Pierre Ferragu: The tourist in the cloud is this arrogant and bragging IT executive who likes to feel important with trendy buzzwords and attract an easy audience by explaining he will “migrate everything to the cloud by 2018”. He is a tourist: he will spend a lot of money for nothing and rapidly go back home.
Amazon’s Very Nice, But ‘All the Nonsense’ of Cloud Too Much, Says Bernstein
By Tiernan Ray
December 17, 2015, 1:34 P.M. ET
Bernstein Research’s networking equipment analyst Pierre Ferragu today offers up some thoughts regarding “all the nonsense we hear too often these days about moving 100% to the cloud,” his damning conclusion being that some people boasting of moving all IT to cloud computing such as Amazon.com‘s Amazon Web Services, are just “tourists” who like to boast.
“The tourist in the cloud,” he writes, “Is this the lunatic analyst running AI experiments on weekends, or GE’s Chris Drumgoole, COO of IT, drumming the Google (ah! ah!) and asserting he will move 90% of GE’s data centers onto the cloud,” referring to a remark by General Electric earlier this year about moving all its data centers to cloud computing.
In passing Ferragu discloses, “as far as I know I am the only tech analyst on the street actually using AWS most advanced features on a regular basis, that’s a good differentiation, isn’t it?”
The enterprise IT infrastructure market, consisting of routers and switches, security equipment, servers, and storage and the like, totals $141 billion annually, observes Ferragu, and an even higher $700 billion, if one counts all the operating expenses to run that stuff. That is out of a total annual global spend of $2.2 trillion on all information technology.
By contrast Amazon’s AWS is less than 1% of IT spending:
In 2015, Amazon Web Services (“AWS”) is on its way towards $8bn annual revenues, with a likely run rate exiting the year at ~$10bn. They likely represent over half the “serious” cloud market. Our esteemed colleague Carlos Kirjner forecasts ~39% growth for these revenues over the next 5 years. (Exhibit 4) Yes, you are getting your numbers right. It means Cloud today is < 1% of IT spending, maybe 3% in 5 years if Carlos’ optimism is reasonable enough. As a share of IT infrastructure spending we are talking 2% today, ~8% in 5 years.
Ferragu sneers at some claims of cloud computing’s enormous potential growth:
As an example of sell-side expectations, the following comment was recently brought to my attention: “we expect 50% cloud adoption in the next 3-4 years’ vs only ~2% penetration now!” Moving from 2% to 50% in 3-4 years means growing almost 3x for 3-4 years in a row!! This is simply not serious and does not reconcile with even the most bullish Cloud forecast.
Regarding Amazon’s AWS’s 70% annual growth, Ferragu concedes “Some amazing stuff is indeed happening here and underestimating it would be a major mistake.”
He cites among those amazing things “razor-thin operating costs,” meaning Amazon and Google run their infrastructure “for about a third of the cost of conventional set ups,” which is “very impressive.”
He notes automation is an important part of that, and things such as how Amazon and Google cool their data centers: “guess what? Cooling a Hyperscale data center in Finland or Washington, optimally designed for passive thermal exchanges, is way cheaper than cooling one in a lousy office building in White Plains.”
He figures with a 60% gross margin for Amazon, “Bottom line, if you buy a server and run it yourself it will cost you 100, if you do it at Amazon it will cost you 75. That’s a 25% discount.
Moreover, traditional IT can’t compete with flexibility of spinning up multiple cores at AWS, he concedes: “Let’s imagine I want to do something similar at work. I think I would have gotten that capacity available in about 18 months (to Bernstein’s IT team: don’t take this wrong guys, nothing personal – I love what you do for us every day, but let’s face it, 10,000 cores in 24 h, you can’t compete!”
That flexibility is “incredibly more valuable than a 25% cost reduction.”
Ferragu also notes the follow-on cost savings of running his own experiments in AWS:
Let me come back to my personal cloud experiment. This nice weekend on the Cloud cost me a couple of thousand bucks. Expensive enough for Mme Ferragu to politely, but firmly, ask me to put an end to this nonsense, but something I still can afford once in a while. How would that go without the Cloud? 300 36-core servers: $3m. Associated networking: $1.5m. Racks and cooling system: $250,000. Full time employees to run the mess: $80,000 a year. Electricity bill and rent: $250,000 a year. Service and maintenance contracts: $1m a year. Gosh! That’s about 3,000 times more expensive that the $2,000 I spent on Amazon [...] This is the most powerful trick of the Cloud. The average utilization rate of servers in the world is probably below 15%. No idea about Google and Amazon servers, but let’s be crazy, it might be 50%.
His conclusion after all this, though, is that eventually, companies will replace Amazon with in-house efforts once they achieve similar aspects of lower costs, flexibility, and better utilization:
If Amazon charges you 25% below your internal cost for like-for-like hardware, if your average utilization rate is equivalent to Amazon’s and if you can deploy data center automation similar to what Amazon does, guess what? Internalizing will make sense at some point. This is why I chose Zynga and Dropbox in the first place. Both started on Amazon and eventually built up their own infrastructure.
Hence, Ferragu arrives at lambasting the “tourists”:
I heard a guy one day saying “these guys are like tourists. They spend a lot of money, but they don’t stay for too long.” I can’t remember who told me that, if it is you please do reach out to claim paternity of the quote. The tourist in the cloud is this arrogant and bragging IT executive who likes to feel important with trendy buzzwords and attract an easy audience by explaining he will “migrate everything to the cloud by 2018”. He is a tourist: he will spend a lot of money for nothing and rapidly go back home. You find that kind of behavior in any organization. Check for instance our progress status on white-box migration we published ecently1and remember these guys at AT&T, Facebook and so many other companies were telling about their grand plans. Reality is most of the time way more humble.