile Robots announced the completion of $500 million in financing, bringing its cumulative total to nearly $2 billion and its valuation to $8.5 billion. The round was led by SoftBank Group, with participation from NVIDIA, Hitachi and others. According to reports, Agile Robots recently opened a humanoid robot factory near Munich and is expected to produce about 2,000 robots in 2026, with output potentially doubling or even tripling in 2027.
Mirendil completes $1 billion Series A financing
AI R&D tools company Mirendil announced the completion of a $1 billion Series A round led by Kleiner Perkins and Clearlake Capital, with participation from Google and other institutions. The company positions itself as a tool platform that makes frontier AI R&D cheaper and more accessible.
Fractile raises $720 million at a $7 billion valuation
UK AI chip startup Fractile has completed $720 million in financing at a post-investment valuation of $7 billion, with investors including Thrive Capital, Founders Fund, Google Capital G and Lightspeed. The company develops its own AI chips equipped with custom memory, and Anthropic is interested in becoming one of its first customers.
NVIDIA announced investments in four South Korean AI startups, including data center interconnect technology company Point2 Technology, AI foundation model company Motif Technologies, physical AI company RLWRLD and AI solutions company Upstage. Upstage had previously received investment from Amazon and AMD and is advancing its Series C round.
Nscale reportedly considering delaying $35 billion IPO
NVIDIA-backed UK AI cloud service provider Nscale is considering postponing its IPO, originally sized at $35 billion, to next year, and executives are evaluating whether to suspend listing preparations. The company filed for an IPO in September and plans to list on the NYSE.
Trends
JPMorgan: US humanoid robot output to reach 300,000 to 400,000 units by 2030
According to the latest forecast from JPMorgan, as humanoid robots move from technology demonstrations to commercial applications, US humanoid robot output will reach 300,000 to 400,000 units by 2030 and further increase to about 1.5 million units by 2035. Labor shortages, falling usage costs and expanding production capacity are driving manufacturing, logistics and warehousing companies to adopt humanoid robots to replace repetitive manual labor. However, large-scale adoption of humanoid robots still faces constraints in hardware, AI training data, computing power and production capacity. Apart from Tesla, most US manufacturers have annual output targets of only 10,000 to 30,000 units.
First 10,000-GPU-class intelligent computing cluster network technology report officially released in Beijing
On November 15, the first report focused on the development of 10,000-GPU-class intelligent computing cluster network technology, the "Research Report on the Technology Development and Domestic Substitution of 10,000-GPU-Class Intelligent Computing Cluster Networks," was released at the 8th China Supercomputing Conference (ChinaSC 2026) and the annual meeting of the China Intelligent Computing Industry Alliance Council held in Beijing. The report notes that as cluster scale continues to expand, the collaborative efficiency of network interconnection, storage throughput and resource scheduling is becoming increasingly important. How high-speed networks can adapt to the construction of ultra-large-scale clusters is an important focus of the report. The report combines Sugon's self-developed scaleFabric high-speed network and the engineering practice of the Sugon 8000 project to analyze the application progress of domestic intelligent computing network technology. From large-scale networking to communication optimization and reliable operation, Sugon's technological achievements and engineering experience provide a reference for the industry in exploring paths for building ultra-large-scale intelligent computing cluster networks.
CITIC Securities: Certainty of AI applications improving; A-share tech to dip then rise in 2027
On November 9, CITIC Securities said in a research note that new AI application directions and commercialization progress, as well as changes in US Treasury yields, are the two most core variables for the A-share technology sector in 2027. Currently, valuations of both Chinese and US technology sectors are at relatively low points, while EPS remains on an upward trajectory. In terms of timing, A-share technology will likely dip first and then rise in 2027, bottoming out from this year to early next year. It recommends focusing on areas such as physical AI (embodied intelligence supply chain, L3/L4 intelligent driving), AI application diffusion (AI Coding, AI film and television, AI4S, agents) and the domestic computing power chain.
CSC: Optimistic about AI moving from model capabilities to scenario deployment in 2027
On November 11, CSC said in a research note that it is optimistic about the application opportunities brought by AI moving from model capabilities to scenario deployment in 2027. For large models, it expects pre-training to continue scaling and post-training to accelerate; on the computing power side, overseas CSP capital expenditure will continue to grow. Applications will revolve around "fast, big and new": fast deployment, large market size and new application scenarios.
Average online price of laptops rises to 7,759 yuan as memory price increases pass through to terminals
Affected by continued increases in memory chip prices and structural restructuring of the supply chain, the average online selling price of laptops in China rose 13.4% year on year to 7,759 yuan in the third quarter of 2026. Prices rose rather than fell during the Double 11 promotions, while market sales fell 14.0% year on year.