18 Comments

Summary:

If you throw in SaaS applications and private cloud along with public cloud into one big bucket, IBM has a claim. But in public cloud, it’s still mostly all-AWS-all-the-time.

Mushroom Cloud
photo: Flicker: Andyz

Cloud reporters got a pitch from IBM the morning after Amazon’s fourth quarter earnings announcement claiming that it — not Amazon Web Services — leads the leagues in cloud since it plunked down $2 billion on SoftLayer over the summer.

As an IBM PR rep helpfully pointed out: Amazon’s full-year “other” North America revenue, which includes AWS, was $3.9 billion while IBM’s comparable number was $4.4 billion.

He wrote:

“By definition, even if EVERY penny of ‘other’ revenue was in Amazon Web Services — and it obviously isn’t — IBM sold at least $0.5 billion more in cloud than AMZ in 2014, and IBM also grew at a faster rate (69 percent for FY 2013). That’s on $4.4 billion in cloud revenue for the year with $1.7 billion in cloud/SaaS services for IBM.” (emphasis his)

To be sure, AWS is in the public cloud biz exclusively (well except for that little CIA cloud sideline), while IBM (and HP, Dell, Microsoft, Rackspace and the rest of the known IT universe) beats the private-plus-public-equals-hybrid cloud drum. So IBM is lumping together all those cloud modes — including SaaS software into its totals. That’s fair. But maybe it’s also fair to note that many SaaS applications (Workday et al.) run on AWS but they are from third-party vendors, so that revenue doesn’t flow to AWS, but that also makes AWS a very important foundational technology for third parties.

After pushing for a wee bit more clarity on what IBM’s “cloud” revenue entails, here’s the breakdown:

  • $1.7 billion from SaaS
  • $2.7 billlion from software, hardware and services.

That second one is a rather big, indeterminate bucket of stuff. Sort of like Amazon’s “other.” IBM argues, however that its bucket, at least conforms to the cloud definition put forth by the National Institute of Standards and Technology.

So cloud revenue depends on how you define cloud. And, as far as I know the SEC still wants more clarity on how IBM itself defines cloud revenue. I bet many other legacy IT players will be getting that same request.

Amazon’s Achilles heel?

One meme that Amazon doesn’t want bandied about is that AWS is great for launching startups that don’t really know what sorts of loads to expect or will see spiky demand at the get go but once they get going and have a better grasp on their resource requirements, many often consider other deployment models.

David Mytton, CEO of Server Density discussed that notion in a December Gigaom guest post. One example he cited was Moz, which moved a lot of its work (except many batch processes) from AWS to its own hardware and saw a pretty dramatic price saving. Moz estimated it would have spent $3,200 per month on AWS vs. $668 per month using its own gear — which over the course of a year would have meant $38,400 for AWS compared to $,8096 for in-house.

Last week, Xconomy reported more on Moz’ move away from AWS referencing a blog post by Moz CEO Sarah Bird who had some damning words vis-a-vis AWS.

“Over the years, we’ve spent many small fortunes at Amazon Web Services. It was killing our margins and adding to product instability. Adding insult to injury, we’ve found the service… lacking.”

Just what was lacking in those services was not spelled out, but Bird said Moz is building out its own private cloud infrastructure in Virginia, Washington and Texas.  That “big bet” meant more than $4 million in capital lease obligations. But she said the gamble is starting to pay off.

“On a cash basis, we spent $6.2 million at Amazon web Services and a mere $2.8 million on our own data centers. the business impact [of this shift] is profound. We’re spending less and have improved reliability and efficiency. … Our gross profit margin had eroded to ~64%, and as of December, it’s approaching 74%. We’re shooting for 80+% and I know we’ll get there in 2014.”

I’ve reached out for AWS for comment but I know from past history that company execs would strenuously object to the notion that growing companies could do better moving workloads elsewhere. I will update this with AWS comments. Update: Amazon had no comment for this story.

The Structure Show!

Many of us — those without small children — probably view Disney World with something akin to disdain. If you’re going to spend all that dough, why not go somewhere real? But parents have little choice and parents who also happen to be data scientists could find some fascinating lessons in Disney’s use of customer surveillance, er consumer analytics at its theme park. That’s the case with John Foreman, chief data scientist for MailChimp who recounts his Disney World experience

Download This Episode

Subscribe in iTunes

The Structure Show RSS Feed

Other cloud computing news

With Open Compute, Facebook is saving billions and moving markets

Om’s chat with Microsoft cloud guy (and probable future CEO) Satya Nadella

Box has reportedly, secretly, filed for an IPO

5 painful lessons startups are learning during from slow death of free cloud services

Note: This story was updated at 11:05 a.m. February 3 to reflect that AWS had no comment on IBM’s cloud claims.

  1. Note that the “Moz estimated it would have spent $3,200 per month on AWS vs. $668 per month using its own gear” is per server, and given the large numbers of servers Moz have, this is quite significant! Indeed, Moz spent $7.2m on AWS last year.

    This highlights how most deploys progress – when you have a known workload then it’s cheaper to deploy your own infrastructure and leave any flexibility for public cloud. Exactly what Moz are doing.

    The difficulty with IBM, AWS, Microsoft etc all claiming they have massive cloud revenue is there’s no real definition of cloud. We know AWS only sells what we would consider true cloud services, they just don’t break it out enough. It’s more murky with Microsoft, IBM, et al because of their legacy revenue base and especially professional services which I think shouldn’t really count.

    Share
    1. You are too fast for me @david. This has been updated to reflect the per server per month cost. thanks.

      Share
  2. In the Bird quote (“On a cash basis, we spent $6.2 million at Amazon web Services and a mere $2.8 billion on our own data centers.”), that billion must be a typo.

    Share
    1. indeed it was. fixed thanks.

      Share
  3. Neither of these stories add up, and it’s for the same reason. These companies are making compromises and should be upfront about them.

    Take IBM… Is virtualization + chargeback all we need to be considered “cloud”?

    I deploy a lot of non-cloud software into cloud environments and leverage many cloud infrastructre features available from Amazon Web Services that are not available **AT ALL** from Azure, OpenStack, CloudStack, SoftLayer or Rackspace. Server failure at Rackspace? Unless you over-provisioned, you get to wait until they fix it. That used to be “fanatical support”. Now it’s wasting customers’ time and money.

    Barb, please dig into this and find out what compromises these companies are making with these architectural choices. Moz apparently has no international data centers. Who is to say they haven’t put themselves in a terrible Disaster Recovery position?

    I was excited by David Mytton’s guest post. However, found it frustratingly oversimplified to the point of being misleading. My workload is a high-performance, always-on database server with simple web front ends. We were the best case scenario for Rackspace and SoftLayer, to which they both eagerly agreed.

    But their list pricing is completely out of whack. You’re left to do some hard negotiating. They started out eager to talk. But they both literally abandoned the sales effort as it became clear there was NO WAY to deliver a reasonably similar level of service for the same price as Amazon.

    Certainly there are more cost effective options than Amazon, but it comes with real compromises. This topic needs someone to dig into the details.

    Share
    1. “Moz apparently has no international data centers. Who is to say they haven’t put themselves in a terrible Disaster Recovery position?”

      Those are not meant to be connected. My point about international data centers is that costs change dramatically outside the US and Western Europe. Disaster recovery is an area where Amazon helps companies excel and I don’t see comparable services at any other cloud provider.

      I’d love to know what Moz’s costs will be when they have to double their capacity, rather than right after negotiating the initial round with an eager competitor. I doubt it will continue at a 50% discount to AWS.

      I’m not saying they can’t do better, but it’s what’s better for THEM. Since my case is also in the sweet spot and no one but AWS can deliver, I am highly skeptical.

      Share
    2. The article I wrote in Dec was “simplistic” in the fact that it looked specifically at compute, which is essentially a commodity product now. Amazon was significantly more expensive when it came to long running compute instances vs Softlayer (and other dedicated server providers) and even more expensive compared to buying your own hardware. The example from Moz proves this point.

      What is much more difficult to compare is the supporting services that Amazon offers, which is what really differentiate them from “build your own” or any of the other cloud providers. So if you look at running your database through RDS it may well be cheaper than building your own, or other providers. I didn’t look at that. Amazon innovates much faster than any other competitor and offers many more products than the others which is why it’s hard to compare. And using these services often reduces a lot of the overhead involved.

      But when it comes to pure compute, Amazon is not cheaper. I’d be interested to see your numbers as I’m always willing to be corrected!

      Share
      1. I included numbers in my replies to your guest post on dedicated hardware. I’m giving a 3-year commitment which gives me a price of $32k for Amazon over 3 years vs $153k from SoftLayer.

        I could not have been more upfront. I sent SoftLayer and Rackspace my AWS calculator with every $ of services listed. I said simply, “Beat them.” Both were sure they would. And both simply stopped calling. No one ever came through with pricing.

        Even if they had, such a huge discrepancy made me concerned about renegotiations should I want to change or add to my infrastructure.

        Share
        1. Let me clarify that $32k and $153k are for my high-performance database server’s compute services only. Beyond compute, the story just got worse.

          Share
        2. I don’t see your comment – did you use a different username?

          Share
  4. Both companies seems to be struggling, IBM because of a declining PC market and Amazon recently decreased guidance. Things are not looking good for both of them

    http://bit.ly/AmazonGuidance

    Share
    1. Uh, you do realize that IBM is not in the PC industry? They sold that off way back in 2005.

      In fact if we are talking hardware in general. IBM only gets 5% of revenue from hardware and that is before the x86 server sell off. Once that goes through we will be looking at maybe 2% of revenue coming from hardware.

      IBM is primarily a software company and then a services company after that.

      Share
  5. Steven Dickens - IBM Monday, February 3, 2014

    Strange comment about IBM struggling because of a declining PC market considering we got out of PC years ago and have just announced we will be selling our x86 Server business to Lenovo.

    Share
    1. No joke. Its really annoying when I read comments that are so completely uninformed.

      Share
  6. Barb, do you really need to reference again the flawed study by David? You just lose credibility. Check out the case studies at AWS and other public clouds (AWS’s are at http://aws.amazon.com/solutions/case-studies/) that have customer testimonials of having saved a bunch of money when they moved to the public cloud. You will not have established on-prem vendors like VMware and IBM going all out on public cloud unless there were solid ROI cases of going to the public cloud. I used to like reading your articles. Please, do not continue to lower your credibility by referencing half baked, flawed “studies”.

    Share
    1. So vendor-supplied and written case studies about their own product are more credible than a third party study?? I don’t see that logic. It’s important to question the conventional wisdom and i talk to startups all the time who love aws but move off it at least partially when they have a better idea of their workload requirements.

      but thanks for your comment.

      Share
      1. Yes when it is customers providing the ROI numbers, not the vendor. And if you absolutely must use a third party, please use someone more credible, and a study that is more comprehensive.

        Share
  7. We all know IBM is the king of marketing spin

    Share

Comments have been disabled for this post