Welcome to the Amazon.com annual meeting of shareholders. I'm Susan Xiong, Vice President, Associate General Counsel, and Corporate Secretary. I now call the formal portion of this meeting to order. Today's agenda covers our items of business and rules of procedure. Please note that in the event an unanticipated issue prevents us from being able to continue this annual meeting, we will post updated meeting information on our investor relations website. After the proposals have been presented, we will have a presentation by Brian Olsavsky, our CFO. Angie Quino, Director of Financial Communications, will then moderate the Q&A session with Andy Jassy, our CEO. Angie, can you provide details on how to submit a question?
Thanks, Susan. If you wish to submit a question and have not already done so prior to the meeting, please type your question into the Ask a Question field on the website you used to access this meeting, and then click Submit. Out of respect for your fellow shareholders, we ask that each person please be concise and limit themselves to one question.
Thanks, Angie. It's now my pleasure to introduce our directors who are attending this meeting, including those joining us remotely. Jeff Bezos, Andy Jassy, Keith Alexander, Edith Cooper, Jamie Gorelick, Daniel Huttenlocher, Andrew Y. Ng, Indra K. Nooyi, Jonathan J. Rubinstein, Brad D. Smith, Patty Stonesifer, and Wendell P. Weeks. Also joining us remotely today are representatives of our auditors, Ernst & Young. The polls open for voting on all matters at the beginning of the meeting, and we will close the polls after presentation of today's proposals. The proposals to be voted on today are set forth in our proxy statement, and we have been notified that a shareholder intends to present a non-binding proposal during the meeting as described in the proxy statement. Instructions for voting on that proposal, if properly introduced, will be provided during the meeting. No other nominations or proposals will be introduced.
If you wish to vote during the meeting, please follow the instructions on the meeting website before the polls close. If you have already voted in advance of the meeting, you do not need to vote again unless you requested a legal proxy, or wish to change your vote, or wish to vote on any other matter properly presented at the meeting. A representative of Broadridge was appointed our Inspector of Elections, has taken the required oath, and has certified that notice of this meeting was mailed beginning on April 9, 2026 to all shareholders of record as of the record date, and that a majority of our common stock is present or represented by proxies. Therefore, a quorum exists for the meeting.
I'll now introduce the company's proposals, which are the election of directors, the ratification of the appointment of Ernst & Young as independent auditors for fiscal 2026, and an advisory vote to approve executive compensation. We have five shareholder proposals to be voted on if properly presented. To ensure that we have adequate time for our Q&A session later in the meeting, each proponent will have two minutes to present their proposal. If a proponent goes beyond two minutes, we will need to place the line on mute so that we can continue the meeting. Thank you in advance for your understanding. Stefan Padfield will now introduce proposal number four, requesting a report on charitable partnerships. Mr. Padfield has pre-recorded the following statement.
My name is Stefan Padfield. I am a principal at the Free Enterprise Initiative, which is part of The Heritage Foundation. The Heritage Foundation is the proponent of item 4, which requests a report on the risks of Amazon's use of diagnostic tools created by politicized corporate partners, particularly the Southern Poverty Law Center. This proposal was submitted before the SPLC was indicted on 11 counts of wire fraud, false statements to a federally insured bank, and conspiracy to commit concealment of money laundering. Yet, as made clear in our proposal, no one should be surprised by this indictment, which in a very real sense merely confirms longstanding suspicions that the SPLC has apparently become an anti-conservative smear factory, endangering lives and fomenting the very hate it claims to be combating.
The board's opposition statement is most notable for its glaring failure to address whether Amazon continues to use SPLC data, which is a question expressly raised in our proposal. That omission arguably constitutes a red flag in terms of current oversight. More broadly, the proposition that shareholders should simply trust Amazon's status quo is belied by the fact that Amazon has been rated a high risk by the 1792 Exchange on its corporate bias ratings, including specifically for concerns about ongoing ties to the SPLC, as well as a corporate leadership that apparently gives 35 times more to Democrat than Republican causes. Amazon also scored a mere 6% on Alliance Defending Freedom's Viewpoint Diversity Score Business Index. One could be forgiven for picking up an anti-conservative theme here, and that raises the specter of an echo chamber that precludes the fully informed decision-making required by fiduciary duties.
Accordingly, regardless of the outcome of this proposal, The Heritage Foundation remains committed to good faith engagement with Amazon in order to provide some perhaps much needed viewpoint diversity.
Thank you. Kelly Poole will now introduce proposal number five, requesting additional reporting on impact of data centers on climate commitments. Operator, please open the line for Ms. Poole.
Good morning, shareholders and members of the board. Proposal number 5 asks Amazon to issue a report explaining how it will meet the climate-related commitments it has made, given the massive projected energy demand from AI and the data centers Amazon is planning to build. Amazon is rapidly expanding its data center footprint to maintain an edge in the AI race. The scale and speed of its build-out is placing significant strain on the power grid. To meet surging demand from Amazon and its peers, utilities are increasingly turning to fossil fuels, building new methane gas plants and delaying coal retirements at a scale that risks disrupting the clean energy transition. The impact of these investments is already becoming clear.
Between 2022 and 2024, Amazon's emissions from electricity increased by nearly 40%, raising serious questions about whether the company will meet its existing climate commitments while rapidly scaling AI infrastructure. Although Amazon discloses certain sustainability initiatives related to its data centers, investors still lack a clear understanding of how these efforts add up to a credible net zero pathway. Investors see growing opportunities to power data centers without locking in new fossil fuel dependence. Advances in clean energy storage, grid modernization, and flexible load management offer pathways to support AI growth while maintaining progress toward climate goals. Given Amazon's rising emissions profile, investors seek a forward-looking roadmap for how the company plans to expand its data center footprint while credibly maintaining its climate commitments. Thank you.
Thank you. Paul Chesser will now introduce proposal number 6, requesting a report on impact of climate commitments. Mr. Chesser has pre-recorded the following statement.
I'm Paul Chesser of National Legal and Policy Center, sponsor of item 6, requesting a report on the impact of Amazon's climate commitments. Amazon just reported another strong quarter, $181.5 billion in revenue, with Amazon Web Services growing at its fastest rate in several years. Buried in the same earnings release is a number that should concern every shareholder. Free cash flow, the actual cash Amazon generates after spending, has fallen dramatically by as much as 95%. The primary reason? Capital expenditures have surged to extraordinary levels, driven in large part by AI and data center infrastructure. Amazon has told investors that it expects capital expenditures to approach $200 billion in 2026. Every dollar of that investment depends on reliable, affordable energy. Amazon has never told its shareholders what its climate commitments cost relative to instead purchasing the most affordable power available in each market.
The Climate Pledge was made in 2019. Since then, Amazon has become the world's largest corporate buyer of renewable energy, a commitment with real costs attached to it. In March, Amazon's leadership signed a pledge at the White House acknowledging that its data center energy obligations are real and must be accounted for. Amazon acknowledged that to the federal government, but it has not made it to the people who own this company. Item 6 asks for something straightforward, a report disclosing what The Climate Pledge actually costs and whether the board has reconsidered it given how dramatically circumstances have changed. That is not an unreasonable request from shareholders. It is a fiduciary one. I urge you to vote for item 6. Thank you.
Thank you. Brandon Rees will now introduce proposal number 7, requesting a mandatory independent board chair policy. Mr. Rees has pre-recorded the following statement.
Hello, my name is Brandon Rees, and I am here today on behalf of the AFL-CIO Reserve Fund. I hereby introduce shareholder proposal 7 that urges the board of directors to adopt an independent board chair policy. We believe that appointing an independent board chair will strengthen the independent leadership of the board of directors and enhance management accountability to shareholders. According to The Wall Street Journal, a growing percentage of S&P 500 index companies have appointed independent board chairs. In 2025, 42% of S&P 500 index company boards were chaired by an independent director, compared to just 29% a decade ago. Numerous institutional investors support having an independent chair as a simple, good governance practice. For example, the Council of Institutional Investors and the California Public Employees' Retirement System both have stated that the board should be chaired by an independent director.
We believe that appointing an independent board chair is particularly important in light of the various related party transactions that our company's current board chair, Jeff Bezos, has with Amazon. Our company has entered into contracts with Blue Origin and The Washington Post. Both companies are owned by Jeff Bezos. Mr. Bezos is also the co-CEO of Project Prometheus, an AI company which could be a potential competitor or a business partner with Amazon. An independent board chair will help ensure that any potential conflicts of interest between Amazon and Mr. Bezos's various other companies are managed appropriately. For these reasons, we urge you to vote for proposal number 7, requesting the adoption of an independent board chair policy. Thank you.
Thank you. Joan Morris will now introduce a non-binding proposal requesting that the company establish and maintain a worker-oriented AI advisory council. Shareholders can vote on this proposal if properly introduced by typing for, against, or abstain in the text box that appears on the meeting website under the caption Other Business and clicking the Submit Button. Ms. Morris has pre-recorded the following statement.
My name is Joan. I work as a warehouse associate at Amazon's ATL6 facility in East Point, Georgia, and today I am speaking on behalf of United For Respect co-executive director Bianca Agustin, whose floor resolution I urge you to support. Please vote yes to require Amazon to create an AI advisory council composed of hourly associates and frontline workers. Shareholders request that Amazon establish and maintain an independent artificial intelligence and automation advisory council to provide formal oversight, review, and recommendations regarding development and deployment of artificial intelligence, robotics, algorithmic management systems, and automated decision-making technologies across Amazon operations, including but not limited to systems used in hiring, scheduling, discipline, promotion, compensation, performance evaluation, and termination decisions.
The council shall be composed primarily of hourly warehouse associates and frontline workers from Amazon operations, alongside recognized worker advocates and independent experts in civil rights, labor standards, occupational safety, and technology ethics. I present this proposal as a warehouse associate who is deeply concerned about Amazon's decision to go full throttle with AI without any consideration of its impacts on its workers. Amazon owes its employees transparency about where the company is headed. When we ask management about the changes we've seen, we don't get straight answers, and this dishonesty has created a sea of confusion and disconnect. Here's the truth. AI and robotics fail all the time. You still need humans to get the job done. We keep things moving. Your workers are your backbone of this company. This race towards AI and automation is a short-term solution and a costly one.
This floor proposal is a straightforward fix. I urge you to vote yes on the floor proposal to create an AI council. Thank you.
Thank you. The company recommends that shareholders vote against the shareholder proposal. That concludes the presentation of the proposals. The proxy statement for this meeting explains the reasons for the company's recommendation against each of the shareholder proposals appearing in the proxy statement. We have received some questions regarding the proposals. Thank you for these questions. In response, I invite you to review the board's statement and recommendation as applicable on each of the proposals as set forth in the company's proxy statement. The polls are now closed on all proposals, and the formal portion of this meeting is now adjourned. Based on preliminary voting results, each nominee for director received a majority of the votes cast for such nominee's election, so all 11 nominees have been duly elected.
The ratification of the appointment of Ernst & Young and the advisory vote to approve our executive compensation have each been approved by the requisite vote. None of the shareholder proposals has been approved by the requisite vote. I would now like to introduce Brian Olsavsky, who will give a financial update after a short video.
As businesses succeed and get larger, they sometimes forget how things got started. It's easy to not want to take risks, to not want to wreck what you've already built. You cannot achieve something extraordinary for customers by playing not to lose. The reality is the world is changing so fast around us. Technologies like generative and agentic AI are rare. They come about once in a lifetime and completely change what's possible for customers and businesses. When you have a big shift like that, you have 2 choices. You can decide to embrace it and help shape it, or you can wish it away and have it shape you. We are going to embrace it. In virtually every corner of our company, we're using AI to make customers' lives better and easier. The progress we're making is evident. We're still at the relative beginning.
We are not close to being done inventing. What we have a chance to do together the next 10 years is very unusual, and the single biggest difference maker will be our ability and willingness to learn. We're eager to learn what customers like, even if it's wildly different or unexpected to us. You have to constantly ask, what are the next big breakthroughs for customers? What is a customer problem that needs solving? Customers are always looking for something better, and we spend a lot of time identifying how to unlock these experiences for them. I think working at a place whose mission is to make customers' lives better and easier every day is very inspiring. Not just saying it, but doing it. Not running away the first time you see any lack of traction or any problems.
It's very rarely that you catch lightning in a bottle where you have an idea, launch it in a few days, and it's a home run. It takes a lot of hard work, and everything that's worth doing takes persistence and resilience. I've never encountered a more intelligent, creative, ambitious, missionary group of teammates than we have at Amazon. For builders who want to change the world and who have fire in their belly, it's challenging to find a company where you can make a bigger impact on the world than you can here at Amazon. We don't always get everything right, and we learn and iterate like crazy. We constantly choose to prioritize customers, delivery, invention, ownership, speed, scrappiness, curiosity, to build a company that outlasts us all.
Hello, everyone. I'm Brian Olsavsky, Amazon's chief financial officer. Today, I will share a recap of our 2025 financial results. First, let me highlight some of the key things we did for customers during the year. Let's start with our stores business. We remain focused on the inputs that matter most to our customers: broad selection, fast, convenient shipping, and low prices. In 2025, we expanded selection across a range of price points, including Everyday Essentials, more than 400 new beauty brands in the U.S., and new fashion brands like Michael Kors, Nike, and The North Face. Our millions of global third-party sellers are important contributors to our vast selection, which helps customers find the items they need at competitive prices. In 2025, we significantly expanded selection by making same-day perishable grocery delivery available in more than 2,300 U.S. cities and towns.
In areas where this service is available, perishables made up 9 of the top 10 most ordered items. Our speed of delivery keeps getting better, and in 2025, we delivered to Prime members at our fastest speeds ever, delivering over 13 billion units the same or next day around the world. Same-day was our fastest growing delivery offering, and nearly 100 million customers used it last year in the U.S. alone. Our speed improvements were driven by the expansion of our same-day delivery network and improved inventory placement, which helps lower transportation distances and results in lower shipping costs. We work hard to be the place that customers rely on for sharp pricing. In 2025, for the 9th year in a row, Amazon was named the lowest priced U.S. retailer by Profitero, with online prices on average 14% lower than other major U.S. retailers.
In addition to our low everyday prices, we held shopping events throughout the year offering great deals for our customers, including our Big Spring Sale in March, Prime Day in July, Prime Big Deal Days in October, and our Black Friday and Cyber Monday shopping event. Customers saved billions of dollars across millions of deals during these events. In AWS, we saw accelerated revenue growth across both our core cloud services and AI offerings. Our industry-leading cloud helps make AWS the best place for startups, enterprises, and governments to modernize their infrastructure and harness the power of AI. Customers want to run their core and AI workloads on AWS, given its stronger functionality, security, and operational performance. In 2025, we brought on more data center capacity than any other company, adding 3.9 GW of power to support growth and customer demand.
As part of this capacity, we launched Project Rainier, an AI compute cluster containing nearly 500,000 of our custom Trainium2 chips, used to build and deploy Anthropic's leading Claude AI models. In December, we launched Trainium3, which is up to 40% more price performant than Trainium2. Now let's turn to our financial results. In 2025, we had strong improvement across our key financial metrics. Full-year worldwide net sales were $716.9 billion, up 12% year-over-year. Turning to our segment financial results, our North America segment sales grew to $426.3 billion, up 10% year-over-year. Customers appreciated the convenience of receiving their items quickly, from gifts for family and friends to Everyday Essentials and perishable groceries. In 2025, Everyday Essentials grew nearly twice as fast as all other categories in the U.S., representing one out of every three units sold in our store.
International Segment sales grew to $161.9 billion, up 10% year-over-year, excluding changes in foreign exchange rates. We're enthusiastic about the business we're building internationally. We now have over 20 countries at varying degrees of growth and investment, including our launch of Ireland in 2025. Improvements in delivery speed and selection have also resonated with customers internationally and are contributing to our growth. Across both our North America and International Segments, there were common elements to our growth in 2025. We continue to expand our Prime benefits with digital content like live sports, faster shipping speeds, and more. This drives strong growth in Prime membership worldwide. Third-party sellers remain a key contributor to our store offerings. In 2025, sellers represented 61% of all units sold on Amazon. Independent sellers help us to increase our selection for customers. We work hard to help sellers grow their businesses on Amazon.
Finally, advertising services grew at a rapid pace, reaching $69 billion in net sales for 2025, an increase of 22% year-over-year. AWS segment sales grew to $129 billion, up 20% year-over-year, driven by ongoing cloud migrations and high demand for AI services. We are pleased with our recent growth as we further expand our generative AI offerings, and we are encouraged by the response from our customers. In 2025, we saw increased operating income across all three of our segments. Full-year worldwide operating income was $80 billion, up from $68.6 billion in 2024. This was the result of hard work and discipline across all of our teams. There are three primary drivers for this increase. First, our top-line revenue growth accelerated in 2025, which drives operating income dollars.
Second, in Stores, we've reduced our global cost to serve on a per-unit basis for the 3rd year in a row, driven by improving inventory placement across our regionalized fulfillment network and reducing distance traveled. Third, we grew our fixed cost at a slower rate than our revenue growth. We are pleased with the increase in profitability we saw in 2025, and we are confident that there are opportunities to grow our operating income in the future as we continue to deliver for our customers. In 2025, we invested $128.3 billion in cash capital expenditures. This investment was primarily driven by AWS to support generative AI, but also reflects investments to meet customer demand for core cloud services and our Stores business. The majority of this CapEx was for technology infrastructure, data centers, chips, and servers.
We purchase a lot of chips from third parties, and we also produce our own custom chips, like Trainium and Graviton. We also invested in capacity for our operations network to support future growth, accelerate our delivery speeds, and lower our cost to serve. Free cash flow remains our most important long-term financial metric. In 2025, operating cash flow was $140 billion, up $24 billion year-over-year, and driven primarily by growth in operating income dollars. Free cash flow was $11 billion. This is down from $38 billion in 2024, largely reflecting the $50 billion year-over-year increase in cash capital expenditures that we saw in 2025, as we build the infrastructure to support the next phase of Amazon's growth. As we've done throughout our history, we invest to meet customer demand, work hard to execute well, and generate strong free cash flow and long-term returns on these investments.
As we've turned the calendar to 2026, we are excited to build upon the great work the teams were able to deliver in 2025. We will work hard to remain the place customers trust for sharp prices, broad selection, and customer convenience. We believe putting customers first is the only reliable way to create lasting value for our shareholders. Thank you.
Thank you, Brian. I would now like to introduce our CEO, Andy Jassy, who is live with us to make a few opening remarks and address questions we have received pursuant to the meeting rules of procedure.
Hi, I'm Andy Jassy, and thank you for taking the time to join us. 2025 was a strong year for Amazon, and that momentum is continuing in 2026 across all of our businesses. Our Stores business continued to grow in 2025, and in Q1 2026, we saw the highest growth since the tail end of COVID lockdowns. Our grocery business, with more than $150 billion in gross sales in 2025, makes us the second largest grocer in the U.S., with fast-rising perishable sales. Alexa for Shopping, our agentic shopping assistant, has seen monthly active users up 115% year-over-year in March, and engagement up nearly 400% year-over-year as it helps our customers quickly discover what they want in the store. AWS continued to see accelerated growth throughout 2025 and closed the year with $129 billion in revenue.
As enterprises continue to move to the cloud and increasingly rely on AWS for AI. In Q1 2026, AWS grew 28% year-over-year, its fastest growth rate in 15 quarters, achieving a revenue run rate of $150 billion. Growing 28% on $150 billion annual run rate is not simple to do. That's a lot of absolute revenue growth. We're particularly excited about what we're seeing in AI. We've never seen a technology grow as rapidly as AI. We felt like AWS grew quickly its first three years as a business and had a $58 million annual revenue run rate. In the first three years of this AI wave, AWS' AI revenue run rate grew to over $15 billion in Q1 2026, nearly 260 times larger than AWS was three years after it launched. Amazon is already a significant leader in AI and growing rapidly.
Customers want to run their AI in AWS, and this AI growth is also driving substantial growth in our non-AI or core AWS business as well. Our chips business continues to grow rapidly, with an annual revenue run rate of $20 billion through Q1 2026 and growing triple-digit percentages year-over-year. If we were to sell the chips we will produce this year in racks to AWS and other customers, as other leading chip companies do, our run rate would be $50 billion. We're very pleased with what we're seeing in Amazon Ads, which grew 22% year-over-year in 2025.
The same goes for Prime Video, which continues to produce compelling content like our recent movie release, "Project Hail Mary," which has now crested $665 million at the box office already, as well as our many popular live sports offerings, which continue to set records for viewership. Alexa+ has expanded to millions more Prime members across 4 new countries, and customers are talking to Alexa twice as much and completing purchases 3 times more than with classic Alexa.
I'm very excited about Amazon Leo, which keeps gaining momentum with commercial service on track to launch in a few months, and we already have a meaningful revenue set of commitments from enterprises and governments, including Delta Air Lines, JetBlue, AT&T, Vodafone, DIRECTV Latin America, Australia's National Broadband Network, DP World Tour, NASA, and others. We also announced that we plan to acquire Globalstar, which will expand Leo's satellite network with a direct-to-device capability. We entered an agreement with Apple for Amazon Leo to power satellite devices for iPhones and Apple Watches. I've never been more optimistic about what's ahead, and I want to thank our teammates. Their accomplishments represent a lot of invention, hard work, and thoughtful execution across Amazon, and they're making so many customers' lives better and easier every day.
The team remains focused on our customers, are innovating at a fast clip, and leveraging AI to improve, and in many cases, reimagine all of our customer experiences. I also want to thank our customers. Maintaining your trust and delivering for you is our mission, and it is at the center of everything we do. Thanks also to our independent sellers, who are such important partners in helping us deliver great selection at low prices. I met with several amazing sellers at our annual Accelerate gathering this past year. It is inspiring to see what they're doing for customers. I'd like to recognize our AWS partners and customers who are working side by side with us to harness AI to change every customer experience we know and set out to create new experiences that we couldn't have even imagined just a short time ago.
Thank you to the communities where we operate for being such great partners. We're working hard to help with education, housing, disaster relief, food insecurity, upskilling, and many other areas in which we can contribute our skills and expertise. Right now, we're in the middle of our global month of volunteering, and it's great to see our teammates partner in their communities. Last year, Amazonians participated in more than 500,000 volunteer activities across 55 countries, wading into murky water to help restore polluted ecosystems, joining classrooms to teach students about AI, supporting local food banks, and helping neighborhoods recover from wildfires, just to name a few.
In closing, we're in the middle of some of the biggest inflections of our lifetime, and Amazon has the culture, the know-how, and the resources to make so many customers' lives better and easier and to build multiple new long-term businesses with substantial return on invested capital and free cash flow. We will continue investing and inventing to make it so. With that, I look forward to taking your questions.
Thanks, Andy. We'll now move to our first question, which is: how is AWS differentiating itself from competitors in generative AI, and what impact do you expect this to have on medium-term growth?
Well, I'm really pleased with the growth in AWS. In this past quarter, in Q1 2026, AWS grew 28% year-over-year on $150 billion annual revenue run rate. That's the fastest growth we've had in 15 quarters, back when we were a business about half the size we are now in AWS. A lot of it is being driven by AI. As I mentioned in my opening comments, through Q1 of this past year, it's an over $15 billion annual revenue run rate. There are several reasons why people are choosing AWS for AI. I think first, we have the broadest functionality top to bottom in the AI stack. If you're looking to try and build AI models quickly and more easily, we have SageMaker AI. For very high-performance inference with the broadest selection of frontier models, we have Bedrock.
We're excited to now offer OpenAI's models in Bedrock, which we've just started doing. For most companies, the significant implementation of AI is going to end up being in agents. We have a lot of functionality for agents. To make it easier to build agents, we have a service called Strands. If you want to be able to run agentic applications securely and scalably, especially for enterprises where the data really matters, we have a service called Agent Core that's very compelling. We have a number of turnkey agents for coding. We have a service called Kiro for knowledge workers, where you're really equipping each employee with an AI assistant. We have something called Quick, which is growing very fast. For contact center, we have Connect. For migrations and technical debt, we have Transform. We really have a very broad set of services and capabilities across AI.
I think a second reason that companies are choosing AWS for AI is that as companies start to expand their use of inference and get them into their production applications, they want their inference to live close to where their data and their applications are, and just much more of that lives in AWS than anywhere else. Third, I think as people are starting to figure out, the more that AI is expanding and growing at this explosive rate, it's turning out the more that it's driving non-AI, or what we call core infrastructure. That's because so much of the post-training and the reinforcement learning and all the actions that agents take is being run on CPU. We have the broadest selection of CPU and core infrastructure offerings by large amounts.
That's another reason that as you're growing in AI, you're going to grow your core usage, and we have the broadest selection in core usage as well. Fourth, we have the strongest security and operational performance. When you think about then the medium to long-term impact of AI on the AWS business, we've been thinking for a while that AWS was on pace to be a multi-hundred billion dollar annual run rate business. With AI and the investment that we're making and the way that we believe AI will reinvent every customer experience that we know, we think our AWS business will be meaningfully larger than we'd anticipated, with strong long-term free cash flow and return on invested capital. We're just very excited about our AWS business.
Thanks, Andy. Your next question is: Are you planning to be more active in commercializing semiconductors, GPUs, chips, et cetera, to final consumers, creating an additional business line with high potential growth?
I'm very bullish about our chips business. It's a $20 billion annual revenue run rate business through Q1 of 2026, it's growing triple-digit percentages year-over-year. That somewhat masks the size of it. If you took all the chips that we'll produce in 2026, if you sold them to AWS or to other third parties, like other chip companies do, that run rate would be closer to $50 billion. We've got momentum. If you look at the two top leading AI labs in the world, Anthropic and OpenAI, have signed multi-year, multi-gigawatt commitments with AWS. We have over $225 billion of revenue commitments in the future for Trainium. People really want Trainium because it has real advantage price performance relative to others. Our second version of Trainium, Trainium2, that's largely sold out.
Our third version of Trainium, which is Trainium3, which we just started shipping for real in the early months of 2026, and is 30%-40% more price performant than Trainium2, it's also largely sold out. Trainium4, our fourth edition of Trainium, which is still about 18 months away from broad availability, has much of it already reserved. There's just very broad demand for Trainium. Then Bedrock, which is our major inference service and growing like crazy right now, with over 125,000 customers using Bedrock. The lion's share of inference in Bedrock is also on Trainium. Apart from the Trainium growth, what you're also seeing, as I mentioned earlier, is you're seeing this substantial amount of growth that AI is driving in non-AI services and core.
We have with Graviton, where we're on our fifth version of Graviton, we have the leading price performance CPU chip as well, where you have 98% of the top 1,000 AWS compute customers are expansively using Graviton already. If you look at the growth in AI and what it's driving in core AWS as well, with the leading CPU chip and the most price performant AI accelerator chip, we're very well positioned in the chip space. It's great for customers because we're able to continue to take price down for our customers. It's great for our AWS business as it saves us tens of billions of dollars at scale in annual CapEx and hundreds of basis points better on operating margin.
I do think we have the opportunity and we have the demand to sell our chips to third parties, and I think that's very much a possibility down the road.
Your next question is if you have any thought of future dividends?
Well, we always look at different ways to best leverage and invest our cash. We continue to very strongly believe that the best long-term use of our available cash for the business and for shareholders is investing in the seminal, still early-stage business as we are. We've just been talking about the AI opportunity for AWS. It's just an unbelievably large, once in a lifetime business, but that's capital intensive. We still have a very significant opportunity in just core AWS. Today, 85% of the global IT spend is still on premises, and if you believe that that equation is going to flip in the next 10 years, which we do, it means we have a lot of growth in front of us in AWS, where we need to invest capital as well.
Our retail business, which is our most mature business if you will, still if you think about it, we have a single-digit percentage in worldwide retail market segment share, and 80% of the worldwide retail market segment share is still in physical stores. Again, if you believe that that is going to change in the next 10 years, which I do think that's going to flip, we have a very substantial opportunity to grow our retail business, and we're going to do everything we can with continuing to invest in infrastructure and speed to make that so. Then you have a business like our low Earth orbit satellite constellation, Amazon Leo, that we're building that's also capital intensive upfront, but has very attractive long-term free cash flow and return on invested capital.
All these opportunities, we believe, have a good chance of being successful and substantial free cash flow in our OIC downstream, and I think customers, shareholders, and the business will be glad we pursued it long term. We will always be open-minded to alternative ways to invest our cash, and if we find what we think are better ways to invest it, we will.
Thanks, Andy. Next is, other than delivery speed improvements, what other innovations are we bringing into the retail business?
That's a little bit of a trick question because there is so much invention happening right now to improve the speed of delivery for customers. It's kind of breathtaking. Even not mentioning any of those, there is so much invention happening across our retail business. I'll just start with agentic commerce, with the shopping assistant that we built that was Rufus and has now merged with Alexa and become Alexa for Shopping. If you haven't used it in a while, I really encourage you to do so. It's dramatically improved. If you look at the monthly users of Alexa for Shopping in March of this past year, it's up 115% year-over-year. Engagement is up 400% year-over-year. It's very helpful in getting recommendations and comparing products and finding out more in more depth about different products that you're interested in or you have questions about.
It's really like having a salesperson at your disposal. It's very compelling. I also like we have a number of other AI features like Lens, which allows you to take a picture of a product, and it very quickly pulls up search results and makes it very easy to buy. A feature that we have called Hear the highlights that allows you, when you're thinking about a product, to enact an AI experience of people discussing the product in more depth so you get a better sense of whether you want to buy it or not. We've built a foundation model that takes all of our apparel brands and their relative fit to each other to make it easy to recommend the right size to customers on a brand, even if they've never bought from that apparel brand before.
Even features like Add to Delivery, where you've ordered something and that product hasn't arrived yet or hasn't shipped yet, where you can add to that delivery and have it arrive in the same package. Seems easy, very hard. A lot of invention involved there. Apart from some of the features, there's also a lot of product invention in our retail business. If you look at our grocery business, Whole Foods Market continues to grow at a faster trajectory than the rest of the grocery business. They're also continuing to invent. If you look at the new format of Daily Shops in urban centers, they're off to an incredibly strong start. We're expanding those very quickly.
If you look at what we've been able to do with perishables, and the fact that we can now get perishables to our customers same day in 2,300 cities and growing, that's very compelling. Look at our pharmacy offering and the ability to offer such broad selection incredibly easily, same day in so many cities with pricing transparency. Programs like RxPass, where for our Prime customers, they can pay $5 a month and have unlimited access to 50 of the most common medications. These are all very compelling and different experiences for customers. Of course, we're continuing to invent in our fulfillment network.
The re-architecture we've done in the fulfillment centers around regionalization and our inbound network, the work we've done on improving our placement algorithms to end up having items closer to our end customers, all of the invention that we're doing in robotics right now, all of this is to get items to customers more quickly and less expensively. There's quite a bit of invention across our retail business, and that will continue for as long as I can foresee.
Thanks, Andy. Next is, "Can we get rid of job rotation in the company? I know it's for safety reasons, but I find it counterproductive, especially when you have people rotating into positions they aren't fit for and would be more qualified doing something else.
Thanks for that question. Safety is really important to us. It's our number 1 priority for our teammates, and we've been putting a lot of work and resources and innovation into keeping our teammates safe. Job rotation is 1 really good way to help reduce musculoskeletal disorders, known as MSDs. These are things like strains and sprains and lower back injuries, which are really the leading cause of workplace injury across the transportation and warehousing industries. These injuries are sometimes caused by repetitive motions, such as lifting and lowering objects or improper posture when working. Rotation to different tasks and positions can reduce fatigue and ergonomic stress, which helps reduce the potential for MSDs in the workplace. It's an industry-recognized program, and it's really recommended by OSHA. For example, at our robotics sites, employees rotate between our pick and pack stations.
We've used digital ergonomic modeling, which simulates MSD impact across thousands of biomechanical scenarios, to confirm that specifically rotating these 2 job types delivers meaningful shoulder and lower back muscle recovery and supports long-term employee well-being. We're still in the early stages of scaling job rotation across our global operations, we're really encouraged by what ergonomic science shows about its potential to reduce MSD risk by nearly 15%. Overall, our associate feedback in locations where we've rolled out this program has been really positive. More broadly, I'd say that we're really proud of the safety improvements we've made in the last few years. Our global recordable incident rate is down 14% year-over-year and 43% in the last 6 years. Our global lost time incident rate is down 14% as well year-over-year and 70% over the last 6 years.
I think that safety for us will continue to be a very high priority to the team and to me for as long as I can imagine. There's nothing more important to us than our teammates' safety, and we're continuing to invest significantly.
Next is, "Amazon Leo seems like an underappreciated long-term opportunity with relatively small CapEx requirements compared to AI. Can you talk about why you're excited about Leo and what you think the long-term opportunities are in that business?
Well, there are billions of people around the world that don't have access to broadband connectivity, and there are thousands and thousands of enterprises and governments who have assets that have no visibility either. If you don't have broadband connectivity, you can't enjoy the things that many of us in metropolitan areas take for granted. You can't do education online. You can't do business online. You can't do shopping or entertainment. That is the digital divide, and we're very focused on trying to help solve it. We think with what we're building with Amazon Leo, we're on path to doing so. We have over 300 satellites now in space. We have 20 more launches this year, 30 more launches next year, and I think we're going to have a very compelling value proposition for customers.
First of all, I think we're going to have advantage performance relative to others. I think we'll be 6 times better on the uplink performance relative to others, and 2 times better on the downlink performance. I think we'll be lower priced. I think for enterprises and government entities, they really want to take that data off the satellite and store it in the cloud and do analytics on it and use it for AI. Amazon Leo being seamlessly integrated with the leading cloud provider in the world in AWS, there's going to be no better way to be able to leverage your data as an enterprise or government entity than through Leo and AWS. We have momentum right now.
If you look at the series of customers I mentioned in my opening comments, from Delta to JetBlue, to AT&T, to Vodafone, to DIRECTV Latin America, we have a lot of customers who've signed up to use Leo already. We believe this is going to be very good for customers, and long term, this is going to be very good for the Amazon business and our shareholders.
This will be your last question. I loved Hail Mary and as a shareholder am excited about the growth in Prime Video and the success of Amazon Studios. Can you explain why you're excited about this business, and where do you see it growing over the next 3 to 5 years?
I'm excited by the growth and the trajectory of Prime Video too. I think there are a few reasons why it's happening, but it just starts with the content, which is really the most important part of the product. I agree with our shareholder that asked this question. I just think Project Hail Mary was an amazing movie. If you haven't seen it, I really recommend it. It's one of the best movies I've seen in the last five years. It already has over $665 million in box office in just a few weeks. We have a lot more content. On the film side, we had over 100 million plus viewers for the Culpables movie trilogy. All three films reached number 1 in 170 countries.
We took the very successful show The Summer I Turned Pretty, and are making a film, a sequel out of it that should be available in 2027. Of course, we have the Bond franchise, which we're working hard at making the next movie and are very excited about as well. We also have a slate of compelling TV shows from Fallout to Reacher, to The Boys, to Cross, and Young Sherlock. Off Campus launched about a week ago and is off to an amazing start, to Beast Games, to Elle, which is the Legally Blonde prequel series that's coming, to Spider-Noir, to Blade Runner 2099. Just a lot of great TV shows coming. Compelling live sports. There's a reason why 95 of the top 100 watched shows every year are live sports.
We have a very compelling set of those sports that we are able to provide our Prime customers from Thursday Night Football with the NFL to just concluding our first season of the NBA, which went really well, to concluding our first time streaming the Masters this past April, which was also a very good experience for customers, to NASCAR and UEFA and Champions League. Just a lot of sports, and it is driving a lot of customers signing up and watching Prime Video and a lot of ongoing engagement. Then also just the partnerships that we have with very compelling third-party content providers, where we are helping their businesses by being accessible to our Prime Video customers. These range from Warner Bros. Discovery with HBO and Max, to Paramount+, to NBC's Peacock, to Apple TV, and to Fox, and many others.
It is very useful for our customers to be able to go to 1 place and be able to find all that selection and an amazing viewing experience. That business is growing, it's profitable, and it's still early with respect to what's possible. As that concludes our Q&A, I just want to thank everyone for joining today. I hope you can see all the invention and customer obsession happening at an unprecedented scale across the company. We appreciate your support and your interest in what we're doing. It means a lot. Thank you.