Blogs

RedMonk

Skip to content

Unpacking the Oracle and AWS Rumor

I wasn’t at Defrag, but the whispers there made their way back to me quickly. My policy is to ignore these, because the probability of any single rumor being accurate is, in my experience, slight.

But given that we’re now fielding multiple inquiries about it, let me say that like my colleague I do not believe Amazon intends to sell its Web Services division to Oracle.

It is unclear where the rumor originated. Amazon is apparently reading intent into it, and it’s easy to understand why. A substantial portion of Amazon’s developer adoption and goodwill is driven by the accessible economics it established for the industry. Given Oracle’s history and its recent behavior with respect to MySQL, widely circulated rumors of an acquisition could introduce uncertainty about the longer term economics of AWS. Which is undesirable from the perspective of Amazon, clearly. And just as clearly, potentially desirable for one or more of its competitors.

Setting the intent aside, I find the prospect of an AWS acquisition problematic. We at RedMonk have regarded Amazon as a potential acquisition target for several years now. James built the case for an IBM acquisition in 2007, and I have argued in this space that Amazon is a technology vendor first, retailer second. That Amazon has been permitted to grow as significant as it is virtually unchallenged is, in fact, perplexing.

But there remain multiple difficulties with the rumored acquisition scenario.

  • Most obviously, there are the implied transaction logistics. Remember that Amazon the web services entity is, effectively, a repurposing and resale of the infrastructure that the parent retailer relies on. How then would you spin AWS out of the larger entity? There are saleable assets apart from that infrastructure, certainly: IP, personnel, customer relationships and developer mindshare. But either the underlying infrastructure is part of the sale or it must be recreated. The former would appear to be an untenable option for the retailer; it would either have to buy back time on its own infrastructure from Oracle, or it would have to buy or build an alternative. The latter, meanwhile, would substantially depress the value of the acquired assets, because Oracle would be compelled to recreate an infrastructure that Amazon has taken years to build. Even for a well capitalized organization, this is a multi-year effort in a market in which months count.
  • The obvious alternative to the problematic AWS-only deal would be an acquisition of the Amazon parent, bringing with it the AWS assets. If anything, there are more problems with this scenario. First, the acquisition cost would be excessive. Since James’ original suggestion that Amazon should be an enterprise vendor target, its share price is up nearly 4X. With a current market cap of $73.5B, it is pricy, even for Oracle. Second, assuming Oracle determined to take this approach, it would presumably be required to immediately spin out the customer facing retail businesses, which it could not be expected to manage effectively alongside its enterprise systems business. And spinning off the business would be problematic for the reasons outlined above: the operational costs of replacing the underlying technology infrastructure or leasing them from Oracle would substantially depress the value of the retail business being sold. It is unlikely that Oracle would pay a premium to acquire Amazon, then spin off the majority revenue component for pennies on the dollar.

Given sufficient incentives, it is not impossible that AWS is aquired. It’s even possible that Amazon intends to begin or already has begun the process of separating the entities to make such a transaction possible. But any acquisition scenario has to make sense for both Amazon and its shareholders.

At the moment, Amazon is a profitable retailer who is able to monetize and thereby offset its infrastructure costs via its Amazon Web Services division. Any deal for that division, therefore, must offer a premium above the value of that model. Given that it’s difficult to conceive of a deal that makes sense for both parties, then, the conclusion here is that the rumor is just that.

The more interesting question, from my perspective, is where the rumor came from. Who floated it, and why?

Disclosure: IBM, mentioned, is a RedMonk customer; neither Amazon nor Oracle are. RedMonk Analytics runs on Amazon EC2.

Categories: Cloud.

Tags: , ,

  • http://blog.gandalf-lab.com Niraj

    Why would orcl buy AWS? for the brand (that is it). The technology – they already got from Sun – Project Caroline was a better fit in the Enterprise World as compared to AWS.

    And why would Amazon sell AWS ? capitalize on investments.(if at all they want to). Not that they are going to run away from the space in any way. Cloud is a key part of their digital lifestyle strategy and touches the consumer in every way. Even if they sell they will probably relaunch AWS under some consumer specific orientation.

  • Pingback: Links 21/11/2010: KDE 4.5 and KDE SC 4.7 Plans, Fedora Elections | Techrights

  • http://twitter.com/martinbarry Martin Barry

    “Remember that Amazon the web services entity is, effectively, a repurposing and resale of the infrastructure that the parent retailer relies on.”

    If you mean to imply that AWS is sharing infrastructure with Amazon’s online store you might want to re-check that. My understanding is that they are, and have always been, completely separate.

  • http://www.monkchips.com james governor

    classic FUD then?

    or is it possible Amazon is considering a tracking stock, kind of like EMC/VMware?

    Whatever – good job for getting this down. sometimes don’t panic is the best advice.

  • http://redmonk.com/sogrady sogrady

    @Martin Barry: interesting feedback. these comments from @werner validate the opinion that it’s not strictly overprovisioning:

    “The excess capacity story is a myth. It was never a matter of selling excess capacity, actually within 2 months after launch AWS would have already burned through the excess Amazon.com capacity.”

    that said, it seems less than plausible that two centrally owned entities – not related via acquisition – have zero ties. we know, in fact, that this isn’t true. the billing and identity system employed by AWS, for example, is shared by Amazon.

    so they can’t be entirely distinct, as implied.

  • Derik

    so basically, to unbundle AMZN retail from AWS where AMZN retail runs as a cloud consumer of AWS cloud IaaS provider, is not possible?

    If so, then it couldn’t be a cloud, by definition.

    All the same, it seems both ORCL and IBM gobble up strange things. Not quite sure about the AWS even tho neither actually has a cloud thing of any substance.