Core Insights - The article discusses the rapid growth of data usage in AI models, particularly highlighting the performance of various models in terms of token usage and their respective developers [1][3]. Group 1: AI Model Performance - Grok Code Fast leads with 1.25 trillion tokens, showing a 16% increase by x-ai [3] - Claude Sonnet 4.5 follows with 527 billion tokens, achieving a 15% increase by anthropic [3] - Gemini 2.5 Flash has 298 billion tokens, with a significant 43% increase by google [3] - DeepSeek V3 0324 has 110 billion tokens, with a notable 44% increase by deepseek [3] - The performance of Gemini 2.5 Pro is also highlighted with 168 billion tokens, showing a 110% increase by google [3] Group 2: Industry Trends - The article indicates that computational power is expected to continue growing, particularly with companies like TSMC and MediaTek [5] - There is an ongoing tracking of major companies' financial reports, indicating a busy period for industry analysis [5]

DeepSeek, a Chinese software firm, surprised the world when releasing an AI model that was competitive with Western rivals, using a fraction of the computing power. Now China’s chipmakers are trying to do the same https://t.co/Mgz3pOtLPu ...

Investment Rating - The report maintains an "Accumulate" rating for the media industry [7] Core Views - The media sector saw a 4.20% increase during the week of October 20-24, driven by rebounds in gaming and themes related to animation and drama [12][20] - The report highlights optimism in the gaming sector and anticipates a turnaround in the film and television sector due to new policy drivers [1][20] - AI applications are emphasized, particularly those that can realize data monetization, with a focus on AI companionship, education, and toys [1] - The report also notes the importance of IP monetization, particularly for companies with IP advantages and full industry chain potential [1] Summary by Sections Market Overview - The media sector's performance was strong, with a 4.20% increase, while the top three performing sectors were telecommunications (11.56%), electronics (8.11%), and machinery (4.80%) [12][14] - The report identifies key stocks in various sub-sectors, including gaming, film, IP, AI, and education [20] Sub-sector Insights - **Gaming**: Key companies to watch include ST Huatuo, Giant Network, and Jiubite [20] - **Film**: Focus on Mango Super Media, Huace Film, and Huanrui Century [20] - **IP**: Companies like Chuangyuan Co., Shanghai Film, and Huali Technology are highlighted [20] - **AI**: Notable companies include Doushen Education, Shengtian Network, and Visual China [20] - **Education**: Key players include Xueda Education and Fenbi [20] Key Events Review - The report discusses the Chinese government's emphasis on accelerating AI and digital technology innovation as part of the 14th Five-Year Plan [3][22] - OpenAI's launch of the AI-driven browser ChatGPT Atlas is noted, which competes directly with Google Chrome [4][22] - The report mentions the release of DeepSeek's new OCR model, which addresses computational bottlenecks in processing long texts [5][22] Sub-sector Data Tracking - The report provides insights into the gaming market, noting a total sales revenue of 880.26 billion yuan for Q3 2025, with a 6.96% quarter-on-quarter increase [22] - The film market's total box office for the week of October 18-24 was 205 million yuan, with "Wang Wang Life" leading the box office [27][28] - The report tracks the performance of various TV series and variety shows, highlighting the top performers in terms of viewership [30][31]

Core Viewpoint - DeepSeek has introduced a new model, DeepSeek-OCR, which utilizes images to store information instead of relying solely on text tokens, significantly improving data compression and model efficiency [5][11][26]. Group 1: Model Functionality - DeepSeek-OCR can convert large amounts of text into images, serving as a memory carrier for AI, which allows for more efficient data storage [9][14]. - The model demonstrates superior performance by using fewer visual tokens compared to traditional models, achieving better results with less resource consumption [11][26]. - In tests, DeepSeek-OCR used only 100 visual tokens to outperform GOT-OCR 2.0, which required 256 tokens, and it achieved results with less than 800 visual tokens compared to over 6000 tokens for MinerU 2.0 [11][14]. Group 2: Data Collection and Utilization - The model can capture previously uncollected data from two-dimensional information, such as graphs and images in academic papers, which traditional models could not interpret [22][24]. - DeepSeek-OCR can generate over 200,000 pages of training data in a day on an A100 GPU, indicating its potential to enhance the training datasets for future models [24]. - The model's ability to remember the position of images and surrounding text allows for a more comprehensive understanding of the data [18][22]. Group 3: Resource Efficiency - By using image-based memory, DeepSeek-OCR can reduce the number of tokens required to one-tenth of the original, while maintaining a high accuracy rate of 96.5% [26][27]. - The model's design allows for dynamic adjustments in token usage based on the complexity of the document, optimizing resource allocation [14][15]. - The research indicates that even with a 20-fold compression, the model can retain around 60% accuracy, showcasing its robustness [27]. Group 4: Open Source Collaboration - DeepSeek-OCR is an open-source project that integrates contributions from various global open-source communities, utilizing datasets and models from companies like Huawei, Baidu, Meta, and OpenAI [32][34]. - This collaborative effort has resulted in a model capable of "thinking in images," highlighting the importance of community-driven innovation in AI development [34].

Core Insights - The AI landscape in 2025 is characterized by a strong sense of division, moving from a "Warring States" era to a clear bifurcation of narratives: one focused on performance enhancement and the other on cost reduction [2][3] Group 1: Performance vs. Cost - The "upward" narrative is led by industry giants like Alibaba, with their Qwen3-Max model achieving top rankings in various benchmarks, showcasing a commitment to scaling and performance [3][9] - The "downward" narrative is driven by emerging tech companies like DeepSeek, which has significantly reduced API prices by over 50%, indicating a shift towards cost efficiency through technological breakthroughs [3][12] - This dual narrative reflects a broader conflict between maximizing performance and minimizing costs, reshaping the commercial logic of AI [4][10] Group 2: The Scale vs. Efficiency Debate - AI giants are heavily invested in the "Scaling Law," with Qwen3-Max exemplifying the belief that larger models yield better performance, despite the high costs associated with training and inference [10][11] - Conversely, the rise of open-source models presents a challenge to the commercial viability of expensive proprietary models, as developers may prefer free alternatives [11][12] - The competition is evolving from a focus on sheer scale to one that emphasizes algorithmic efficiency and cost-effectiveness, marking a transition to a new phase in AI development [16][20] Group 3: Market Dynamics and Implications - The significant price gap created by high-end models versus cost-effective alternatives is a central tension in the AI industry, impacting the fortunes of small and medium enterprises versus large corporations [17][18] - The ongoing API price war among major players like Alibaba, Baidu, and Tencent reflects their struggle to maintain market share while managing the high costs of flagship models [18][19] - The conflict between efficiency-driven startups and scale-focused giants is pushing the industry towards a reevaluation of value creation, emphasizing measurable benefits for clients [20][22]

Global Economic Outlook - Goldman Sachs expects a more sustained upward trend in Chinese stocks in the future, and the structural shift of Chinese capital into stocks may have begun. The willingness of Chinese market entities to settle foreign exchange has significantly increased. Traders are increasing bets that the Fed will implement at least one 50 - basis - point interest rate cut in its meeting later this month or in December [4] - Although AI infrastructure investment has reached a record high in nominal terms, compared with historical technology cycles, it is not exaggerated. Currently, US AI investment accounts for less than 1% of GDP [4] - Meta has joined hands with private - equity giant Blue Owl to raise $27 billion through the issuance of private bonds to build data centers. DeepSeek has launched a revolutionary OCR model to solve the computing power problem of AI in processing long documents [4] - An executive at Apollo Global Management warns that there is a huge gap between the huge energy demand of artificial intelligence and the current global power supply, and this gap may not be filled in our lifetime [4] - Due to the continuous wrong policies of the US, the global economy is entering the top - region [5] - After the US imposed reciprocal tariffs on various countries, the competitiveness of Chinese goods has increased, and the US imports from China in August increased by nearly 40% month - on - month [6] - The US Market manufacturing PMI in September was 52.0, continuing to expand [9] - In August, the US manufacturing unfilled orders were at a record high, with a year - on - year increase of 7.1%, indicating high business sentiment [12] - The US capital goods import amount in August was $91.9 billion, still at a high level, and the year - on - year growth rate was 10.5%, indicating that the US manufacturing industry is accelerating its return and the US "re - industrialization" is speeding up [15] - The US wholesalers' sales in August were $711.3 billion, a record high, with a year - on - year increase of 6.2%, indicating strong US consumption [18] - The total retail and food sales in the US in August increased by 0.6% month - on - month, exceeding expectations, showing strong US consumption [21] - The eurozone's manufacturing PMI slightly contracted in September, while the service - sector PMI accelerated its expansion [24] - India's manufacturing and service - sector PMIs continued to expand in September, and India's manufacturing and service industries have maintained expansion for more than three years [26] - The upward trend of Japan's long - term government bond yields has not changed [29] Asset Allocation - The Fed is expected to continue cutting interest rates in October [31] - Wait for the Sino - US agreement at the end of the month in the context of Sino - US game [32] - The market's defensive state has not ended, and the Shanghai Stock Exchange 50 Index, representing the value style, has reached a new high. The Fourth Plenary Session of the 20th CPC Central Committee's communiqué emphasizes seizing the high - ground of scientific and technological development, and the technology sector has become active again. Policies are favorable to semiconductor equipment ETFs and science and technology innovation chip ETFs [33] - AI infrastructure drives a rapid increase in power demand, leading to a stronger copper price [34][47] - The container shipping European line 2512 contract has entered the seasonal peak season [35][50] - Awaited for the clarity of the Sino - US agreement at the end of the month, the CSI 300 Index is both offensive and defensive, and the stock - index allocation is mainly long positions in CSI 300 stock - index futures [39] Lithium Carbonate Market - The growth rate of lithium carbonate demand is much higher than the production growth rate. In October 2025, the planned production of power + energy - storage + consumer batteries in the Chinese market was 186 GWh, a month - on - month increase of 22.4% and a year - on - year increase of 45.3%. The planned production of lithium - battery A was 61.3 GWh, lithium - battery B was 30.9 GWh, and lithium - battery C was 13 GWh. The planned production of energy - storage cells accounted for about 40.3%, and the planned production of ternary cells accounted for about 15%. The average operating rate of the lithium - battery industry was close to 90%, and the production lines of leading enterprises were operating at full capacity [41] - In October 2025, the total production of power + energy - storage + consumer batteries in the global market was about 205 GWh, a month - on - month increase of 25.8%. Currently, the increased production of domestic spodumene and salt - lake lithium mines offsets the reduction in lithium - mine production, but the total supply has not increased, while the demand has increased significantly, resulting in nine consecutive weeks of inventory reduction, which will continue in the future [42] - PwC reports that in 2026, thanks to the explosive growth of energy - storage and the continuous growth of power batteries, the total demand for lithium carbonate will reach 2.05 million tons of LCE, a year - on - year increase of 35% compared with 1.55 million tons in 2025 [42] - The expansion of lithium carbonate production capacity is expected to slow down significantly in 2026, with an expected growth of only about 10%, and the new supply will be difficult to match the continuously high - speed growth in demand [43] - The surge in energy - storage demand, with the demand growth rate of lithium carbonate far exceeding the production growth rate, has pushed up the price of lithium carbonate [44]

Core Insights - The software industry in China has shown continuous growth from January to August 2025, with software business revenue reaching 9.64 trillion yuan, a year-on-year increase of 12.6%, marking six consecutive months of recovery [2][3] - The total profit of the industry reached 1.3186 trillion yuan, reflecting a year-on-year growth of 13.0%, which is 0.4 percentage points higher than the revenue growth rate [2][3] Sub-industry Highlights - **Domestic Software Growth**: Basic software revenue increased by 13.6% year-on-year, 6.7 percentage points higher than in 2024, and 1.0 percentage point above the overall software industry growth rate. This growth is expected to benefit from the demand release in the closing year of the "14th Five-Year Plan" [3] - **AI Server Market**: The demand for AI servers in China is increasingly concentrated in the internet sector, with a significant impact from the trend of domestic chip localization on market competition. The cloud business of internet companies is expected to see positive growth driven by AI [3] - **AI Model Updates**: Recent updates in AI models have focused on optimizing compatibility with domestic chips, addressing ecological barriers posed by CUDA, and enhancing hardware-software synergy in the AI industry [4] Investment Recommendations - Companies to watch include EDA firm Huada Jiutian (301269), Runze Technology (300442) with large-scale intelligent computing center delivery plans, and Zhongke Xingtou (688568) actively exploring low-altitude economy and commercial aviation [5]

Core Viewpoint - The implementation of the "Artificial Intelligence Generated Synthetic Content Identification Measures" in China mandates explicit and implicit labeling for all AI-generated content, aiming to enhance transparency and combat misinformation in the digital landscape [4][9][10]. Regulatory Framework - The new regulations require all AI-generated text, images, and videos to have a clear identification, referred to as a "digital identity card" [4][5]. - The measures include both explicit labeling, which is easily noticeable by users, and implicit labeling, which embeds metadata within the content for traceability [6][7]. Industry Impact - As of June 2025, the user base for generative AI in China is projected to reach 515 million, highlighting the rapid growth and potential challenges in content verification [5]. - Major platforms like Douyin, Kuaishou, and Xiaohongshu are already implementing these identification measures, enhancing user awareness and compliance with the new regulations [8][9]. Technical Specifications - Explicit labels must be clearly visible and placed at the beginning, end, or key positions of the content, while implicit labels involve embedding metadata that includes identifiers and verification information [6][7]. - The regulations aim to create a comprehensive governance model that covers the entire content lifecycle, from creation to distribution [7][10]. Social Responsibility - The measures are designed to mitigate the risks associated with AI-generated misinformation, particularly targeting vulnerable populations who may be misled by unverified content [9][10]. - The initiative encourages companies to integrate ethical considerations into their technology development processes, promoting responsible AI usage [10].

RT Saritha Rai (@SarithaRai)A deep dive into DeepSeek's surge in Africa & Chinese AI's inroads into a continent that could become key to more users, more human-generated data & loads of soft power @business @LoniPrinsloo1 @HelenNyambura(gift link: free to read until Oct 30)https://t.co/5hvnphlADV#AI https://t.co/iy720d9v6C ...

Report Key Points 1. Industry Investment Rating - There is no specific industry investment rating provided in the report. 2. Core View - Goldman Sachs expects a more sustained upward trend in Chinese stocks, with the MSCI China Index predicted to rise by about 30% by the end of 2027, driven by 12% trend - based earnings growth and 5% - 10% further revaluation potential [1]. - Due to the focus on STEM education and talent cultivation, China is taking the lead in industrial innovation, with a shift in enrollment trends towards STEM majors and a decrease in majors like law, economics, and management [1]. - The recent rebound in US stocks is a "short - squeeze" market caused by short - covering, which may send false optimistic signals [1]. - The global economy is entering the top region due to the continuous wrong policies in the US [2]. 3. Summary by Related Information Global Economic News - Meta and private equity giant Blue Owl raised $27 billion through private bond issuance to build data centers, with BlackRock subscribing over $3 billion and Pimco subscribing $18 billion. The bond has a S&P A+ rating but a high yield of 6.58% [1]. - OpenAI launched a new AI - driven web browser, ChatGPT Atlas, to compete directly with Google Chrome [1]. - Anthropic is in talks with Google for a potential multi - billion - dollar cloud computing agreement, and Google has previously invested about $3 billion in Anthropic [1]. - The zinc market on the London Metal Exchange is facing its worst squeeze in decades, with available zinc inventory insufficient for one - day demand and the spot zinc premium soaring to $323 per ton [1]. Chinese Market News - Chinese stocks are expected to have a sustained upward trend, and Chinese capital may have started a structural shift towards stocks. The central parity rate of the RMB against the US dollar has been raised to 7.0995, the first time since November last year [1]. - According to a survey of emerging markets, China is the top choice for stock investment, with 100 surveyed institutions managing $423 billion of emerging - market assets [2]. - Huawei announced the evolution and goals of its Ascend chips, with its computing power in "super - node + cluster" leading NVIDIA by over a year [2]. US Market News - US AI investment accounts for less than 1% of GDP, and although AI infrastructure investment has reached a nominal high, it is not exaggerated compared to historical technology cycles [2]. - DeepSeek launched a revolutionary OCR model to solve the computing power problem of AI in processing long documents [2].