On Thursday, April 13, Andy Jassy, who took over as CEO of Amazon in 2021 from founder Bezos, released his 2022 annual shareholder letter, saying that he is confident that Amazon can control costs and continue to invest in new areas of growth.
This is also the second shareholder letter issued by Jassy since becoming CEO. Most importantly, he confirmed in his letter that Amazon will invest heavily in areas such as large language models (LLM) and generative artificial intelligence (AI), which are popular today.
Amazon CEO called LLM and generative AI transformative and will continue to invest heavily and join the generative AI race
Jassy said LLM and generative AI are "at the heart of how Amazon can innovate in every area of its business for decades to come" and will significantly accelerate the adoption of machine learning that Amazon has been working on for 25 years, calling generative AI transformative and meaningful to customers, shareholders and Amazon:
"Amazon has been developing its own LLM for some time and believes it will transform and improve nearly every customer experience, and will continue to invest heavily in these models across all of our consumer, seller, brand, and creator experiences.
As with the path AWS has been taking for years, we're pushing companies of all sizes to take advantage of generative AI. AWS offers the most cost-effective machine learning chips, Trainium and Inferentia, making it affordable for both small and large companies to train and run their LLMs in production.
Amazon's business customers can choose from a variety of LLMs and build applications using all the AWS security, privacy, and other features that customers prefer. In addition, we are offering applications like AWS CodeWhisperer, which revolutionizes developer productivity by generating code suggestions in real time. ”
On the same day that the shareholder letter was released, Amazon announced its participation in the generative AI competition with the launch of a new cloud service called Bedrock, where developers can enhance their software with artificial intelligence systems that can generate text.
Customers also have access to Amazon's own large-scale language model "Titan," as well as language models from startup AI21 and Google-powered Anthropic, as well as Stability AI's model for converting text into images.
The macroeconomic downturn slowed Amazon AWS's growth last year, bullish on the potential of groceries, healthcare and satellite internet
In his shareholder letter, Jassy admitted that 2022 was one of the toughest years in recent macroeconomic times, with Amazon facing some operational challenges, such as slower AWS growth due to more cautious customer spending on the cloud, but the company achieved unprecedented growth.
He said the last few months have been "going deep into the entire company on a business-by-business basis," checking whether each division can generate sufficient revenue, operating profit, free cash flow and return on invested capital, "which in some cases has led us to close certain businesses." ”
In order to cut costs, Amazon has cut a telemedicine service and other experimental projects in recent months, suspended the expansion of fresh supermarkets, and scaled back the size of warehouses. Amazon also implemented the largest layoffs in its 29-year history, laying off 27,000 corporate employees through multiple rounds of layoffs and freezing hiring.
Subsequently, the shareholders' letter business introduced the development situation of various departments and future investment ideas.
In the core e-commerce retail business, Jassy said that Amazon continues to face key challenges such as rising service costs on the fulfillment network, and after a series of adjustments in two years, it has been able to cut costs and increase delivery speed, hoping to accelerate delivery for Prime members in 2023.
As the world's largest cloud service provider, Jassy said that AWS's annual revenue will reach $85 billion (run rate), Amazon's advantage is to help customers optimize cloud computing costs in times of economic uncertainty, adhere to "customer-centric and long-term strategy orientation", and is committed to investing in machine learning training chip Trainium and inferentia chip Inferentia. He predicted: "AWS is still in the early stages of development and has the opportunity to achieve extraordinary growth over the next decade. ”
In the advertising business, Jassy said that Amazon can customize advertising services for brand-sponsored products based on its understanding of shopping behavior and deep investment in machine learning algorithms to make advertising more efficient for consumers to make advertising more relevant to what consumers are searching for, so Amazon's advertising revenue continues to grow rapidly when the growth of most large advertising-focused enterprises slows down (23% year-over-year in the fourth quarter of 2022 and 25% year-over-year in 2022, with full-year revenue of $31 billion).
While cutting employees and cutting costs of its main business, Amazon is also continuing to invest in new growth areas, including: expanding the international market of consumer e-commerce retail business, expanding product categories in existing large retail markets (Amazon Fresh), Amazon Business, a solution that facilitates the purchase of office supplies and other bulk commercial supplies, helping Prime users pay through Amazon accounts on third-party seller direct websites. Healthcare (Amazon Healthcare) and satellite internet Kuiper, among others.
The shareholder letter mentioned that "grocery" is a huge growth opportunity for Amazon, expanding into the international market, exploring the large-scale retail field that has not yet been involved, and using its unique resources to help merchants sell products more effectively on its own website, which are the more natural business expansion directions for Amazon. Investments that appear to be further away from the core business, such as healthcare and Kuiper, also have unique growth potential.
Finally, Jassy said he believes that "Amazon will emerge from this challenging macroeconomic downturn and be in a better position than ever before." He highlighted the company's "two great future opportunities," namely:
Amazon's consumer business generated $434 billion in revenue in 2022, but the vast majority of the total global retail market share still belongs to physical stores (about 80%). The same is true for global IT spending, where AWS generated $8 billion in revenue in 2022, but 90% of global IT spending is still in on-premises data centers and has not yet migrated to the cloud.
The following is the full text of Jassy's letter to shareholders:
Dear Shareholders,
This is my second annual shareholder letter as CEO, and I find myself optimistic and energetic about Amazon's future.
Although 2022 was one of the more difficult macroeconomic years in memory, and the company experienced some operational challenges of its own, we still found ways to increase demand, especially after experiencing unprecedented growth during the pandemic. We've innovated our biggest businesses to significantly improve the short- and long-term customer experience. We've made important changes to our investment decisions and how we're moving forward, and retained long-term investments that will transform Amazon's future for our customers, shareholders, and employees.
While there have been many unusual challenges that have emerged simultaneously over the past year, the truth is that if you operate in a large, dynamic global market segment with many capable and well-funded competitors (which is the condition for all of Amazon's operations), the external environment will rarely remain the same.
In the 25 years I've been with Amazon, changes have been happening, many of them initiated by ourselves. When I joined Amazon in 1997, Amazon had a turnover of only $15 million in 1996, was a book-only retailer, had no third-party marketplace platform, and shipped only to addresses in the United States. Today, Amazon sells almost every physical and digital retail item imaginable, has a vibrant ecosystem of third-party sellers that account for 60% of our unit sales, and reaches customers in nearly every country in the world.
Similarly, when we started thinking about AWS, Amazon's cloud computing service platform, in 2003, the trend of building new businesses around a set of technology infrastructure services in the cloud was not yet apparent, and we were very ahead of our time when we launched the first such service in 2006. When Amazon launched its Kindle e-reader in 2007, "almost every book was easily accessible in a minute, stored and retrieved on a lightweight digital reader," which was not yet popular. When Amazon launched Alexa as a voice-driven personal assistant in 2014, the act of "using it to access entertainment content, control smart homes, shop, and retrieve all kinds of information" also did not enter the public eye.
Sometimes, macroeconomic conditions or poor operational efficiency present us with new challenges.
For example, during the dot-com bubble burst in 2001, we had to obtain letters of credit to buy holiday season inventory and optimize costs to improve profitability, but we still prioritized long-term customer experience and business building (during that uncertain 2001, we actually lowered prices in most categories).
For example, during the 2008-2009 recession triggered by the subprime mortgage crisis, we took a number of steps to manage the cost structure and efficiency of our store operations, reducing these expenses while continuing to invest in the customer experience, as we believe these investments will be a significant business in the future, generating strong returns to shareholders. In 2008, AWS was still a small startup. We know we're creating something special, but it still requires a lot of capital investment.
There are constant questions inside and outside the company as to why Amazon, which at the time was primarily seen as an online retailer, invested so much in cloud computing, but we knew that special businesses were being created that would bring a lot of value to customers and Amazon in the future. That's why we're one step ahead of our potential competitors, and we're more eager to innovate faster, so we made the long-term decision to continue investing in AWS.
Fifteen years later, AWS is now an $85 billion annual business that is highly profitable and has changed the way startups, multinationals, public sector customers, and many other types of customers manage their technology infrastructure. If we had slowed our investment in AWS during 2008-2009, Amazon wouldn't have been what it is now.
Change is always happening, sometimes on your own initiative and sometimes uninvited. But when you see it coming, you have to meet it. Companies that are often successful in the long term respond well to change. I'm optimistic about Amazon's future prospects because I like how our team responds to current changes.
Over the past few months, we've taken a deep dive into the company, reviewing every business, every creation, asking ourselves if we have confidence in the long-term potential of each initiative and whether it will generate sufficient revenue, operating profit, free cash flow and ROI. In some cases, this has led us to close certain businesses, such as stopping pursuing brick-and-mortar concepts like bookstores and four-star stores, exiting Amazon Fabric (selling fabrics online) and Amazon Care (telehealth services) programs, and abandoning some new hardware devices that don't see a meaningful path to return.
We also reviewed and revised some items that failed to generate the expected returns (such as free shipping on all online grocery orders over $35). We also redefined the priority areas for allocating resources, culminating in the difficult decision to lay off 27,000 employees at the company. Over the past few months, we've made a number of other changes to optimize overall costs and, like most leadership teams, will continue to assess what we see in the business and respond flexibly.
We also carefully reviewed how our teams collaborated and asked our employees to return to the office at least three days a week starting in May. During the pandemic, our employees have come together to get work done from home and are doing everything possible to respond to the unexpected. This is impressive, and I'm proud that our collective team has overcome unprecedented challenges for our customers, communities, and business.
However, we do not believe that this is the best long-term approach. We believe that working together face-to-face is good for mutual learning, it also makes collaboration and invention easier and more efficient, and people can communicate with each other and ideas more freely. People stay after a meeting, discuss ideas on a whiteboard, continue to communicate on the way to the end of the meeting, or visit a colleague's office later in the day and come up with a new idea—that's how many of Amazon's best inventions came out of nowhere.
Invention is often chaotic. It wanders, wanders, and winds irregularly. Casual interactions help with this, while face-to-face interactions are more effective than virtual interactions. When we spend most of our time in the office with our colleagues, it's also easier to think, shape, practice, and grow our culture. Innovation and our unique culture have been important throughout the company's 29-year history, and I expect the same to be true for the next 29 years.
A key challenge we continue to face is the rising cost of service (i.e., the cost of delivering products from Amazon) to customers across our store fulfillment network, and we have made several changes that will meaningfully improve fulfillment costs and delivery speeds.
In the early days of the pandemic, when many brick-and-mortar stores closed, Amazon's consumer retail business grew at an alarming rate, increasing annual revenue from $245 billion in 2019 to $434 billion in 2022. This meant that we had to double the size of fulfillment centers that had been established over the past 25 years and dramatically accelerate the construction of an end-of-line transportation network, which is now comparable to UPS (while building a new network of sorting centers to improve the efficiency and speed of moving items over long distances) – all in about two years.
This is not an easy task, and thousands of Amazon employees work hard for it. However, with the speed and scale of this change, a great deal of optimization is required to achieve the desired productivity. Over the past few months, we have carefully reviewed every process path in fulfillment centers and transportation networks, and redesigned numerous processes and mechanisms to continue to increase productivity and reduce costs over the past few quarters. There is more work to be done, but we are satisfied with the progress that has been made and the room for progress in the future.
We are also using this opportunity to make greater structural changes that deliver lower costs and faster speeds for many years to come. A good example is a reassessment of how the U.S. Compliance Network is organized. Until recently, Amazon operated a nationwide U.S. fulfillment network, distributing inventory from fulfillment centers spread across the country. If the local fulfillment center does not have the product ordered by the customer, we will have to ship from the rest of the country, which will increase the cost and extend the delivery time.
This challenge has become even more prominent as our fulfillment network has expanded to hundreds of additional locations over the past few years, with more locations used to distribute inventory and increasing complexity connecting fulfillment centers and delivery station nodes. Last year, we began redesigning our inventory placement strategy and moving from a national fulfillment network to a regionalized network model with a larger fulfillment center size. We made significant internal changes (e.g., inventory placement and logistics software, processes, physical operations) to create eight interconnected fulfillment network areas within smaller geographic divisions. Each region has a wide range of relevant options that can be largely self-sufficient, while still allowing for nationwide transportation and inventory deployment when necessary.
Some of the most meaningful and difficult jobs are optimizing the connections between these massive infrastructures. We're also constantly improving our advanced machine learning algorithms to better anticipate the needs of our customers across geographies so we can have the right inventory in the right region at the right time. We recently completed this regional rollout and are pleased with the early results. Shorter transport distances mean lower service costs, less environmental impact and faster receipt of orders by customers.
In terms of faster delivery, we're excited to see more next-day and same-day delivery, and we are on track to achieve the fastest Prime member delivery speeds in 2023. Overall, we are confident in reducing costs, reducing lead times, and building a larger retail business with healthy operating margins.
AWS will have annual revenue of $85 billion and is still in the early adoption phase, but remains focused on what matters most to customers in the long term at a critical time. AWS revenue grew 29% year-over-year in 2022 from $62 billion, with current macroeconomic challenges putting the business in short-term headwinds as enterprise customers become more cautious about spending on cloud computing.
While some companies may be obsessed with how to get as much money out of their clients as possible during these stressful times, this is neither what the client wants nor is it in the long-term best interest of the client, so we take a different approach. One of the big advantages of AWS and cloud computing is that you can continue to scale as your business grows; Conversely, if your business shrinks, you can choose to return a portion of the capacity to Amazon and stop paying for idle capacity.
This elasticity is unique to cloud computing and doesn't exist when you've already made expensive capital investments in on-premises data centers, servers, and network equipment. At AWS and all of our businesses, we don't optimize for any one quarter or year. We try to build customer relationships (and business) that go above and beyond; As a result, the AWS sales and support team spends a lot of time helping customers optimize AWS spend so they can better cope with the uncertain economic environment.
Many AWS customers tell us that they're not cutting costs, they're optimizing costs so they can apply resources to the innovative customer experiences they're planning. Customers appreciate AWS's customer-centric, long-term strategy-oriented approach, which we believe will benefit both customers and AWS.
While short-term headwinds have slowed AWS's growth, we are very happy with AWS's fundamentals. Our new customer base is huge and our activity is very active. Many companies will take advantage of such uncertainties to take a step back and decide what they want to change strategically. We see that more and more enterprises are choosing to forgo managing their own infrastructure and are more likely to move to AWS to enjoy the benefits of agility, innovation, cost-effectiveness, and security. Most importantly for customers, AWS continues to innovate quickly (more than 3,300 new features and services were introduced in 2022) and invests in long-term creations that change the future.
Chip development is a good example. In last year's shareholder letter, I mentioned that Amazon was investing in Graviton, a general-purpose CPU processor. Graviton 2-based compute instances deliver 40% better price performance than the latest generation of x86-based instances; And in 2022, we launched the Graviton 3 chip, which is 25% more powerful than the Graviton 2 processor. In addition, as machine learning adoption continues to accelerate, customers are craving lower-cost GPUs (the chips most commonly used for machine learning).
AWS began investing in specialized chips for machine learning training and inference (inference is predictions or answers provided by machine learning models) for years. We launched our first training chip (Trainium) in 2022; For most common machine learning models, Trainium-based compute instances are 140% faster and cost nearly 70% lower than GPU-based instances. Most companies are still in the training phase, but as they develop models that can go into mass production, they will find that most of the cost is in the inference phase, because the models are trained on a regular basis, and inference occurs continuously as the relevant applications run.
We launched our first inference chip in 2019 and have already saved companies like Amazon hundreds of millions of dollars in capital expenditure. The just-introduced Inferentia 2 chip delivered four times higher throughput and ten times lower latency than the first Inferentia processor. With the potential for huge growth in machine learning, customers will be able to accomplish more with AWS training and inference chips at a significantly lower cost. Our innovation efforts in this area are not over, and this long-term investment should yield beneficial results for both customers and AWS. AWS is still in the early stages of development and has the opportunity to achieve extraordinary growth over the next decade.
There is also great potential for Amazon's advertising business to be unique to brands, which is one of the reasons for its continued rapid growth. Similar to brick-and-mortar retailers' advertising business of selling shelf space, end-of-line display, and exposure in brochures, our sponsored products and branded services have been an integral part of the Amazon shopping experience for more than a decade. Unlike brick-and-mortar retailers, however, Amazon can tailor these sponsored products to make them relevant to what customers are searching for, based on our knowledge of shopping behavior and deep investment in machine learning algorithms. This makes advertising more efficient. That's why Amazon's ad revenue continued to grow rapidly over the past few quarters while growth slowed for most large advertising-focused enterprises (up 23% year-over-year in Q4 2022 and 25% year-over-year on $31 billion in full-year 2022 revenue).
We strive to be the best place for advertisers to build brand identity. We have a medium to long term opportunity to help achieve this mission. We will continue to invest heavily in machine learning to continuously optimize ad selection algorithms. Over the past few years, we've invested in comprehensive, flexible and long-lasting advertising planning and measurement solutions that give marketers a deeper understanding of ad performance.
One example is Amazon Marketing Cloud ("AMC"). AMC provides a secure digital environment where advertisers can run Custom Audiences and ad effectiveness analytics with privacy and security, using multiple first- and third-party inputs to generate advertising and business insights to guide their broader marketing and sales strategies. Advertising and AWS teams enable companies to store data in AWS, operate securely in AMC, use data sources from Amazon and other third parties, use AWS for analysis, and display ads on Amazon or third-party publishers through the Amazon demand-side platform. Customers really like this synergy. We also see future opportunities to integrate advertising into Amazon's video, live sports, audio, and grocery products in an orderly manner. We will continue to work hard to help brands create unique connections with the right audiences and grow this part of our business.
In turbulent times, it's easy to focus only on the big business that exists, but to build a sustainable, long-lasting, growing, and in many ways helpful to customers, we can't stop innovating and investing in long-term customer experiences that will have a meaningful impact on customers and Amazon itself.
When we look for new investment opportunities, we ask ourselves a few questions:
If we succeed, can this opportunity become a big project with a reasonable return on investment? Is this opportunity currently working well? Are there different ways we can realize this opportunity? Are we capable of this area? If not, can we quickly acquire this capability?
If we are satisfied with the answers to these questions, we invest. This process has led to some seemingly simple expansions, as well as expansions that may not have been promising at first.
For example, in 1998-1999, Yama chose to expand from selling only books to include categories such as music, video, electronics, and toys. At the time, this strategy was not generally favorable, but in retrospect, it seems like a clear path forward.
The same words can be used to describe the expansion of our international store business. In 2022, our international consumer business generated $118 billion in revenue. In larger, more mature international consumer businesses, we are larger enough to be more vulnerable to macroeconomic slowdowns; However, the growth in 2019-2021 is significant, with a compound annual growth rate ("CAGR") of 30% in the UK, 26% in Germany and 21% in Japan (excluding currency effects). Over the past few years, we have continuously invested in opening up new international markets, including India, Brazil, Mexico, Australia, many European countries, the Middle East and parts of Africa. These emerging markets require some fixed investment to launch and expand, but their growth trajectories are good, with growth patterns similar to what we see in North America and other already mature international markets.
Emerging countries sometimes lack some of the infrastructure and services (e.g. payment methods, transportation services, and internet/telecommunications infrastructure) on which our business depends. To address these challenges, we continue to work with a variety of partners to provide solutions to our customers. We believe that investing in international markets will enable us to help more global customers and build a larger consumer business that generates free cash flow."
In addition to expanding into international markets, we are also working to enhance the existing retail market for larger and unique products. In the U.S. alone, the overall grocery market is $800 billion, with the average household shopping three to four times a week. Amazon has built a rather unusual but significant grocery business over the course of nearly 20 years.
Similar to how other big retailers entered the grocery market in the '80s, we started by adding products on Amazon that supermarkets sold and didn't require temperature control, such as paper products, canned and boxed foods, candy and treats, pet care, health and personal care, and beauty supplies. But while a typical supermarket typically offers only 30,000 items, Amazon offers more than 3 million. Given the current high cost of online fulfillment, we are also focusing on larger packaging sizes.
We are pleased with the size and growth of Amazon's grocery business, but still hope to be able to meet more of our customers' grocery needs. This requires expanding the physical store space, as most grocery shopping still happens offline. Whole Foods pioneered the concept of a natural and organic specialty food store 40 years ago, and it's still a growing business that raises the bar for healthy and sustainable food, and Amazon has continued to invest in this business over the past year, while also making changes to improve profitability. Whole Foods is moving in an encouraging direction, but to have a bigger impact on the offline grocery business, we must find a form of mass grocery that deserves broad expansion. Amazon Fresh is a brand we've been experimenting with for the past few years, working to identify and build a large grocery format that works for Amazon's scale. Groceries is a huge growth opportunity for Amazon.
Amazon Business, a sourcing and wholesale buying solution, demonstrates how our investment in e-commerce and logistics capabilities enables us to pursue success in this huge market segment. Amazon Business makes it easy for businesses, local governments, and organizations to purchase office supplies and other bulk commodities at great prices. Despite the poor performance of some sectors of the economy over the past few years, Amazon Business has prospered. Because the Amazon team translates the meaning of providing choice, value, and convenience into enterprise procurement, constantly listening and learning from customer needs, and innovating based on the interests of customers.
While some people have never heard of Amazon Business, our enterprise customers love it. Launched in 2015, Amazon Business now has annual revenues of about $35 billion, and more than 6 million active customers, including 96 of the Global 100, are using a wide selection of one-stop shopping, real-time analytics, and hundreds of millions of office supplies. We believe that the full potential of Amazon Business is far from being realized and plan to continue to develop the features that commercial customers need and want.
While many brands and merchants successfully sell products on the Amazon marketplace, there are also many brands and sellers who have launched their own dropshipping sites and are challenged to convert views into purchases. We're introducing Buy With Prime to help address this challenge. This feature allows third-party brands and sellers to offer products on their own websites to a large number of Amazon Prime paying members, and offers fast, free shipping, and Prime members can check out directly with their Amazon account.
This feature gives sellers several additional benefits, including Amazon handling the storage, picking, packing, shipping, payment, and returns of products, all through Amazon Pay and FBA. Buy with Prime is available to all U.S. merchants, increasing conversion rates on third-party shopping sites by an average of 25%. Sellers are excited about converting more sales and shipping easier, and Prime members love being able to use Prime benefits on more third-party sites and improve the shopping experience on more sites.
Expanding into international markets, tapping into large retail areas that Amazon has yet to tap into, and leveraging our unique resources to help merchants sell more effectively on our own websites are all natural ways to expand. Of course, we are also making some investments that are further away from our core business, mainly because we see unique opportunities in these areas. Like AWS did in 2003, Amazon Healthcare and Kuiper could be potentially similar products by 2023.
Our initial efforts in Amazon Healthcare began in the pharmacy business, which seems not far from the main business of e-commerce. Customers have been asking us for years when we will be able to offer online pharmacies because they are frustrated with current providers. Launched in 2020, Amazon Pharmacy is a full-service online pharmacy that offers Prime members transparent pricing, convenient regular renewals, and preferential pricing. This business is growing rapidly and continues to innovate.
Amazon Pharmacy, for example, recently launched RxPass, which gives Prime members access to prescription medications for many common conditions, such as high blood pressure, acid reflux, and anxiety, for a flat fee of $5 per month. However, our customers are still pressing for a better alternative from Amazon to improve the inefficient and unsatisfactory overall healthcare experience.
We decided to start with primary care, where patients have the most direct contact. After extensively evaluating and studying the existing situation, including early-stage experimental services such as Amazon Care, we believe One Medical's patient-centric experience is excellent and can lay the foundation for our future business and announced the acquisition of One Medical in July 2022. One Medical's digital app makes it easy for patients to discuss issues with healthcare practitioners via text or video, and if a physical exam is needed, One Medical has offices in cities across the U.S. where patients can book same-day or next-day exams. One Medical has relationships with specialists in each city where the office is located and works closely with the local hospital system to provide quick access to these resources when needed. We strongly believe that One Medical and Amazon will continue to innovate together to change the face of primary care."
Kuiper also exemplifies Amazon's long-term innovation in areas where customers are in high demand, with a vision to create a low-orbit satellite system to serve regions around the world that currently do not have access to high-quality broadband internet service. There are hundreds of millions of homes and businesses that do not have reliable access to the Internet. If they have a reliable connection, they will be able to learn online, use financial services, start a business, shop, have fun, and businesses and governments can improve their reach, efficiency, and operations. Our team has developed low-cost antennas (i.e. customer terminals), which will lower the barrier to entry. We recently released new terminals for communicating with satellites flying overhead, and we expect the standard residential version to cost less than $400 each.
They're small: 11-inch square, 1-inch thick, and weigh less than 5 pounds without mounting brackets, but they transfer at speeds of up to 400 megabits per second and are powered by Amazon-designed baseband chips. We are preparing to launch two prototype satellites this year to test the entire end-to-end communications network, with a beta launch to commercial customers in 2024. Customer feedback has been very positive so far, and we think Kuiper represents a huge potential opportunity for Amazon. It has several similarities with AWS in that it requires a large capital investment at the beginning, but has a large base of potential consumers, enterprises, and government customers, has significant revenue and profitability potential, and only a few companies have the necessary technology, innovation, and investment capabilities.
The last area of investment I want to mention, and at the heart of what will enable Amazon to innovate in every business area for decades to come, are the large language models ("LLMs") and generative AI that we're investing in.
Machine learning has been seen as a high-potential technology for decades, and it is only in the last five to ten years that it has begun to be widely used by companies. This change is driven by several factors, including the ability to acquire higher compute capacity at a lower price than ever before. Amazon has used machine learning extensively for the past 25 years, applying it to personalized e-commerce recommendations, picking paths in fulfillment centers, Prime Air drones, Alexa voice assistants, and many of the machine learning services offered by AWS (AWS is the vendor with the broadest range of machine learning capabilities and customer base of any cloud vendor).
Recently, a new form of machine learning, called generative artificial intelligence, has emerged and is expected to significantly accelerate the adoption of machine learning. Generative AI is based on very large language models (trained on tens of billions of parameters and still growing) with an extremely broad range of recall and learning capabilities. We've been working on our own LLMs for some time now and believe they will transform and improve nearly every customer experience, and will continue to invest heavily in these models across all of our consumer, seller, brand and creator experiences. In addition, as with the path AWS has been taking for years, we are pushing companies of all sizes to take advantage of generative AI. AWS offers the most cost-effective machine learning chips, Trainium and Inferentia, making it affordable for both small and large companies to train and run their LLMs in production.
Amazon's business customers can choose from a variety of LLMs and build applications using all the AWS security, privacy, and other features that customers prefer. In addition, we are offering applications like AWS CodeWhisperer, which revolutionizes developer productivity by generating code suggestions in real time. I have a lot more to say about LLMs and generative AI because I think they will be transformative, but I'll leave that to future shareholder letters. Suffice it to say that LLMs and generative AI mean a lot to customers, shareholders and Amazon.
All in all, I am optimistic that Amazon will successfully weather this challenging macroeconomic environment and rise stronger. I've already mentioned many reasons for optimism, but two relatively simple statistics highlight our huge future opportunities.
While our consumer business generated $434 billion in revenue in 2022, the vast majority of our total global retail market share is still in physical stores (approximately 80%). The same is true for global IT spending, where AWS generated $8 billion in revenue in 2022, but 90% of global IT spending is still in on-premises data centers and has not yet migrated to the cloud.
As things change (and are shifting), we believe Amazon's leading customer experience, relentless innovation, customer focus, and hard work will deliver significant growth in the coming years. Of course, that's not counting the other businesses and experiences we're pursuing at Amazon, which are still in their early stages.
I firmly believe our best days lie ahead, and I look forward to working with the team at Amazon to make it a reality."
Amazon President and CEO Andy Jassy
![The whole army of giants is gone: the report card of the opening year, Google, Amazon are all "unqualified"[fig]](data:image/gif;base64,R0lGODlhAQABAIAAAP///wAAACwAAAAAAQABAAACAkQBADs=)