Group 1: Nvidia - Nvidia's revenue and profit exceeded expectations, but the guidance for the upcoming quarter was less impressive, with a focus on the "absence of China" as a key issue [10][12] - The company reported a year-on-year revenue growth rate of over two years' lowest, yet still above analyst expectations, with a significant drop in data center revenue due to a $4 billion decrease in H20 sales [12] - Nvidia announced a new $60 billion share buyback authorization and highlighted a potential $50 billion business opportunity in China for the year [10][12] Group 2: Meituan - Meituan's Q2 adjusted net profit plummeted 89% year-on-year to 1.49 billion yuan, significantly below expectations, with marketing expenses increasing by 51.8% [10][12] - The company achieved a revenue growth of 11.7% year-on-year, but its operating profit fell by 98% to 230 million yuan, with a drastic drop in operating profit margin from 13.7% to 0.2% [12] - Meituan's core local business revenue was 65.3 billion yuan, a 7.7% increase year-on-year, but the company anticipates significant losses in Q3 due to ongoing fierce competition [10][12] Group 3: Snowflake - Snowflake reported strong earnings, raising its full-year guidance, which led to a 13% increase in its stock price post-announcement [6][16] - The company’s remaining performance obligations reached $6.9 billion, a 33% year-on-year increase, indicating strong long-term customer investment [16] Group 4: Honey Snow Group - Honey Snow Group reported a 39.3% year-on-year growth in the first half of the year, with net profit increasing by 44.1% and global store expansion reaching 53,000 [15] Group 5: Global Economic Context - The U.S. stock market showed volatility ahead of Nvidia's earnings report, with the S&P 500 barely reaching a new high, while the Chinese concept stock index fell by 2.58% [2] - The U.S. Treasury yields fell across the board, with the 2-year yield dropping over 6 basis points, amid expectations of continued monetary policy easing from the Federal Reserve [2]

Core Viewpoint - The article discusses the rising trend of low-precision computing, particularly focusing on the FP8 format, driven by the increasing demand for AI computing power and the advancements in domestic chip technology. The release of DeepSeek-V3.1 marks a significant step towards the adoption of low-precision calculations in the industry, which is expected to enhance efficiency and reduce costs in AI model training and inference [3][11][12]. Group 1: Industry Trends - The chip index and AI computing power index have shown significant growth, with the chip index rising by 19.5% and the AI computing power index increasing by 22.47% over the past month [3]. - The introduction of DeepSeek-V3.1, which utilizes UE8M0 FP8 parameters, is seen as a pivotal moment in the transition to the agent era in AI [3][11]. - The industry is shifting from a focus on acquiring GPUs to optimizing computing efficiency, with low-precision formats like FP8 gaining traction due to their advantages in memory usage and processing speed [9][10]. Group 2: Technical Developments - FP8 is an 8-bit floating-point format that offers significant benefits over traditional formats like FP32 and FP16, including reduced memory usage (0.5x for FP8 compared to FP16) and improved transmission efficiency (2x for FP8) [10]. - The adoption of FP8 has been facilitated by the establishment of the MX specification by major tech companies, which allows for large-scale implementation of low-precision calculations [8][9]. - The successful application of FP8 in complex AI training tasks by DeepSeek is expected to attract attention from AI developers and research institutions [9][12]. Group 3: Market Dynamics - The market share of domestic chips is projected to rise to 38.7% by Q2 2025, reflecting a growing trend towards domestic alternatives in the AI chip sector [12]. - The shift towards low-precision computing is driven by the need for more efficient hardware to support the increasing computational demands of large AI models [12][17]. - The domestic AI chip industry is moving towards a sustainable path, with a positive cycle established between financing, research and development, and practical applications [17].