EP324 - Amazon Q4 Earnings
In this episode we do a deep dive into Amazon Q4 and Full Year 2024 results. Amazon had an overall strong quarter with record profits from their retail business unit. Amazon GMV grew 11% in 2024 vs. 2023, almost three times as fast as the retail industyr as a whole. Amazon's North American GMV also nearly matched Walmart's in 2024 (Walmart reports it's Q4 results next week).
We also cover the impacts and implicates of the de minimus ban in the US (which was lifted after only 4 days).
We wrap up by comparing Amazon's 2024 growth to the rest of the retail industry. Amazon, Walmart, Costco, Temu, Shien, and Tik Tok Shops representing 62% of all retail growth in 2024 (this is a revision from the 70% estimate we shared in the podcast).
Don't forget to like our facebook page, and if you enjoyed this episode please write us a review on itunes.
Episode 324 of the Jason & Scot show was recorded on Monday, February 10th, 2025.
http://jasonandscot.com
Join your hosts Jason "Retailgeek" Goldberg, Chief Commerce Strategy Officer at Publicis, and Scot Wingo, CEO of GetSpiffy and Co-Founder of ChannelAdvisor as they discuss the latest news and trends in the world of e-commerce and digital shopper marketing.
Transcript
Jason
[0:23]Welcome to the Jason and Scott Show. This is episode 324 being recorded on Monday, February 10th. I'm your host, Jason Retail Geek Goldberg, and as usual, I'm here with your co-host, Scott Wingo.
Scot
[0:36]Hey, Jason, and welcome back, Jason and Scott Show listeners. Jason, how was your Super Bowl?
Jason
[0:43]My Los Angeles Chargers did not do well in the Super Bowl this year. Or my adopted Detroit Lions.
Scot
[0:50]Well, as a chief digital payment retail advertising officer for a big ad firm, I'm sure you were there for the ads. What any ads catch your attention?
Jason
[1:02]Were there ads in the Super Bowl?
Scot
[1:04]Yeah. Yeah. It's supposedly one of the more expensive places to buy an ad.
Jason
[1:08]Yeah. Yeah. I mean, this is kind of hackneyed, but it's near $8 million for each 30 second spot now. And what people miss is that most of these ads cost more than twice that. That both the production of the ad is quite expensive. And then there's now this whole, it used to be that you just surprised everyone by running an ad. And what everyone has learned is you get a lot more mileage by hyping the bejesus out of your ad for weeks before the Super Bowl. And so now the Super Bowl is like step four in a 10-step process and it's all super expensive. So the money is higher than ever. I would not say it was a standout year for e-commerce. Like the last two years, like two years ago, we had this surprise new, not only Super Bowl advertiser, but surprise new retailer in Timu.
Jason
[2:02]And, you know, then last year they were surprised in that they bought so many ads. So that was remarkable. Over the years, we've had some really interesting Amazon ads. And then occasionally we have some weird venture funded, like digitally native direct to consumer companies try to make a splash in the Super Bowl. So I'm not sure I would characterize anything that this year as super standing out from an e-commerce standpoint. Instacart was like the most significant new e-commerce player to do a Super Bowl ad. I thought they did a good ad that was well received with the famous brands. And then, you know, there's some weird stuff. Most of these ads are just to increase brand awareness. And in fact, the way they measure them is things like unaided brand recall, which are, these are metrics that I personally hate. But one of the ads for Fetch, which is arguably an e-commerce company, it's a, you know what Fetch is, Scott? It's a, it's like a purchase affinity program.
Jason
[3:08]So... Brands buy promotions. They're from Fetch. Customers subscribe to Fetch, and when they take advantage of the discount on General Mills goods, they earn Fetch points, which they can then redeem for prizes. So it's, it's a multi-brand customer affinity, digital native customer affinity program. And they did an ad in the last two minutes of the Superbowl that was essentially open our app right now. And in the next five minutes, we're giving 120 people, $10,000.
Scot
[3:44]Well, and so the.
Jason
[3:46]The, yeah, you had to watch to the bitter end and the game was no longer interesting at that point. But the interesting thing about that is that this is one of a handful of attempts over the years to do real activation in a Super Bowl ad versus just generate awareness. So I don't think anyone thought it was super well done or super effective, but I admire the effort to try to make some of these ads work harder. Like you'll remember past years, there were interesting, like there was a one year where the ad was just a QR code that stood on the screen for 30 seconds.
Scot
[4:23]Yeah. Yeah. And bounced around. It did the bouncy thing.
Jason
[4:25]Yeah. To make it harder to scan, which is, yeah.
Scot
[4:31]Cool. I liked the Instacart. That one kind of caught my attention.
Jason
[4:35]Yeah, I was just sad that there was no Timu jingle to get earwormed into my head this year.
Scot
[4:42]Yeah, I was confused by the water bottle thing. I had never seen that.
Jason
[4:46]Yeah, that got kind of panned. That didn't get circle or whatever it's called. Yeah.
Scot
[4:53]Yeah.
Jason
[4:54]Yeah. So it didn't get great reviews as an ad, but it was super confusing. People didn't understand what they were even trying to do.
Scot
[5:00]Yeah, I thought it was like a delivery network. And then I was thinking, oh, we'll have to mention this. And I was like, wait a minute. They're only delivering water bottles. I must have misunderstood the whole thing.
Jason
[5:08]Yeah. It's a filtering, it's a filtering water bottle.
Scot
[5:11]Yeah. I guess I didn't understand why they were distributing a hundred.
Jason
[5:14]I guess they gave away like a million starter kit sample packs, but it was not clear to anyone watching the ad that that's what they were talking about or doing. They, they tried to get too much in a short ad, but I think the, the plot in theory was that the the celebrity who's the the guy from mean girls and or pitch perfect yeah um that he who apparently is a real life fan of the product and got this contract as a result of his like, social shares about the product that he inadvertently ordered a huge surplus of the of these these products and then had to give them all away.
Scot
[5:54]Well we'll have to have you do a live stream next year to explain the commercials to me.
Jason
[5:58]Yeah i think we should get someone that understands.
Scot
[6:05]I got none of that whatever you guys got so.
Jason
[6:07]Yeah yeah that's just all the water cooler conversation i get in my company today the coolest ad in our world i should have said is starbucks that did not advertise at the super bowl but they've created a new holiday which is called i think I think it's called like super Monday and so it's, everybody's like tired and low energy on the Monday after the super bowl. So Starbucks gave all, all rewards members a free coffee today.
Scot
[6:40]Yeah, that was smart.
Jason
[6:41]Yeah.
Scot
[6:42]The, it doesn't work for the kind of drinks we have though sadly.
Jason
[6:45]Not trying to use a little. Basic brewed iced coffee does count. So I didn't, I didn't go, but I could have gotten a nice coffee.
Scot
[6:52]Yeah. Okay. I liked, I thought it was interesting that the Dunk Kings were, or Dunkin' Donuts was basically making fun of Starbucks. They were kind of dissing them.
Jason
[7:03]Yeah. Yeah. I didn't like that at all. No. And that was kind of a sequel, right? Like that was a sequel from last year's ad. Although I think Matt Damon got bumped for Ben Affleck's brother. So that, I will say the, the. The david beckham ad was pretty cute.
Scot
[7:20]Yeah that was funny yeah.
Jason
[7:21]The other david yeah.
Scot
[7:23]And he kicked the ball.
Jason
[7:25]Yeah oh it's the thing i do yeah and so this is for people that didn't watch matt damon portrayed david beckham's long lost brother that his parents never told him he had and, and the other david grew up in america versus david beckham growing up in in england obviously But at the end of the ad, David Beckham explains that he's a pretty famous soccer player. And Matt Damon goes, you mean like Matt Damon famous? And David Beckham goes, more like Ben Affleck famous. And Matt Damon goes, oh, that's too bad.
Scot
[8:00]Ouch.
Jason
[8:01]Yeah.
Scot
[8:03]Cool. Well, today we are going to talk about Amazon's fourth quarter. So the, before we go into there, the setup was you, did the Department of Commerce ever come out with their data for, for the holiday or we're still waiting on it?
Jason
[8:21]No, they did. The December data is out. And so it ended up being a good holiday. November sales were really soft by historical standards, but then December sales were really strong. The partly, I think that's partly because the, October. Literally, the holiday season started in October this year. For the first time ever, they really got consumers to buy more stuff in October. And so then there was kind of a lull in November and sales picked back up in December. Also, just weird vagarities of the calendar. Cyber Monday was in December this year, which it's normally in November. So that had some impact. But cumulatively, between November and December, sales were up 4% year over year, which NRF had forecasted that they would be up two and a half to three and a half percent. So being up four percent is pretty solid.
Scot
[9:13]Yeah. Good. Cool. So then coming into, oh, the other thing I wanted to hit before we get into the deep dive of Amazon earnings is the end of the de minimis thing that you've been talking about.
Jason
[9:26]Yeah. See, we can do a super short version of this. When you import stuff into the United States, you have to pay duties and taxes on those things and you have to meet import requirements. So you can't import products that are made with, you know, dangerous chemicals or made by a child political prisoners or things like that. But we don't have enough inspectors to make sure that everything comes into the United States meets all those criterias. So for a long time, we've had this exception on the import thing, which is if it's a very small package of low value, it does not need to be inspected and you do not need to pay customs. So if you make a container full of t-shirts and send them to the United States to sell, you pay duties on that container. But if you put a single t-shirt in an envelope and ship it to a single customer, it's under the de minimis threshold and nobody has to pay customs on it. And it was a clause that was just to save on labor to process all this paperwork. No one wanted to collect $0.30 of duties and have to spend $5 processing that $0.30 payment. But over time, that threshold crept up and up. And under the Obama administration, it got raised to $600.
Jason
[10:42]And so there's companies like Shein and Timu shipped the vast majority of their packages into the U.S. When they first started via that mechanism. So they, in essence, didn't pay any duties or taxes. And, you know, a lot of people felt that's unfair, that Gap is paying a lot of duties because they're importing containers, a lot of stuff, and Timu is shipping individual packages to consumers and not paying duties. And so Biden actually...
Jason
[11:11]Issued an executive order to cancel De Minimis from China, but he wrote it in a way that it would be like carefully considered and slowly implemented. So it hadn't been implemented by the time he left. And then Trump issued an executive order on like a Thursday saying it's, it's closed on Tuesday over the weekend. And so everyone's like, oh man, Timu and Sheehan are going to be out of business. This is going to be great for Western brands and blah, blah, blah. and it's still all playing out. Who knows what really will happen. My own personal opinion is Timu and Sheehan needed De Minimis to get started and they took full advantage of it. They are now so big that they don't need it. And both of those companies have half the goods they sell are already in warehouses in the US. They were already paying customs on it. They mostly decided that the popular items needed to get to Americans faster. So they have one and two day shipping. That's all coming from the U.S. So they're not dependent on de minimis anymore. And so banning de minimis, I think, is an inconvenience to them, but I don't think it's a death blow.
Jason
[12:19]Also, it doesn't necessarily save you money to use de minimis because while you save money on duties, it costs you a lot more shipping money because you're usually air freighting something instead of ocean freighting it. And that's much more expensive. So arguably, there's a trade-off between the air freight and the ocean freight. And then the secret dirty thing of all of this is every Western brand also figured out this was a thing. And so all the athletic shoes you buy get shipped in a container, not to the United States, but to Mexico, and then de minimis into the United States from Mexico. And so a lot of people assume, oh, this is only going to affect these Chinese brands that we don't particularly like doing well in our market, but it actually is likely to have an equal impact on many big brands.
Jason
[13:07]Trump also banned a minimus from Mexico and Canada, but then that got reprieved. So at the moment, it doesn't affect those. Nobody knew how to and still doesn't know how to process all of this. So literally the post office stopped accepting packages from China altogether for a day because they didn't know how to do this. Nobody solved the fundamental problem that we don't have enough inspectors to inspect all these packages. So in theory, if we try to inspect them all, there's not going to be any more goods in the United States because it's going to take a year for anything to clear custom. So there's a lot of confusion and uncertainty out there. I just sum it up by saying Amazon took full advantage of not paying sales tax when they got started. And it was a huge advantage and it helped Amazon scale.
Jason
[13:56]When we definitively started charging sales tax on e-commerce companies, everyone's like, oh, that's going to put Amazon out of business. And of course it didn't, right? Like Amazon had already gotten so big that they were happy to pay taxes and kind of pull the ladder up behind them and not let other people start with the same jumpstart that they got. And now they're happy to play on a level playing field because they know they can win on the level playing field. And I would argue Timu and Shein are kind of in the same boat with regard to De Minimus. So I think it's not going to materially stop them, but it's going to change how everyone does business. And at the moment, it just created a lot of confusion and uncertainty for everyone.
Scot
[14:40]Yeah. Interesting. Yeah, I saw a screenshot that like most of them, you know, overnight they had switched everything to be from Mexico and native in the U.S.
Jason
[14:49]Well, that's the other problem. There's a way to whack them. When you write, oh, we're just getting rid of de minimis for three countries, you're just starting a whack-a-mo game. Like, everyone will move everything to Guatemala or Indonesia. And then you'll have to issue new orders.
Scot
[15:04]Yeah, it does close the door for other competitors to use it as a bootstrap method. So in a weird way, it almost like solidifies Amazon, Timu, and Shein as kind of like the leaders of this.
Jason
[15:14]Yeah, I think that's it makes it much harder for a new competitor to be born to compete with with these guys that have already used it to get to scale.
Scot
[15:22]Yeah, cool. So last Thursday was Amazon.
Jason
[15:25]Oh, Scott, sorry. There's one other thing we have to cover before we get to Amazon. I did an event for a client, a speaking event, and a picture of it went up on the internet. And my friend Scott Wingo has too much free time on his hands and posted a whole LinkedIn article, zoomed in on my sock in this picture, which revealed the logo of the client I was doing a presentation for. And it's gone like Jason and Scott viral. So all of my supervisors at Publisys are super impressed that I got caught wearing a logo. And I have Scott Wingo to thank for that. So thanks, Scott.
Scot
[16:05]You're welcome. Hey, you wore the socks. It was funny. I was talking. I did another post where I went to a Starbucks one day and the new CEO, I don't know if you know this, but he's asking them if they have time to write little notes on the cups. I went to the Starbucks like two weeks ago and everything was lined up alphabetically. It was like one of those little delivery stations. And it all had like personalized notes. And I just like put it up there as like a really good example of small things are important for customer service and like wearing your client's socks. So little things like that matter.
Jason
[16:35]Amen. On the one hand, it was super awesome. On the other hand, the unintended consequence of that, Scott, is I now cannot go to any other client presentation without their socks. I just cost you thousands of dollars. Yes, yes. So let me just say, I've spent the last like five days, like scraping the corporate web company stores of a lot of retail and brands.
Scot
[16:58]Cool. So coming into Amazon's earnings, you know, it's really become all about AI these days. There's a lot going on in AI. First, we had Microsoft with Azure. They largely missed expectations on the growth of their AI piece. And then Google had a similar reaction. And what they're all saying is the demand is outstripping our ability to meet it where we're gpu and energy constrained and and fulfill a data center constrained so then everyone was kind of like oh boy it's going to be bad for amazon and there's a lot of little bit of a fear uncertainty and doubt in the market and then the big disruption happened when deep seek which is the chinese company there's there's been so much ink on this i'd be surprised if people hadn't heard about it they're.
Scot
[17:42]Their reasoning model, R1, came out and it was basically what they call open source. It's really open models, meaning open weights. You don't actually get the source code. So it's a little bit of a misnomer. I prefer open weights. But what that means is you can just basically download this thing and run it anywhere. And this is interesting because it's as good as the latest models from everyone that's charging quite a bit of money. And you know what what this the direct disruption here is it became very clear that the way this is priced is dollars per million of token million tokens so if you're going to use one of these llm kind of engines from an api perspective to power a feature or something like that it was really it was relatively very expensive in the early days of chat gpt and now it's gone from it's basically turning to zero very quickly so a lot of people were beating up Amazon for not having their own multi-billion dollar investment in an LLM. They are an investor in Anthropic, but they don't have their own LLM. So it's interesting because Jassy's been saying, we're kind of neutral. We just want to host them all and let the customer pick and the customer is going to go where the data is. And we have a lot of data.
Scot
[18:56]So then that actually ended up, DeepSeek R1 ends up really being a big benefit to companies is Amazon for the reasons I mentioned, but also Apple, because it turns out the way DeepSeq developed this thing, the whole model is pretty efficient because it only lights up a little part of the model instead of all the parameters. It uses this expert network thing.
Scot
[19:19]And anyway, so there's a, you know, Apple has put a lot of neural nets on chips near memory, which this model works really well for. So there's kind of like two unexpected winners in this, which are the two companies that didn't spend billion dollars on models, which were Amazon and Apple. So then, so that was kind of the setup going into it. And then...
Jason
[19:41]Scott, there's one super important stat about DeepThink.
Scot
[19:45]Yes, DeepSeek.
Jason
[19:46]DeepSeek. They should have called it DeepThink, by the way. But apparently Meta has 20 AI execs that all have an, each have an annual salary that's greater than the total training cost of DeepSeek.
Scot
[20:00]Yeah. Yeah, that's pretty crazy. It also kind of shows that these export controls may have backfired because they made them, they constrained them so much, they came up with like a really clever way to do it. So it was all very interesting. And then we didn't have any Walmart coming in. That hasn't happened, right?
Jason
[20:18]Yeah, Walmart has a very peculiar financial calendar, but what would be the reporting for their sales in Q4 of the calendar year happens on February 20th. So that's next Thursday.
Scot
[20:33]Yeah. Okay, so then Amazon came out, and much to the relief of everybody, their revenue was in line. So it was kind of a meet. If you have kind of miss, meet, and beat, it was a meet expectations on the top line, but they beat on the bottom line. So earnings per share came in at a buck 86, which beat the estimate of $1.48 by 20%. So a really nice kind of profit beat there on earnings per share.
Scot
[20:58]And then the guidance is what Wall Street would call soft. So Wall Street already thinks they're going to know what the guidance is. And they look at this midpoint, there's a range, and they do all this kind of triangulation. And it came in a little bit below. But they have a history of that. So people weren't too concerned. So I think there was a bit of a sigh of relief that things at Amazon were pretty good and there wasn't a big miss on AI. And then I think they got a nice benefit. Now, the stock was down a little bit. I think that just kind of happens. It's recovered most of that since then. It did signal that a relatively conservative guidance going forward where they were going to grow between 5.4% and 8.5%. Again, they beat this guidance like for the last eight to 10 quarters. So I think people know there's some conservatism in there. They did say, you know, giving the environment with like this de minimis thing, these tariffs and all this stuff that they wanted to be a little careful there. Also, everyone, they did signal in the call that they're going to invest a lot. And we'll talk about that in a second. So they're going to kind of enter this invest phase again. And a lot of it is around AI, but it's not models and things. It's really just kind of supporting AWS as a part of it. But they're also investing a lot in AI applications, they call them, for the retail and the supply chain part of the business. And we'll go into that.
Scot
[22:23]So then they dove into kind of more of the retail side and some of the macro numbers. Why don't you take that over?
Jason
[22:30]Yeah. So what they typically disclose in their earnings call is the retail business unit revenue, which is not the same thing as their total sales or what we would call gross merchandise value. So the revenue number they reported was that Q4 was up by 10.5%, which was good and above expectations. That was like $188 billion for the quarter, which would be a total of $638 billion in revenue for the year. They said online stores grew at 7% year over year, while physical stores grew at 8.3% year over year. That physical store number is interesting because, you know, the bulk of the Amazon physical store fleet are these Whole Foods that they bought. And then over the last five or six years, they've opened a bunch of store concepts, Five Star, Amazon Books, Amazon Go, and Amazon Fresh Grocers.
Jason
[23:32]They've closed most of those concepts. They only have a handful of the Go stores left and the grocers. So the number is back to mostly being Whole Foods, which hasn't been a great growth story. So the 8.3% growth for physical stores is somewhat encouraging for Amazon on Whole Foods. And I don't know if this is coincidental or not, but the CEO of Whole Foods got named the head of olive retail for Amazon or olive grocery for Amazon rather.
Jason
[24:01]So those were good numbers. I'm most interested in how much stuff they sold, which they don't tell us. They give us some data, like their breakdown of first-party and third-party sales. And we're able to back into an estimate of GMV. And all the really smart, big-brain stock analysts put together these big models and do the math for me. So in the old days, you and I used to do models. But now I just get a way prettier spreadsheet from Morgan Stanley and Citibank. So the Morgan Stanley model is already out. And they think GMV for Q4 in North America, which is, you know, mostly what I like to talk about on this show, was up 11.4%, which is huge. The, you know, again, like retail is growing about 3.8% in Q4. So it's Amazon's growing three times as fast as the overall retail industry. And then, you know, you look at that on a full year basis and Amazon sales for the, in North America, GMV was up 11% year over year, which is huge.
Jason
[25:09]And somewhat even more impressively, those sales are getting more profitable. The fulfillment network is getting faster and more efficient. The cost to serve is going down.
Jason
[25:21]I think they had a stat that like same day and early next day deliveries grew 65% versus the number of deliveries in the same quarter last year. And so if you remember, about a year and a half ago, they pivoted from this national model to this regional model of distribution, and it's totally working. They're delivering goods way faster and cheaper. And so what kind of the interesting unexpected thing here is as they're able to deliver goods faster, the mix of goods that customers are buying from them is changing. So they used to not be good at selling everyday essentials, but now that they're so good at same day and fast next day delivery, they're selling a lot more of those essentials, which is taking a huge bite out of grocery. And I would argue even a huger bite out of like drugstores and convenience stores. So the, for a long time, you know, analysts are always like AWS is the growth and retail is this necessary evil and blah, blah, blah. Well, I would argue retail was really the star of, when you peel back the onion on the numbers, retail was really the star of this earnings call for them and is really helping to, it's contributing a lot of dollars that are now getting invested into AI cloud capabilities.
Scot
[26:44]Yeah. Cool and then on the marketplace side the third-party seller business accounted for 62 percent of paid units kind of has ebbed and flowed between 58 and 62 for a while i mean q4 it does surge up because the first party business gets swamped and sells out and then the third party is kind of there to to take up the slack that's up 60 from 60 so in q3 it was 60 and then here it was 62. I believe last Q4, it was also 61, 62. So not a big change there. That's kind of steady state now. The revenue grew 9% year over year. These numbers used to be crazy at like 20, 30, 40, 50%. But again, we've kind of settled in with overall retail on a lot of the business. And then that came in at 47.5 billion is the revenue from third-party seller services, which includes some of the pieces of FBA and whatnot. And then the ad business is my favorite. That's the best business.
Jason
[27:40]I know Scott waits with bated breath to talk about the ads every year. So again, ads was a super robust growth driver for Amazon. I think the trailing 12-month run rate for ads, so Q4 was up 18% on a year-over-year basis. That Q4 number would put them at nearly $70 billion run rate for 2025. But trailing 12 months, it was $56 billion of ads. And, you know, remember you sell a banana at like 4% gross margins. You know, you sell an ad at like 80% gross margins and that's being like conservative. And so this, this ad business contributes a tremendous amount of free cashflow and profit to Amazon.
Jason
[28:29]A couple of notes here. The vast majority of this ad are sponsored product listings and search. And that's true for most retail media networks. But increasingly, Amazon's a more complicated ad network because they do now have a lot of network programming. They have an NFL contract. And this ad number isn't just the retail media dollars. It's all of their ad dollars. So there may come a day when, like, just like we have to have a kind of a third party model to estimate what GMV is, there may come a day when we have to estimate what percentage of the ad revenue is retail media versus other flavors of ads.
Jason
[29:11]But as a result of this huge success that Amazon has had, everyone else in the retail industry is worked up into a huge lather over retail media networks. So, all the big agencies like mine, you know, have now hundreds of people that, you know, are talking about and helping brands and retailers execute, helping retailers build retail media networks and helping brands spend more money on those retail media networks. And every tiny little retailer is hiring a bunch of ad execs to launch their own network.
Jason
[29:41]And while I do think this ad business is great for Amazon, I think it's pretty good if you're Walmart or Instacart, I think it's wildly overhyped. And what people are missing is the only people making money on retail media networks are retailers with huge traffic at the top of the ecosystem that all have one thing in common. They're marketplaces because all of those third-party, those sponsored product listing ads are being bought by third-party sellers. It's not being bought by the first-party sellers. There's a little bit of revenue coming from first-party sellers, but that's the revenue that they used to get that they called shopper marketing dollars or trade dollars. The growth engine here is marketplaces get to make money selling ads. And so I just point all this out because I see more and more like small regional wholesale retailers with no marketplaces spending a fortune to launch a retail media network thinking they're going to have the same success on a smaller scale as Amazon. And it's not true that this retail media business is tied directly to the success of the robust third-party marketplaces that you predicted. Well, when did you predict marketplaces? About 70 years ago, was it?
Scot
[30:56]Yeah, it feels like it.
Jason
[30:58]Yeah. So just an interesting part of this whole thing. But for sure, if you're Amazon or you're another big retailer like Walmart or Instacart, it's another competitive advantage that helps the big get bigger at the expense of the smaller players.
Scot
[31:14]Yeah. And then I love that business and not facetiously because it's basically pure margin. Like it just makes the whole thing work. I do feel like it's got to, as a consumer, whenever I go to kind of top of the funnel, wherever I'm kind of like researching and try to find products, I feel like it's definitely, kind of ruined is a strong word, but greatly diminished the shopping experience at Amazon, which is always debatably not great. It's very kind of transactional to begin with. It's very hard to find the product you're looking for now because you're just like inundated with ads at this point. I think they've, they've reached the, you know, almost jumped the shark levels of the amount of ads, but that's a personal preference, I guess.
Jason
[31:57]No, no, no. I think most people agree with your evaluation there. Like, you know, they're for a long time, Amazon tried to make this argument that it's always about the customer experience and customers like these helpful ads. But I think that's, that's a pretty hard argument to make. It's, it's every digital platform in the history of mankind, like started out super useful and ad free and they slowly turned up the sponsorship over time, right? So when Google was 10 blue links, when you did a search, it was super useful. Now, you know, all the pixels on the front page of Google search are paid for, you know, Facebook used to, you know, Elevate all this organic content from people in your social graph. Now it's predominantly ads. TikTok with Instacart with the interest graph and now predominantly ads. It's the same thing. The problem is if the product is the ad and not the customer satisfaction from selling goods, this is the foreseeable outcome. And all of these businesses have just realized selling ads is a better business than being a wholesaler.
Scot
[33:06]Yeah. Cool. So then they rolled into the AWS part of the program. The AWS revenue reached $28.8 billion for the quarter, growing about 19% year over year. And its operating margins were 36.9%. That's actually an improvement of 732 basis points. So, you know, even though, you know, this thing is growing really quick, it's both growing fast and improving margins, which is good. And then you can think about AWS as having kind of three things. There's like a storage component where they can store your data, and that's commonly called the S3 family of offerings. And then you can basically rent, compute in traditional fashion, which is CPU. And then you can rent AI, which is on a GPU. So there's kind of like storage, CPU, and GPU.
Scot
[33:56]I think margins went up because of the strong demand for GPU. When supply outstrips demand, you can have pretty aggressive pricing in there and higher margins. So I think what's happening with the advent of AI is the workloads are really tilting towards that GPU side, which are much higher margin. And speaking of AI, Amazon is one of the two companies. So Google has their own GPU called a TPU and Amazon has Tranium and they came out with the version two of Tranium they announced was generally available. And it is much cheaper than other GPUs, which basically what they're the other big player in town is NVIDIA. So that's kind of what they're saying. And they still offer NVIDIA, but you get to pick. So what they said when they announced Tranium two is if you choose that is 30 to 40% better price performance and other GPU instances. So let's say your company and you want to host this deep seek open weight model, you now can host it on Tranium and run it yourself. It won't connect to China because you're controlling the infrastructure and you can basically have your own model. All you can eat just paying for the GPUs. Whereas when you go to like a open AI API or Anthropic or anything, you're paying for the GPU, the model and their margin. So this is what's driving that price per million tokens down towards zero.
Scot
[35:13]And in fact, to that end, they basically said that they would have had a much better quarter on AWS, but they are at maximum capacity for these GPUs, both their in-house and third party. And this is the same story that we're hearing from everybody is there's so much, so many people want to put out these GPU based AI workloads that they don't have the infrastructure for it. And it's not only the chips. So they're chip constrained, data center constrained, and energy constrained, and refrigerant constrained, and so on and so forth. So it's kind of a crazy time that there's so much constraint around this one part of something everyone wants.
Scot
[35:52]They did, when they file their 10Q, they put out a backlog and it was pretty material, which got people excited. So there was like some backup data to that. And then, you know, they did announce they're going to make some investments to the tune of $100 billion in 2025. And this is crazy because Meta came out and said something like 25 to 30. Microsoft said a number that was big, like 65 to 75. And then Amazon one-upped them all and said $100 billion. So that's pretty eye-popping. They didn't specifically say exactly how that's going to be, but it was like in the conversation of we're at capacity for AI. So I imagine they're going to build a literal tons of these Tranium chips and deploy them as well as whatever GPUs they can get their hands on from the folks out at NVIDIA. I already mentioned that it feels like they're really doing well here. And they talked about that. They didn't specifically talk about DeepSeq, although they quickly announced that they had it on their infrastructure.
Scot
[36:51]Jassy's been saying all along that they're not going to bet on an LLM. They're going to be neutral. But what they're going to bet on is people are going to want to run them near their data. And so because they have all these traditional CPU workloads, they have all their data. So if you're a big ad company, you're running all these calculations on ads, you're probably doing that in the cloud. You want the data to be near the compute so that it's fast. And now if you're going to run any AI workloads, they're going to want to be near the data because they're usually not just running an LM, you're running it through your data through like a retrieval augmented generation so that you can get insights from your data and either present them internally or to customers. So Amazon has really set up very well for this. And this part, a lot of the, you know, prior to kind of recently, most of the GPUs went towards training these models. Well, training is becoming a commodity, and it's really the inference part, which is where the models are actually running. And so they're really well set up for this whole thing. So I thought that was really interesting that they had the foresight, and they ended up being right. You know, they and Apple were the ones that were right, and everyone kind of thought they were crazy for not investing hundreds, tens, if not hundreds of billions of dollars in these things.
Jason
[38:07]Yeah. Yeah. That worked out pretty well for them. Yeah.
Scot
[38:11]The jury's still out, but like, as of today, it feels like they made a pretty good decision. Now, if, you know, OpenAI is the only one that has a, you know, ASI and AGI and all these fancy, you know, superhuman like capabilities, then maybe it's a problem, but seems like, you know, even then, are you really going to want that to be kind of running through your corporate data or something? You know, that starts to be even kind of scary. So I think this use case of making my internal data smart has many, many decades to run before it runs out of steam.
Jason
[38:41]Yeah, I hear them all talking a lot too about this AI application layer that they're investing in. So not the underlying LLM, but all these industry-specific, vertical-specific application APIs that sit on top of those LLMs. So like at NRF, they were focused on the hundreds of retail-specific patterns they had for AI applications that get to take advantage of every time there's an iteration or an improvement in the underlying LLMs.
Jason
[39:16]I wanted a couple of things that jumped out at me on the AI side. Number one, in our last podcast, you made a prediction about the profound impact of agent-based AI in commerce. And we started to talk a little bit about like, is this friend or foe to retailers? Are retailers all going to be blocking the AI agents from scraping their sites? Or are they going to be partnering with the agents? And I just want to say that that whole narrative got like hugely picked up and discussed on LinkedIn. It's super interesting. And we're definitely in the near future going to have to do a deeper dive on the podcast on the implications and ramifications specifically on retail of all this AI. But I do want to point out in the short term, you know, it's not all roses. There's two big negatives to AI. Like number one, every time Scott Wingo does a search on perplexity, it is taking out a small rainforest in Brazil.
Jason
[40:12]So like it's not clear at the moment like that all these things can scale, that there's enough electricity for all these things to scale. Unless these more efficient models like DeepSync end up winning. Right. Because one of the benefits of it is like it's a lot more compute efficient to to build a new model. But the the bigger negative ramification to me is you cannot buy a darn graphics card for good video gaming in the world anymore yeah all the all the the ai hosts have have bought all the chips and and you know if you just want to run call of duty at 240 frames a second you you are mostly out of luck which is quite irritating in the goldberg household.
Scot
[40:56]Yeah and whatever was available, the crypto bros grabbed. So you're kind of third on the wait list there.
Jason
[41:02]Yeah. Yeah. So I'm, I'm a little bitter about that, despite the fact that I'm, I'm very bullish about the profound impacts AI is going to have on commerce. I will say it is interesting in the, in the AWS world, because it feels like there's, there's two kinds of compute loads in the, in the world. There's the traditional CPU based compute loads that, you know, Amazon's been making a lot of money as people migrate from on-prem compute loads to the cloud. And then there are these new GPU-based AI compute loads. And the GPU-based compute loads are totally constrained by capacity. Like Amazon could sell more loads if they had more chips. They only have so many chips, and that's why they're going to spend a fortune on chips. The it does appear that the rate of growth of the cpu loads has slowed down which what is a worry sign to some investors because they they have excess capacity there if there were more people that wanted to move more of their loads from their on-prem to the cloud amazon could make more money today whereas if more people wanted to do ai they like they're they without more chips they really can't do anything.
Jason
[42:20]And so what's interesting to me is that it has not slowed down because all the world's compute is on the cloud. Arguably, well less than 25% of all the world's compute is on the cloud. So there's still plenty of compute happening on premise that could be moved to the cloud. But I think the double whammy is that all the easy compute loads got moved to the cloud And all the ones that had a particularly high ROI of being hosted on the cloud moved. And now all these older legacy compute loads are harder to move, are less appealing to move. And we were joking earlier, a lot of them are probably written in COBOL. And so, yeah, it's going to be interesting. They may ironically require AI to port all the business logic and COBOL code to cloud loads to kind of reinvigorate the cloud growth.
Scot
[43:18]Unless there's some kind of a simulator thing or, you know, emulator mode, there's no running COBOL up in AWS. So those loads are going to be stuck on some old mainframe for a very long time until something can rewrite it.
Jason
[43:32]Yeah yeah but it won't surprise me if the the ai bots get really good at porting all those those cobalt applications to to some modern architecture on the cloud.
Scot
[43:41]Yeah open eight sam just yesterday he was sam altman was speaking in japan and said they have an internal open ai, developer agent that they think is the top 50 developers in the world and they think they'll have the best one in three years yeah and these he says all this wacky stuff and it's kind of like, it's easy to say because it's unverifiable. So, you know, it's very hard to tell if he's hyping, but he's a pretty, you know, a lot of what he said has actually come true. So it's like, there's evidence that he's not making stuff up.
Jason
[44:15]I think that's very fair. And then there, like there are real world examples, maybe not quite that profound, but like of this stuff working, right? Like, so going back to Amazon earnings, like on the investor calls, they've talked a lot about, I'll tell you what workloads AI has been awesome for us on coding is porting all of our end of life Java 4.1 code to Java 5 that in the old world would have taken a hundred million man hours to do. And now takes like 10 man hours because of AI, right? Like I'm, I'm dramatically exaggerating because I can't remember the numbers, but, but it's, there are things that used to be hard and expensive to port that are like, are real business savings for big companies like Amazon now as a result of the AI being an adequate coder. So forget even being one of the best coders in the world. Just being adequate is saving Amazon a fortune.
Scot
[45:08]Yeah. I hate technical debt like that. Rewriting for some new version. Yeah.
Jason
[45:13]Doesn't add any benefits. You have on Sundays. So I want to bring this all home for our audience that's mostly focused on commerce, right? So if we zoom back to the retail side of Amazon, I'm just going to recap for you that Amazon probably sold $540 billion worth of stuff in North America, which was 11% more than they sold the previous year. So they're, they, they added $56 billion in sales this year.
Jason
[45:48]The U.S. Department of Commerce said that all of retail grew 3%, 3.6% this year over last year. So 11% growth is way higher than the industry average. And the whole industry added $183 billion in sales. So Amazon alone is $56 billion of the incremental $183 billion that happened in North America in 2024.
Jason
[46:20]So then the other huge retailer is Walmart. So you'd love to compare Walmart and Amazon growth to U.S. Department of Commerce. But as we mentioned, Walmart's earnings aren't going to get released until the end of next week. But we have three quarters of earnings and we have estimates, right? And at the moment, people are estimating that Walmart's Q4 is likely to grow about 4%. So if they see 4% growth Q4 this year versus last year, it means that Walmart will grow 5% for the year. So half as fast as Amazon, still meaningfully faster than retail, right? Retail is growing at 3.6%. Walmart's growing at 5%. Amazon's growing at 11%. That means that Walmart contributed another $26 billion of incremental sales, if that plays out. And then our friends at Timu, Sheehan, and TikTok probably captured about $50 billion of sales in 2024 that they didn't have in 2023.
Jason
[47:23]So you take $26 billion from Walmart, $56 billion from Amazon, $50 billion from Timu, Sheehan, and TikTok, and that's $132 billion in growth. When the entire industry had $183 billion of growth. So it basically means Amazon, Walmart, and those three Chinese companies are 72% of all retail growth, which is super scary. Obviously, the big are getting big, and we're seeing this huge gap open up between those leaders and everyone else. The other super interesting thing, if these numbers are right, it means Amazon sold $540 billion worth of stuff in 2024. Walmart in North America will have sold $547 billion worth of stuff. So Walmart in North America is still slightly larger than Amazon. And when I say Walmart, I mean Walmart Inc., including Sam's Club in the U.S.
Jason
[48:21]Walmart has international sales. Amazon has even more international sales. The U.S. economy is faring pretty well. A lot of other economies in the world where Amazon in particular does business are not faring as well. So there's more headwinds on the international side of Amazon's business. But apples to apples in North America, this is really turning into a two-horse race with these Chinese entrants taking a huge bite of the market, but clearly not taking a bite out of Amazon or Walmart, which is super fascinating and super concerning.
Jason
[48:58]And Scott, I roll all this up and I say to myself, huh, all of these companies that are growing, the thing they most have in common is their growth and or their primary business model is all coming from marketplaces. Like increasingly you can't name a huge super successful e-commerce wholesaler that buys goods and resells them the like amazon keeping 62 of their sales as third party is even more impressive if you remember i said that amazon selling way more essentials than they ever did before historically essentials were all one p so for them to keep 3p at 62 that means 3p is growing really robustly based on the mix change at Amazon. So they're increasingly, I'm looking at all these numbers and I'm going, man, it is just getting harder and harder to make a living selling other people's stuff that you either have to make your own stuff and sell it to consumers and be content not to be one of the largest retailers in the world. Or if you want to be one of the largest retailers in the world, you have to be really good at a marketplace. And then that unlocks all these supplemental revenue streams and retail media revenue streams and all these other things we've been talking about on the earnings call. So you may have seen this a long time ago, but like, It's increasingly seeming possible that marketplaces will eventually kill off the wholesale model, which is crazy to think about.
Scot
[50:24]Yeah, it's kind of like what happened in China, though. So it's, you know, it's something that a lot of people have predicted for a while. It's actually taking longer than I thought. I'm a little perplexed why all of e-commerce isn't growing a little bit faster, but that's a topic for another time.
Jason
[50:39]Yeah. So last week we did an overly long show. And so this week we worked really hard to bring it in slightly under our typical hour. So I'm happy to say we have not used up all our audience time, but hopefully this was a helpful recap on Amazon's earnings, a forecast into Walmart's earnings next week, and gives everyone a really good kind of feel for how 2024 played out in North American commerce. I'm always Debbie Downer. I want to leave everyone with one piece of good news. The U.S. Government data for January sales comes out next week, but the National Retail Federation has a new data source called the Retail Monitor, which uses a huge panel of credit card data. And that data came out this week and they report that January sales are up 5.7%. So, again, all of last year grew at like 3.6%. Holiday grew at 4%. January being up 5.7% over last year is really encouraging. and certainly we'll hope to continue that momentum for the rest of 2025.
Scot
[51:42]Look at that Jason ending on a positive note. Who would have thunk it?
Jason
[51:46]Yeah, it's the new wear your clients logos and think happy thoughts version of Jason.
Scot
[51:53]Cool. Well, thanks everybody. And until next time.
Jason
[51:57]Happy commercing!
