Business Week has published an excellent article on the Battle of the Tech Titans in cloud. It’s well worth reading especially because of its focus on clouds enablement of innovation, however there are a couple of comments I’d like to add.
First in the discussion with Andy Jassy, SVP of Amazon’s Web Services two important points came out:
- Each day, Jassy’s operation adds enough computing muscle to power one whole Amazon.com circa 2000
- This will be a very high-volume, relatively low-margin business
The first point is not surprising to us at the LEF because whilst AMZN remains tight lipped about actual revenue and growth, we’re expecting to see in excess of 100% YoY growth. The second point we believe is carefully worded because whilst cloud is fundamentally all about utility provision and we should expect to see competitive marketplaces with exchanges, brokerage and clearing houses creating effective pricing competition, this isn’t the situation today.
From some very rough calculations, I estimated earlier this year that a large server farm with a reasonably high level of energy efficiency and a relentless focus on commodity provision should be able to achieve a total operating costs (including hardware depreciated over 3 years, network, power, aircon, operations, buildings, maintenance, software, people, opportunity cost of money etc) of around $220 per year per virtual machine of an equivalent EC2 small instance. Through discussion with some utility providers in the space, it turns out that whilst I was in the right ballpark, the cost can be even lower. Now obviously you have to consider level of utilization, difference cost boundaries for different size machines (the cost is incredibly sensitive to this for reasons of component cost and power), historical investments and economies of scale but given this, it doesn’t seem wildly unreasonable to speculate that Amazon is operating near to or below the $220 per VM per year whilst selling at $740 per year. If true, that’s a very comfortable mark-up.
I mention this because it reinforces the point made by UBS that Amazon is currently a high volume, high margin business heading towards a massive volume, low margin business. A gross margin of 50%+ is not unheard of in the commodity cloud space and if this is widespread then there is plenty of scope for a price war and significant reductions in margins. Even with this, given AWS’ growth it would still be fairly reasonable to speculate that they’ll surpass a $1 billion in revenue this year.
The second comment I wish to make is regarding private clouds. The evolution of any business activity from either custom built solutions to products or from products to utility services incurs three common groupings of risk – disruption, transitional and outsourcing risks. The transitional risks are related to the shift of one model to another and typically cover issues of governance (including data), transparency, trust and security of supply. A hybrid model is a standard supply chain management technique of combining both public and private sources of supply with the purpose of trading off certain transitional risks against economies of scale i.e. a hybrid approach will mitigate risks but it’ll cost more than a pure public play. Hybrid was used in the electricity industry, it is being used in cloud. Obviously, once we have truly competitive public cloud markets with easy switching between providers (a key focus of the NASA & Rackspace led open stack effort) then the risks change and private clouds become less important but they shouldn’t be dismissed at this moment as they provide a valid transitional function.
Unfortunately private clouds do get confused with “enterprise” clouds i.e. cloud services providing infrastructure at high levels of resilience. Private clouds are all about commodity provision i.e. volume operations of good enough components and require consumers wishing to implement highly resilient services to follow architectural approaches including scale -out (horizontal scaling) and design for failure. In short, capacity and resilience must be dealt with in the software layer. The “enterprise” cloud market is more to enable legacy systems which are often built with scaling and resilience models based upon physical hardware (i.e. multiple PSUs, hot swappable components, scale-up) or those systems with specific security / SLAs / latency requirements to move to a cloud like environment with minimal disruption. In short, they are about mitigation of disruption risks such as loss of previous investment, skillsets & political capital.
Obviously the latter incurs a significant operational costs when compared to a pure commodity play but that’s the problem with legacy architectures – you either incur the cost of redesigning the architecture as you migrate your application to a commodity infrastructure environment (i.e. public and/or private clouds) or you’ll pay higher running costs (use of enterprise clouds) or you’ll switch the entire activity to a service provider which incurs switching costs. You could of course do nothing but that creates competitive costs in itself. It’s the complexity of the benefits, risks, market segmentation and different types of costs which led many to use a trusted broker to interface between themselves and the cloud – another story for another day. Back to my main point, private and enterprise clouds often get mixed together leading some to question the validity of a hybrid model but there are clear distinctions between these and this is where I would disagree with some of the comments in the article.
In general :
Do Private Clouds have a future? Yes, they have a medium to long term benefit in mitigating transitional risks (such as issues over data governance) through the use of a commodity based model. However, be careful what you’re building and remember the impact of private clouds will diminish as competitive markets develop in the public space with easy switching between providers.
Do Enterprise Clouds have a future? Yes, but they’ll tend towards high value niches (specific classes of SLAs, latency and security requirements etc). Their role is principally in mitigating disruption risks but the transitional costs they seek to avoid can only be done at an increasing operational penalty. It should be noted that there is a specific tactic where an enterprise cloud can have a particularly beneficial role: the “sweating” of an existing legacy system prior to switching to a SaaS provider.
Do Public Clouds have a future? Absolutely, this is the long term economic model and will become dominant through the same mechanisms that has made public utility provision dominant in other industries especially as clearing houses, exchanges and brokerages form.
In general, the Business Week article was spot on and well worth the read.