Blog Post

Confused by cloud computing price comparisons? Here they are in one easy-peasy chart

Stay on Top of Enterprise Technology Trends

Get updates impacting your industry from our GigaOm Research Community
Join the Community!

Following the pricing moves of cloud providers is a bit like managing a fantasy football team; so many players with so many services, and so little time. But RBC Capital Markets took a stab at it, thank goodness.

RBC’s formula condenses cloud services into one unit price based on “total spend per GB of RAM,” which includes storage, compute, memory, I/O and other base features. That makes it easier to compare cloud pricing across vendors. Per a research note from RBC analyst Jonathan Atkin this week, the second half of 2014 saw less price cutting than the first half — which included a round robin of competitive cuts from [company]Google[/company], [company]Amazon[/company] and [company]Microsoft[/company] in March.

As of this week, RBC calculated that [company]Rackspace[/company] charges $79 per GB of RAM per month, quite a premium over number two [company] HP[/company] at $55. Going down the list, Microsoft Azure is $35; Google and [company]IBM[/company] SoftLayer both charge $32, and public cloud leader Amazon Web Services is least expensive at $27. Over the past year, every cloud saw considerable price cuts based on this RBC metric. Rackspace unit cost per month, for example, was $98 in October 2013 (excluding support costs) compared to $79 now. AWS was $42 last October compared to $27 now.

Stay tuned because Google cut compute prices 10 percent earlier this week and that will likely spur some countermoves.

But a picture’s worth a thousand words so here’s RBC’s latest chart.

Cloud price cuts second half 2014; RBC Capital Markets
Cloud price cuts second half 2014; RBC Capital Markets

12 Responses to “Confused by cloud computing price comparisons? Here they are in one easy-peasy chart”

  1. Uncle Bob

    This is a terrible method to ascertain cost. It’s variable based on the speed of the VM and the ram. But more importantly, the BANDWIDTH isn’t factored in here AT ALL.

    That’s the most important cost point and what makes most of these solution very expensive for egress usage. It’s a smart model – allow the data in for free, but charge big when you shovel it out.

    For 1TB of data, azure costs $87. Softlayer and Rackspace include a hefty amount of free transfer, as do most bare metal providers, giving 1000TB away a month. If you factor that into the equation, Azure would charge you $8700 for what you get for FREE at Softlayer or Rackspace.

    AWS Amazon and Google also charge a hefty fee for bandwidth – it’s cheaper but in the same ballpark.

  2. Ray Nugent

    The fallacy of cloud cost comparisons across vendors is that it’s not cross vendor that counts, it’s cloud vs legacy DC or colo that counts, Those differences will be MUCH more meaningful and since enterprises already factor labor, real estate, power, depreciation etc into on prem they will be more prone to doing that for cloud as well.

  3. Actually boiling it all down to RAM is exactly how you want to track it as a commodity and know that it came out as the proxy after a very complicated analysis to find one.

  4. aaronklein14

    How do you make a price comparison like that? Just to start, for AWS, are they using On-Demand, 1 year RI, 3 year RI, Dedicated, or Spot? Or arte they blending them together — and in what ratio? Then, once we answer AWS questions, what about discounts offered by other vendors?

    Are they providing the context? Without it, this is completely useless.

    • Rich Morrow

      Agreed. Public cloud is much too complicated to condense even down to a dozen or so vectors. What would be useful is a matrix comparing different offerings and richness, prices of each. Even then, we’d need to augment with a couple venn diagrams to show overlap. Oh… and by the time you do that study, it’d be obsolete due to all the changes.

    • i’m sure you can get the whole report from RBC which includes more on methodology etc. Factoring in all the variables would be a beast of a job and as stated, this is a pretty good start and you have to start somewhere.

  5. Todd Day

    Not really an Apples-to-Apples analysis. What one gets in value add features from each vendor for the elements/costs considered varies widely. A scorecard on the features provided with said costs to compliment the above chart would be a bit more complete.