Commentary: There are reasons to think the clouds would be wise to go back to the core infrastructure that made them popular, but there are other reasons to think this will never happen.
Here’s a crazy thought: What if the cloud vendors decided not to continue to move “up the stack” to applications and other higher-order services? It’s not my thought, but it’s an interesting one.
Erik Bernhardsson, the former CTO of Better.com, posited the idea that “Cloud vendors will increasingly focus on the lowest layers in the stack…[while] [o]ther pure-software providers will build all the stuff on top of it.” This is absolutely not the cloud world we currently live in, with AWS CEO Adam Selipsky taking the stage at AWS re:Invent 2021 to talk about how AWS will “continue to build more of these [vertical industry] abstractions on top of our existing foundational services,” with Google and Microsoft already well down this path of industry solutions, not to mention applications, databases etc. That sounds like more innovation up the stack, not less. (Disclosure: I formerly worked for AWS.)
But Bernhardsson makes a compelling point: The cloud vendors may get spread too thin to compete effectively with pure-play managed service vendors. This could be true, but it’s hard to see the cloud vendors giving up their need to grow, and there’s a lot more money in applications, for example, than operating systems.
SEE: Hiring Kit: Cloud Engineer (TechRepublic Premium)
According to Bernhardsson, there’s money to be made by focusing on what cloud vendors used to do: just core infrastructure. No, this isn’t a commodity business. One of the most important announcements AWS made at re:Invent this week was the introduction of the Graviton3 processor. As Tom Krazit said, “Graviton is a years-in-the-making cloud-infrastructure moat that is drawing converts focused on the most basic cloud computing question: How much will it cost to run my application?”
There is real money to be made, and real differentiation to be had, at the core infrastructure level.
It’s also the area where the clouds face the least competition, which is where Bernhardsson’s argument makes the biggest dent. “[D]eveloper experience has become an attack vector,” he said, with startups like Databricks and Snowflake outflanking their more established cloud provider peers through tighter focus and better developer experience.
For the cloud providers, which provide the underlying infrastructure for all of these application/data warehouse/etc. upstarts, there’s plenty of money to be made in partnering well, as Bernhardsson argued:
Let’s say a customer is spending $1M/year on Redshift. That nets AWS about $500-700k in gross profits, after paying for EC2 operational cost and depreciation. If that customer switches their $1M/year budget to Snowflake, then about $400k goes back to AWS, making AWS about $200k in gross profits. That seems kind of bad for AWS? I don’t know, we ignored a bunch of stuff here. Snowflake’s projected 2022 research and development costs are 20% of revenue, and their sales and marketing costs are 48%! For a million bucks revenue, that’s $700k. Translated back to AWS, maybe AWS would have spent $300-400k for the same thing? Seems reasonable. Now the math suddenly adds up to me. AWS basically ends up with the same bottom line impact, but effectively “outsources” to Snowflake all the cost of building software and selling it. That seems like a good deal for them!
Bernhardsson’s math seems roughly reasonable to me, and I’m a strong advocate of the idea that the cloud vendors cannot and should not build a managed service for every area of software (from call center services to databases to operating systems to … the list is seemingly endless). AWS, for example, now has well over 200 services. That alphabet soup of services (many of which compete with each other) makes it confusing for customers to know which to use for something as straightforward as running containers. (AWS has 17 different ways to do this.)
But here’s where the logic, compelling though it may be, starts to break down.
If you’re in enterprise IT, you know that your spend on applications, databases etc. dwarfs what you spend on operating systems and storage. It’s all important, but the closer software gets to the customer, and the more that software helps you to deliver a better customer experience, the more you’re going to pay.
Small wonder, then, that Selipsky, in John Furrier’s annual interview with the AWS CEO, stressed how much the company plans to focus on industry solutions.
But even looking at current AWS services like Amazon Managed Service for Kafka (MSK) or the Amazon OpenSearch Service, it’s hard to see AWS giving up on big businesses to retrench around core infrastructure. It’s not that Bernhardsson’s logic is incorrect, in other words, but rather that there’s another logic involved, and that is “revenue growth.” AWS, Google Cloud and Microsoft Azure don’t have the luxury anymore of moving back to core primitives like compute and storage. Not without setting their stock prices on fire.
SEE: Multicloud: A cheat sheet (free PDF) (TechRepublic)
Perhaps, to use Bernhardsson’s example of Redshift, over time these cloud giants will discover that their native, higher-order services keep losing out to nimbler, more focused competitors. Perhaps. That’s certainly happening in some areas already across the big clouds. But it’s also true that these cloud providers are at times delivering superior services. One very basic, but obvious, example is how each of the clouds has a superior MySQL offering than Oracle … which actually owns MySQL. (I can say “superior” with complete confidence because Oracle doesn’t actually offer a MySQL managed service, which is somewhat baffling. Just spinning that up and throwing some marketing at it should be worth a few hundred million, if not a billion. And yet….)
But even without losing to their partners, the clouds may find that they can build an even bigger business, and drive more growth, by better enabling partners. We’re nowhere near that decision at any of the clouds, but perhaps it will play out over time, on a service by service, and partner by partner, basis. We shall see.
To see if Bernhardsson’s argument will win out, watch for a cloud vendor scuppering any of their native services in favor of better elevating and supporting a partner solution. It hasn’t happened yet (to my knowledge), but when/if it does, that will be a sign that Bernhardsson may have been ahead of his time.
Disclosure: I work for MongoDB, and formerly worked for AWS, but the views expressed herein are mine alone.