Amazon: Retail Efficiency Revisited, AI to be Heavily Invested

This article is the summary of Amazon's 2Q23 performance meeting. For the corresponding financial report comments, please refer to "* * Sales are firm, Amazon is bullish again * *" * * 1. Management Statement * * * 1, Retail Business * * This quarter reported a revenue of US $134.4 billion and an operating profit of US $7.7 billion, both of which exceeded the upper limit of our guidance range. We are encouraged by the progress we have made on several key priorities, including reducing the cost of service in our stores operations; continuing to improve our various customer experiences and building new ones. We will reduce service costs in our store fulfillment network * *. We **decided to transform the distribution and transportation network of our stores from one nationwide network in the United States to eight separate regions serving a smaller geographic area * *. We have a wide range of inventory options in each region, making these products faster and cheaper to reach customers. Regionalization is working. The number of contacts for packages we delivered has decreased by 20%, the number of miles to deliver packages to customers has decreased by 19%, and the number of deliveries within the region has increased by more than 1000 basis points, currently 76%. This is a big step forward. **Amazon is pursuing faster speed while pushing up costs, which is not important to customers. This view is incorrect. Our same-day delivery facility is located in the largest metropolitan area in the United States , with 100000 skus and millions of other skus in nearby execution centers, and is designed to simplify the process from order to ready delivery in just 11 minutes. This experience is so positive for customers in our business that we plan to double the number of these facilities. Although we have seen strong early results from the regionalization efforts, we believe that there is still room for further efficiency improvements * *. We have been particularly focused on providing faster delivery. Our delivery has never been so fast * *. Last quarter, in the 60 largest U.S. urban areas, more than half of Prime members' orders arrived on the same day or the next day * *. So far this year, we have delivered more than 1.8 billion products to Prime members in the United States, almost four times the delivery rate for the same period in 2019. **We now have more than 0.3 billion items available through US Prime Free Shipping, including tens of millions of same-day and one-day free shipping. **This year's Prime Day,Prime members purchased more than 0.375 billion items worldwide and saved more than $2.5 billion in Amazon stores, making it the largest golden member day ever. 2. AWS Due to the uncertain economic situation over the past year, AWS customers need help with cost optimization to withstand this challenging period. We actively help our customers do this. While customers continued to optimize in the second quarter, we are starting to see more and more customers shift their focus to driving innovation and bringing new workloads to the cloud * *. As a result, we saw AWS's revenue growth rate stabilize in the second quarter, up 12% year-over-year. AWS is a leader in compute, networking, storage, databases, data solutions, and machine learning, and continues to invent and deliver in these areas. A few years ago from customers want to find more cost-effective methods to perform general calculations. To achieve this, **we set out to design our own general-purpose CPU chip. Today, more than 50000 customers use AWS Graviton chips and AWS compute instances , including 98 of our top 100 Amazon EC2 customers, and the price/performance ratio of these chips is about 40% higher than other leading x86 processors * *. Now, generative AI has captured people's imagination, but most people are talking about the application layer, specifically what OpenAI is doing to ChatGPT. It's important to remember that we're in the early stages of adopting and successfully generating AI, and consumer adoption is only one layer of opportunity. We believe that there are three key layers to a large language model in generative AI, with the bottom layer being the calculations needed to train the base model and make inferences or predictions * *. Customers are excited about Amazon EC2P5 instances powered by NVIDIA H100gpu to train large models and develop generative AI applications. We started a few years ago to develop our own custom AI chips for training, called Trainium and inference, called Inferentia, and these chips are already available in a second version, **which is a very attractive price-performance option for customers building and running large language models. We are optimistic that much of the future large language model training and inference will run on AWS Trainium and Inferentia chips. **** We think of the middle layer as a large language model as a service. **It takes billions of dollars and years to develop these large language models. Most companies don't want to consume these resources to build themselves. Instead, they want access to those large language models and want to customize those models with their own data without leaking their own proprietary data into a common model. Get all the security, privacy, and platform features in AWS to work with this new enhanced model, and then wrap them all in one managed service. This is what our service Bedrock does by providing customers with not only a large language model, but also models from multiple leading large language model companies such as Anthropic, Stability AI, AI21Labs, coherence and Amazon's own large language model Titan. Customers including Bridgewater Associates, Coda, Lonely Planet, Omnicom, 3M, Ryan air, Showpad, and Travelers are using Amazon Bedrock to create generative AI applications. We just recently announced Bedrock new features, including coherence's new model, Anthropic Claude2 and AI-Stability StableDiffusionXL1.0, as well as Amazon's Bedrock agents that allow customers to create conversational agents that provide personalized, up-to-date answers based on their proprietary data, and perform actions. Then **the top layer is the focus of a lot of publicity and attention, and these are the actual applications that run on top of these large language models * *. As I mentioned earlier, ChatGPT is an example. We believe one of the early compelling applications of generative AI is the coding companion. That's why we built Amazon Code Whisperer, an AI-driven coding companion that recommends code snippets directly in the code editor **to speed up developer coding. It had a very strong head start and changed the game in terms of developer productivity. **AWS not only provides customers with the broadest range of storage, database, analytics, and data management services, but also has more customers and data stores than any other company * *. Coupled with providing customers with unparalleled choice in this three-tier generative AI stack, and Bedrock enterprise-grade security, which is necessary for enterprises to put generative AI applications into production, **we believe AWS will be a long-term partner for customers in generative AI * *. 3. Other businesses **Amazon Business is one of our fastest growing products, with total annual sales of $35 billion. The team is working to further build the optionality, value, convenience and functionality that business customers need. **** Buy with Prime service **continues to make great progress. Merchants using Buy with Prime service found that their * * customer conversion rate increased by an average of 25% * *. In addition, the daily Buy with Prime orders of merchants participating in member day activities during the promotion increased 10 times compared with the month before we announced the membership. Frankly, our decision to invest heavily in the healthcare market sector was made in a short period of time. A lot of what we tried before was small-scale experiments. But we're thrilled that Amazon Pharmacy's active customers have doubled in the past year and we're pleased with the response to Rx Pass, which gives Prime members access to all eligible generic drugs for as little as $5 a month with free home delivery * *. The One Medical only been part of Amazon for a few months, and we're also encouraged by what we're seeing there. Our grocery business continues to grow. We already have a very large business in non-temperature-controlled areas such as consumables, pet food, beauty products and canned foods, and our business continues to grow as we continue to increase speed and reduce service costs, which allows us to sell more items more effectively. Whole Foods, which continues to lead the organic grocery segment, is growing at a healthy pace and has significantly increased its profitability over the last year. We are pleased with the situation at Whole Foods. As I shared before, we are developing a new format for our mass store product "Amazon Fresh" (AmazonFresh), which has significantly improved the number of key business inputs and introduced new concepts in the store. In other areas like Kuiper, Zoox, and Alexa, we 've also seen substantial innovation and progress. We are still in the relatively early stages of investing in many technological inventions that are changing the possibilities that these areas offer customers, but they are huge long-term opportunities that we remain optimistic about. Q & A Q1: In terms of performance efficiency, how is efficiency optimization progressing? Is grid optimization likely to help grocery store expansion? A1: In terms of performance efficiency, we are again encouraged by regionalization and efforts to control cost structures, and some **inflationary pressures are also declining * *. However, **we are still in the process of restoring the previous cost structure * *. We 've seen some margin improvement, particularly in the last 3 to 5 quarters, and we 've been encouraged, but we're continuing. **There is still a lot to recover in our compliance area and the team is working very hard. * * There are many facets to our grocery business. There's consumables, there's fresh goods, there's Whole Foods. As Andy mentioned, we are able to expand our business in these areas because we can afford it with shorter distances to customers and more fixed transportation costs. We get some optimization and leverage from the fulfillment network. to increase the grocery items that people can get on the day. Over time, we will be able to take advantage of people being able to choose different items from different grocery stores, which are different grocery store models. **But it is also true that we have built different infrastructure in other businesses and also optimized for the grocery business. **** Q2: What is the growth strategy for Amazon Business? **A2: For Amazon Commerce, $35 billion in total annual sales is pretty strong growth. Our team * * is working hard to build a business worth more than $100 billion. **We currently only have a small part of the functions needed to solve more enterprise problems. Therefore, we have added a lot of functions, which need to be enriched around helping procurement and doing some services. **Our development there is not over, and this is an area that our store team and our senior leadership team are very focused on. A3: What we're seeing in the quarter is that cost optimization, while still ongoing, is slowing and we're seeing more new workloads, new businesses * *. **We do not intend to give segment guidance for the third quarter. But I would add that we saw the trend in the second quarter continue into July. So I think our business has stabilized, and we look forward to the end of the year, **because as Andy said, we will launch a lot of new features, and we will invest a lot of money in this area to provide excellent solutions for generating AI and large language models, and machine learning solutions that we have been providing customers. So we're optimistic and are starting to see some good traction with our clients' new business. The AWS business is a $88 billion in revenue business, and to achieve double-digit growth with cost optimization, you must add a lot of new customers and a lot of new workloads. As we start to see the weakening of cost optimization, more and more workloads, new workloads that people are starting to plan, not to mention generating things that AI brings, AWS has a lot of growth in front of us. Q4: What is the outlook on AI monetization? How is the impact on/by AWS business? A4: What I want to say is that for several years, we have a lot of business on AWS driven by machine learning and AI. Because customers have been doing a lot of machine learning training and then running their models and products on AWS and our compute instances. I think everybody's excited, including us, when you talk about the great potential of generative AI, and I think we're still in the very early stages. It seems to me that we are a few steps into the marathon. I think it's going to be revolutionary, and I think it's going to change almost every kind of customer experience we know. But I think it's too early. I think **most companies they don't want to build their own very large language models. They want to take other models and customize them, and services like Bedrock allow them to do that, but it's too early. So I expect the profit will be very large, but it will be in the future. **** Q5: Regarding the grocery business, how do you think/think about the latest initiatives? **A5: **We have been working on Amazon Fresh for several years. What I want to say is that we are not satisfied with the progress in terms of input and input * *. In the past year, the team has worked very hard, including the right inventory level, the right cost structure, the right scrap rate and other data. We're trying to build a model in Amazon's fresh mass store that resonates more with customers. **But unless we see this resonance, we will not expand. **** We will also continue to provide delivery services to non-prime members * *. So I think we have a lot of opportunities to grow the business over time. We are optimistic that we can do this, but we are also disciplined to not expand physical fresh stores until we think there is a model that resonates more with customers. **** Q6: How do you see the upward trend in North American retail margins? How do you invest in some new initiatives in the retail business? A6: I will start with the North American retail segment, and I would like to remind everyone again that we will not expand the number of fresh stores in a very significant way until we believe that we have something that resonates with customers and we will like the return on invested capital * *. **We would also like the corresponding cost structure * *. **I do believe we will return to pre-pandemic margins. **** Q7: Regarding the investment in AI (perhaps through logistics or AI-based agents, etc.) to improve your own retail and device network, how high is the priority? **A7: Regarding the issue of AI, what I want to tell you is that every business we have within Amazon, every business has multiple generative AI initiatives underway. From helping us improve cost efficiency and simplify the way we operate in our various businesses, to the absolute core of every customer experience we provide. The same is true in our store business. The same is true in our AWS business. The same goes for our advertising business. This is true across all of our devices, and you can imagine what we're doing with Alexa. The same is true in our entertainment business, every one of them. This will be at the heart of our work. For us, **this is an important investment and focus * *. Q8: How do you see the normalized growth of AWS in a better macro environment? A8: I expect to continue cost optimization, but I think the balance of cost optimization is really new workloads and new migrations, and as we saw the shift in Q2, I expect we will continue to see this shift over time. Like I said, I mean, everybody has to make their own conclusions about what they think is the percentage increase in revenue. But to achieve double-digit growth at $88 billion in revenue growth, when you see every company in the world last year trying to save as much money as possible, you will see the amount of cost optimization. Therefore, I am very optimistic about the growth of AWS in the next few years. It's hard to predict a single quarter, but I'm definitely bullish on it in the medium to long term. **** Regarding how much generated AI may affect capital expenditures, this figure includes considerable capital expenditures for large language models and generated AI in the AWS business * *. So we will invest more in generative AI, this is our current best estimate of capital expenditure, and if we find that the situation is different, we will update it. Risk disclosures and statements in this article: Dolphin Research Disclaimer and General Disclosures