It got me to wondering about how this might end up.
Google has been in the cloud business for plenty of time, but don’t have the same sort of mind share that AWS enjoys. I think it’s largely a function of how Google’s cloud services were delivered back when I first learned about them: BigTable and AppEngine appealed to highly technical people, but to use them meant writing your applications in a completely new way. There’s an argument for that being better suited to massively scalable workloads, but that’s not all that utility computing is good for.
AWS, meanwhile, looked a lot more like regular computers: Instances are just a development server that lives “out-there” somewhere, and that I don’t need to talk to IT to get. Developing on one of those is exactly like developing on a server under my desk. Similarly, Azure looks to me like hosted Windows servers you can scale relatively easily (up and down), with a bunch of related services like SQLServer available as well.
Because of that, it’s easier to make the conceptual leap and start using AWS or Azure. From there, you add services that you’re mostly familiar with, or quickly learn about, like load-balancing and replication, support for multiple regions, databases, object datastores, etc.
Google only added Compute Engine relatively recently: beta in June 2012 and GA only last December. On the one hand, it’s smart that they responded to clear market demand for an Infrastructure as a Service offering, but they have to play catch up to the market leader in AWS.
Economics and Competition
AWS enjoys high margins, but only relative to Amazon’s core business of retail. In Brad Stone’s book The Everything Store, Bezos is on record as saying he didn’t want to repeat “Steve Jobs’ mistake” of pricing the iPhone such that is was so profitable as to attract competition to the smartphone market.
Amazon appears to have followed a textbook penetration pricing strategy, as distinct from a skimming pricing strategy, and Bezos’ public statements certainly support this strategy as being the goal. With regular price drops along the way, it looks like AWS has used our old friend the Experience Curve to track their reductions in cost over time, allowing them to drop prices and maintain the same margins. Again, this is common for high-volume, low-margin businesses.
And here’s where things get really interesting. Amazon knows how to play the tough low-margin game, but Amazon doesn’t make much profit, possibly because of their investments, like that in AWS. If margins are squeezed, the available dollars for investment go down, reducing Amazon’s ability to out-innovate their competitors. If they’re going hell-for-leather now to stay ahead, how far ahead are they, really? The trap of penetration pricing is that if you can’t ride out a price war through being the lowest cost producer, you lose a lot of money and fail.
Compare this to Google and Microsoft. Google enjoys high margins from their core business: selling ads on the internet. They’ve enjoyed market dominance for many years, despite very serious attacks from well funded competitors, chiefly Microsoft and Yahoo. They are strongly profitable, despite having to continually invest in the same innovation arms race that all high-tech companies are involved in. Similarly, Microsoft makes scads of money from selling software, particularly enterprise software. Software, particularly enterprise software, also has high margins.
In contrast, Amazon has to invest in a bunch of physical space for its fulfilment centres to run its core business, and while there’s a lot of modern IT required to run one, Amazon (or Walmart for that matter) doesn’t need as many data centres as Google to ship boxes of Oreos around. Google already needs massive datacentres to run its core business, so what’s a few more racks? Done right, it can use up spare capacity and make the investment in fixed assets (power and cooling, physical space, etc.) pay off even more by using it for both core business and cloud services. Amazon can’t exactly drop a few dozen EC2 racks into a fulfilment centre. And Amazon needs a small army of humans to pick and pack boxes, while a modern data centre is mostly empty, automated by shell scripts from a remote Network Operations Centre.
Amazon’s greatest asset is also its greatest flaw: they specialize in “undifferentiated heavy lifting” as AWS Principal Architect Simon Elisha told me last November. They are actively trying to commodify the product so that they can own and run the marketplace. It’s what they did with Amazon.com, and its what they seem to be trying to do with AWS. They don’t much care what people sell on Amazon.com, so long as Amazon gets their cut, and that cut doesn’t have to be much so long as there are enough of them. And what better way to guarantee your cut than by being the only market in town?
But cloud isn’t really playing out that way. There are well funded competitors who have shown in the past they can be every bit as aggressive and ruthless as Amazon. Amazon doesn’t have a monopoly on talented programmers and engineers who can put together massively scalable systems.Commodities are undifferentiated, by definition. If cloud storage “just works” everywhere, then who cares where it’s stored? It could be on Amazon, Google, Azure, Box, Dropbox, iCloud, or any one of a dozen other places. The only thing preventing it from moving is switching costs or people’s laziness. The same goes for compute. There are already efforts to create “cloud brokers” that can move compute and storage around to wherever it’s cheapest, and it’s only a matter of time before some clever finance boffin creates a market in tradable derivatives for cloud resources.
I’d say that if Google and Microsoft (and others like Rackspace, though they don’t seem to be playing the price game and have differentiated with service) really take the fight to Amazon for cost leader in cloud infrastructure, Amazon’s margins on AWS will get squeezed right down, requiring more capital to be sucked out of the rest of Amazon’s businesses to invest in AWS at the rate they do now. Those other businesses are low margin as it is, so it would seem a prolonged price war would likely hurt Amazon more than Google or Microsoft.
The other potential disruptor for Amazon is if their overall growth slows. Microsoft and Google make profits while still investing in new products and services, but Amazon doesn’t. If shareholders lose faith in Amazon’s ability to grow (which increases the stock price), they’ll sell their stock and go looking for a better deal elsewhere, which will put pressure on Amazon’s ability to fund their investments. Google has nearly $59bn in cash or short-term investments, while Microsoft has $84bn. Amazon has a relatively paltry $12bn, so they are the least able to fund a shortfall with free cash, or to grow through acquisition.
Ultimately what we’re seeing is competition in action. Price is being competed down to marginal cost, and customers are reaping the benefits.
Which is all Jeff Bezos really wants, right?