Cloud sellers race to the bottom, but buyers appear to want 'race to the top'

The chart below shows how the CPI indexes have varied over the past 18 months. The dotted lines shows this variation for our large and small baskets, assuming these market share 'weights' set in October 2015 have remained the same. The baskets represent common cloud applications, the small consisting of virtual machines, block storage and bandwidth, with the large basket adding object storage, and SQL and NoSQL as a service. As anticipated, they show a big drop in the average price of cloud, as a result of price-cutting by cloud providers.

In a commodity market, a lower price should attract more customers. In the example above, let's say provider A wins more market share because it is cheaper. The average comes down to (50% x $10) + (50% x $20) /100 % = $15, three dollars lower than previously. So even though the prices have remained the same in this example, the average price has come down as a result of the cheaper provider winning market share. If we factor in this evolving market share to the Cloud Price Index, you would expect the drop to be even cheaper – a cheaper basket price would lower the average price, and the lower average price would attract greater market share, further lowering the average. Surprisingly, this is not what we see.

In the same chart, our actual average prices (shown in solid lines), updated quarterly with new price and new market share information, are actually increasing at points. This isn't mathematical trickery – if you surveyed thousands of users who use cloud services across the providers covered in the index every quarter, the average price they pay would be seen to be rising similarly.

Cloud providers are, in general, reducing the price of their services, but providers that are not the cheapest are winning greater market share. In fact, our results suggest that a greater proportion of buyers are choosing more expensive cloud providers over the cheaper ones, thereby increasing the market share of these more expensive price cuts. The so-called price war isn't causing our average price to shift in proportion.

However, the trend is different depending on which basket of goods is chosen. The average price of our small basket of goods decreases similarly whether assessed using fixed or variable weights, suggesting that consumers are taking advantage of price cuts and, in general, don't see value in paying more for services that are, broadly, equivalent. However, the difference in the large basket of goods when assessed using fixed and variable weights is significant. This suggests that although prices are coming down, end users are choosing to pay more where applications are more complex. Essentially, what we are seeing is that end users see little value in paying more for simple services such as compute, but do see value in paying more for larger, more complex, multi-service applications.

The figure below shows the relationship between the changes in large-basket price over the past year and changes in market share – the dotted line shows the line of best fit through those points. Data points are so few simply because so few cloud providers can offer the large basket of goods specification, which does affect our rather crude analysis, but remember these points (some overlap) represent over 70% of the IaaS market in the US today.

In a commodifying market, we would expect to see a lower price deriving greater market share. The line of best fit shows that a decrease in price does generate more market share, and similarly a pricing increase causes a loss in market share. So price does does have a bearing on choice of cloud provider, regardless of the value each claims to offer. However, the correlation between price and market share is very poor; even though this general trend appears to be true, we can't say what impact a price cut of x% has in terms of increased market share. In fact, some cloud providers have decreased pricing but lost market share. So we can't make the statement 'cut prices and you'll get more market share'; other factors, unrepresented in the above chart, seem to have a greater than price alone.

This is the very concept of value – a good that is deemed more valuable by the buyer than simply being the cheapest.

Markets are complex. The above image reflects this complexity. Yes, price matters, but people are being attracted to other things – reputation, perception, geography, SLAs and product portfolios all have an impact, as do a multitude of other variables. Quantifying the impact of these variables to price is a difficult, even impossible, task. But these dynamics do give us a basis for some simple truths, and provide reassurance that individual cloud buyers are looking at the broader impact of their choice of cloud provider. And remember a loss in market share doesn't translate into a loss of revenue. The cloud market is growing, and the tide is raising all boats – just some boats more than others.

• Don't price basic services such as compute at a premium. Seriously, ask yourself how much value you can really add to a virtual machine considering the market is saturated. It's better to be competitive and still in the game, than premium and not even considered.
  
• If you can't price competitively on these services, partner with someone who can, and act as a broker.
  
• Services beyond compute appear to be regarded as value-adding, for which end users are choosing to pay a premium. Object storage, however, has been subject to intense price cuts over the past quarter, and is likely to be commodifying. However, databases, support, and crucially, a product portfolio that integrates, retain value even when prices are coming down.
  
• Don't compete in a race to the bottom unless you are AWS, Google, Microsoft, Alibaba or now, Oracle – you can't win. Focus on the value, for which we've shown end users want to pay more.
  
• This seems obvious, but is often overlooked: put yourself in the CIO's shoes. Would you rather make savings of a few percent by moving cloud providers, or have career assurance knowing your apps are available, secure and guaranteed?