Charles Fitzgerald, Managing Director of Platformonomics, placed a thoughtful piece on Seeking Alpha examining the positions of the leading cloud providers including Amazon, Microsoft, Google, IBM, and Oracle. It's no surprise that taking a strong position like this, Fitzgerald's story is capturing attention.
Fitzgerald’s contention is that to understand who the cloud computing leaders are, you simply need to follow the capital expenditures (CapEx) each company has made in cloud infrastructure over the last decade. When you do, the companies that Fitzgerald says spend the most on advertising are definitely not the leaders. Says he,
In cloud computing, CapEx is the ultimate form of putting your money where your mouth is because no amount of jawboning alone will conjure up data centers or pack them with millions of servers.
A good point but is it enough?
Fitzgerald’s laggards are IBM and Oracle. Each company has spent huge sums of CapEx recently but IBM’s spending has declined since 2013 and while Oracle has made significant increases in CapEx since 2014, Google, Amazon and Microsoft have spent much more to build what Fitzgerald calls hyper-scale clouds. It even gets worse if you’re a fan of Oracle because Fitzgerald also says,
Amazon, Google, and Microsoft each spent more on CapEx in 2017 than Oracle has in its entire history.
Even accounting for Oracle’s long history and inflation, that’s a sobering thought. But what’s the context for all this?
We’re witnessing the birth and growth of an information utility which I’ve discussed here before. But it would be wrong to assume that the vendor that captures the greatest market share and builds the most hyper of hyper-scale clouds will be the winner.
Some are building cloud infrastructure to accommodate their legacy customers while all are trying to capture a slice of the emerging market for cloud services. Three good questions we should all be asking include
- Haven’t we been here before?
- And what were the results?
- Will this time be different?
Let’s take a look.
At the turn of this century, analysts and media got all hot and bothered about cloud computing and the Internet. Companies built out hosting capacity to accommodate all of the demand that was sure to come. They built it but contrary to a meme of the time, they (customers) didn’t come. Phrases like “dark fiber” entered the lexicon to describe new fiber optic cables that were not carrying signals and that were hence dark. Also, server farms built on venture capital went unused and eventually their owners went belly-up.
Those entrepreneurs were early and being early is sometimes equivalent to being wrong. The future they tried to mid-wife wasn’t here yet, largely due to technical constraints. In order to make it all work, vendors needed to rewrite software or they needed to write off systems with life left. No one wanted to do that but cloud computing survived because the companies that built their software to be cloud-native prospered.
To its credit, Oracle was the exception, building out its cloud applications aka Fusion while continuing development of its legacy systems.
If you look at the graphs of CapEx spending, investment in the modern cloud by the market leaders started after the sub-prime financial meltdown. Around 2011 and 2012 cloud investment began in earnest based on Fitzgerald’s numbers.
But keep in mind that we track three kinds of cloud products and Fitzgerald’s analysis is primarily about infrastructure (IaaS). Google and Amazon primarily provide infrastructure while Microsoft has a significant stack that includes platform and applications. Each has a database project that has been in the works for ages and while you can’t rule out one or both bringing a database to market, there’s a coals-to-Newcastle aspect of such a move.
Of the three product classes, infrastructure is the only one that’s built with hard goods, the others are software whose marginal cost of delivery is zero. Unlike building a second or nth cloud datacenter, the cost of providing the second and subsequent copy of any software is nugatory compared with development. Hence a vendor could not lead at all in infrastructure and still capture significant cloud market share.
More importantly, there is a real possibility of overbuilding infrastructure when companies compete. You can trace a very interesting history of overbuilding all the way back to railroads whose differentiators at one point were the gauges of their tracks. We know how that worked out. Badly.
What runs on the deployed infrastructure is vitally important and there’s been little discussion here. That discussion leaves Google and Amazon at a disadvantage when contemplating vertical integration, compared to Microsoft, Oracle, and possibly IBM.
If you believe, as I do, that an information utility is evolving, then the build-out of infrastructure is one vital element of the rollout. But what’s missing is the admission on all sides that a utility isn’t a monopoly, it’s an oligopoly made up of primary producers, transport, maintenance and more.
The early leaders in infrastructure are already opening their infrastructure to those with software and less infrastructure. SAP, for example, has made clear it is not in the infrastructure business and holds a long-term relationship with AWS and now with both Google and Microsoft. That process will continue making the oligopoly a reality.
Netting this all out, there are leaders and laggards in the early cloud derby when measured by infrastructure CapEx. But infrastructure alone isn't enough.
The story of CapEx is not the last word on which company will lead the information utility market. More likely there will be multiple leaders in multiple categories each concentrating on an aspect of the utility market just as today there are electricity generators, transmission companies and others involved in delivering power.