Source; Bold Beast Finance Author: Bold Beast Finance
Beast Finance learned that Amazon (AMZN) recently announced the release of a number of generative AI and AIGC-related products, entering the global large-scale model competition. CITIC Securities released a research report optimistic about the incoming generative AI. CITIC Securities stated in the research report that Amazon, as the largest e-commerce and cloud computing giant in North America, has deep accumulation in application scenarios and computing power reserves. In the future, the company plans to gradually introduce AI products into product content production, search recommendation and other fields, and provide developers with AI tools to assist in product development. We believe that, on the one hand, this move is expected to strengthen the competitiveness of the company's main business, and on the other hand, it can bring medium and long-term demand for the company's cloud computing business. We continue to be optimistic about the company's continued competitiveness in the AI field, and are optimistic about the investment opportunities brought about by the company's performance reversal in the second half of the year.
The main points of CITIC Securities are as follows:
release large language models and AIGC tools, and enter the global AI competition. According to the Amazon Cloud Technology Blog, based on the company's previous accumulation, Amazon has launched related products related to generative AI, covering model, calculation, development and other aspects. Model Layer, publishing Amazon Bedrock and Amazon Titan models, is the easiest way for enterprise customers to build and scale generative AI applications with base models. For the computing layer, Amazon EC2 Trn1n and Amazon EC2 Inf2 are released, based on Amazon's self-developed hardware products, providing the most cost-effective generative AI cloud infrastructure. In the development dimension, Amazon launched the code production tool Amazon CodeWhisperer for individual developers.
Large model: external model + self-developed, constructing a multivariate and efficient model system.Amazon Bedrock is the easiest way for business users to use basic models to build and extend generative AI applications. It lowers the threshold for all developers and allows users to access basic models from AI21 Labs, Anthropic, Stability AI and Amazon through APIs. In addition, users Simply show Bedrock a few examples of labeled data in Amazon S3, and Bedrock can fine-tune the model for a specific task. The Amazon Titan basic model currently includes two new large language models. The first is a generative large language model for tasks such as summarization, text generation (such as original blogs), classification, open question answering, and information extraction. The second is a text embedding (embeddings) large language model, which can translate text input (words, phrases, and even large articles) into digital representations that contain semantics (ie, embeddings embedded codes). The Titan model is mainly used for personalized recommendation and search, and provides additional capabilities such as security.
Computing power and development: create more cost-effective computing and development costs. In the past five years, Amazon Cloud Technology has continued to increase investment in self-developed chips to support model training and reasoning. Trn1 compute instances powered by Trainium can save up to 50% on training costs compared to any other EC2 instance. Inf2 instances powered by Amazon Inferentia2, optimized for large-scale generative AI applications running hundreds of billions of parameter models. Compared with the previous generation, Inf2 instances not only increase throughput by 4 times and reduce latency by 10 times, but also enable ultra-high-speed connections between accelerators to support large-scale distributed inference. In addition to computing power, Amazon provides Amazon CodeWhisperer's code rewriting service. Participants who use CodeWhisperer complete tasks 57% faster on average and have a 27% higher success rate.
Follow-up Outlook: Strengthen Amazon's competitiveness in the AI and cloud fields, and focus on the opportunity for performance reversal in the second half of the year. Facing the layout of Microsoft, Google, Meta and other manufacturers in the AI field, Amazon has not released related products. The release of this product, on the one hand, can complement Amazon's capabilities in areas such as generative AI and large-scale models, and on the other hand, provide a potential impetus for Amazon's medium and long-term cloud computing demand growth. Unlike Microsoft, which directly provides integrated product ideas, Amazon AI provides services for cloud enterprise customers, focusing more on enterprise services and cloud business growth. From the perspective of the company's fundamentals, although in the short term it faces pressures such as a consumer recession and weakening consumption, judging from inflation data and the asset-liability structure of U.S. residents, the U.S. economy is likely to be in a mild recession. At the same time, Amazon continues to reduce costs and increase Efficiency and other measures are expected to continue to release profits. We are optimistic about the opportunities brought by the improvement of the company's fundamentals after the North American economy bottomed out in the second half of the year.
Risks: The duration of high inflation is longer than expected; risks brought about by changes in epidemic policies; risks of declining user spending power due to macroeconomic fluctuations; regulatory risks brought about by anti-monopoly and digital regulation; risks of continued intensification of competition in the North American cloud market, physical retail channels Uncertain factors in development, the slowdown in revenue growth and the acceleration of capital expenditures have led to pressure on profits, and the risk of slowing down the progress of digital content.