What company has the lowest IT spending budget in the world, but has also paradoxically spent more money than any company in history investing in creating a new, modern, cloud-native system that is capable of running just about any application at just about any necessary scale?
The answer is, of course, Amazon, and the real genius of its Amazon Web Services division, which is the world’s largest provider of cloud compute, networking, and storage services. Thanks to the large investment it has made in creating AWS, Amazon’s effective IT budget has been zero – or lower than zero, depending on how you want to look at it – for many years now. But to get there, Amazon had to let AWS create a platform that spanned the globe that millions of companies could – and would – use.
Let’s do a thought experiment, because it is a Philosophical Friday. Just how much are you paying for Amazon’s IT department to ride for free, and what kind of competitive advantage does your investment in AWS give to Amazon in the IT sector and in any sector for that matter?
We will start with the IT budget at the e-tailing, entertainment, and advertising giant. Amazon does not provide any guidance on what its various business units – online retail, streaming media and content, advertising, and so on – spend on their IT infrastructure to run those collective businesses. But the market researchers at IDC estimated in 2018 that Amazon had the largest IT budget in the world, at $13.6 billion; Google and Wal-Mart tied for second at $12 billion IT budgets for 2018, by the way. If you extract out the AWS revenue stream from Amazon’s overall sales, then that non-AWS portion of the Amazon business had $207.2 billion in sales, and that IT spending represented 6.6 percent of revenues. This is right in the ballpark of what big banks and retailers spend.
If you fast forward to 2021, we have to guesstimate what the IT spending is for the non-AWS part of the Amazon business, which had an incredible $407.6 billion in sales last year. Assuming that Amazon is getting better at IT, in large part by moving to AWS and using the services it created first for itself and then commercialized to the world as part of AWS, then we figure the IT budget at the core, non-AWS business in 2021 was around 6 percent of revenues, or $24.5 billion. The increase is mostly for scaling capacity to scale the Amazon business, offset by the efficiency improvements that AWS and the rest of Amazon are maniacal about.
In the fourth quarter of 2021, the AWS cloud business posted $17.78 billion in sales, up 39.5 percent, and had an operating income of $5.29 billion, up 48.5 percent. This part of the Amazon conglomerate is truly profitable, and because of supply chain issues, overtime and pandemic safety costs, and increased facilities costs and media production costs for the retail and entertainment parts of the Amazon empire, the non-AWS parts of the business posted an operating loss of $1.65 billion, and it was only through a gain of $11.8 billion on the sale of stock as electric car maker Rivian Automotive as that company went public last fall and the ongoing profits from AWS that allowed the overall Amazon to post a killer $14.23 billion net income, up 106.8 percent year on year.
If you tally up the numbers for AWS in 2018, then the cloud unit had $25.57 billion in sales, more than offsetting the Amazon IT budget of $13.6 billion given by IDC. And if you want to only cover the Amazon IT budget with AWS operating profit, then the $7.3 billion it posted in 2018 covered about half of the IT spending bill for the non-AWS portion of the business.
Now, jump to 2021. AWS reported revenues of $62.2 billion, a factor of 2.54X larger than the $24.5 billion IT budget we estimate that the non-AWS parts of Amazon had last year. And if you think that the core Amazon budget has to be covered by AWS operating profits, then the $18.53 billion in operating profit that AWS posted for all of 2021 covers three-quarters of the IT bill for the core Amazon businesses. (We are not implying that AWS is somehow non-core. It is absolutely vital to Amazon.) And if current trends of revenue growth for Amazon and for revenue and profitability for AWS all continue, it won’t be too long – two or three years – before profits from AWS will cover the entire non-AWS IT budget tab.
Amazing, isn’t it? Jeff Bezos should send thank you notes to all AWS customers.
Now, to get the world to pay for your own IT costs, you have to invest a lot more than the IT budget you might have otherwise had. The AWS business has taken an enormous amount of capital to build a global network of infrastructure that can support millions of companies and therefore throw off enough cash to cover the non-AWS IT bill with the revenue stream, and as we say, it will take a few years before it can do it with operating profit. (Most of the Amazon IT budget, we reckon, is for AWS capacity, but there is still plenty of unique software development that the other Amazon businesses must do, which ain’t cheap.)
In the call going over Amazon’s financial results for the fourth quarter, we did get some insight into just how much that infrastructure spending on AWS was. We have had to infer it from generic capital investment numbers Amazon has provided in the past.
In the quarter ended in December, AWS shelled out $21.46 billion for capital expenses, and the vast majority of this was for property and equipment (which includes software and web development), but also capital and finance leases on facilities and build-to-suit leases on property and equipment. That was an increase of 21 percent year-on-year, which represents a peak in absolute spending but a deceleration in overall growth.
Brian Olsavsky, Amazon’s chief financial officer, said that just under 40 percent of the capital spending in the past two years has been on infrastructure, and most of that feeds into AWS business. That adds up to around $50.45 billion. Fulfillment centers for the Amazon retail business cost another $37.45 billion, and the expanding Amazon transportation network, which allows for very fast delivery, cost another $31.29 billion, leaving another $8.3 billion for office space, stores, and other capital expenses. Ok, so here is the whopper number. Assuming that the AWS portion of the capital budget has been fairly constant at around 40 percent – which is a reasonable but possibly incorrect estimate – than Amazon has ponied a little bit more than $104 billion on infrastructure on AWS since it was launched. Half of that investment has come in the last two years, and based on what Olsavsky explained on the call, AWS infrastructure is going to represent more of the capital spending pie in the future, not less.
“We are still working through some of our plans for 2022, but it’s coming into focus a bit,” Olsavsky said. “We see the CapEx for infrastructure going up. We still have a very fast-growing business that is growing globally, and we are adding regions and capacity to handle usage that still exceeds revenue growth in that business. So, we feel good about making those investments.”
There is little doubt why. AWS has generated $218 billion in revenues and $59 billion in operating income between 2009 and 2021, inclusive, and it has taken more than $104 billion in capital investment in hardware, facilities, and software to get that revenue and profit stream. But look at the return AWS is able to get on that capital. It is a heck of a lot better than what the rest of the Amazon businesses can do. Well, excepting maybe its burgeoning advertising business. The rest of the Amazon business appears to be under water due to the pandemic and was only treading water before then.
One last thought. The real value in all of this for Amazon might just be the advanced research and development that gets funded by profits coming back into Amazon from AWS, which creates the next set of scalable datastores, databases, development tools, and application frameworks that Amazon needs to run its conglomerate. After Amazon gets a head start on these technologies, they are productized and then rented out as services to the rest of the world. But we suspect that Amazon itself will always have a lead of several years over the AWS customers who use its various services. Amazon can be perceived of as offering the most advanced technology for sale, but the real advanced stuff could lag into the market significantly.
We have nothing against monopolies – they happen naturally in a lot of markets, and all we ever objected to was an unregulated monopoly with too much power, whether we were talking about IBM in the 1960s and 1970s or what Amazon could become several to many years from now. Monopoly cannot be so strong as to stifle competition, but in some cases, and perhaps with search engines, social networks, and cloud computing, maybe monopolies will emerge. (Or have already.) But rather than talking about monopolies in terms of relevant market share, which is an old way of thinking about it, maybe we need to start thinking about them as insurmountable and unassailable architecture leadership. No one can get the intellectual capital together to ever build a Google, or Amazon, or Facebook from scratch, much less the actual capital to buy the gear to run such big machines and their hyperscale applications.