“Commodity” is a bad word among technologists. It implies standardized, unchanging, noninnovative, boring, and cheap. Commodities are misunderstood. This post seeks to dispel some of the myths around the commoditization of IT services (i.e., the cloud).
1. Infrastructure as a Service (IaaS) Is Not a Commodity Like Oil
Yes, according to the technical definition of “commodity,” IaaS is not one. But then nor is oil, or gas, or coal, or pork bellies. None of these so-called commodities is perfectly fungible (i.e., so close to identical that a buyer is indifferent as to what is delivered), and being fungible is a prerequisite for any true commodity. When we refer to oil as a commodity, for instance, what we should mean is “Brent Crude” or “West Texas Intermediate” — that is, an exchange-traded contract for delivery of an idealized, perfectly fungible product, which in real life doesn’t really exist. Most of the oil that comes out of the ground today is pretty far off the specification for Brent or WTI. The reason these other “commodities” have been chosen as something to be traded is that it is possible to find enough consumers of the commodity who, at the right price, will accept delivery of slightly different products. This means that if IaaS is truly not on the way to becoming a commodity, then buyers really have very little choice in what they buy, and the market is therefore not competitive. If this is really the case, then why is there a “cloud price war” going on?
Oil, being a mixture of alkane molecules of various lengths (octane, decane, etc.), is obviously not fungible. Why then is oil the most heavily traded of all commodities? One reason is that oil refineries are able to take oil that is mainly composed of long-chain alkanes and “crack” them into shorter chains. How is that then different from an “IT refinery” that buys dedicated servers and sells smaller virtual machines(VMs)? Or even one that buys VMs and sells Docker containers?
The issue of physical products not being truly fungible has not stopped the trading as a commodity of other so-called commodities. As a former commodities trader, and someone running a company that for years has been actively trading cloud computing as a commodity, I don’t see the problem.
2. Commoditization Leads to Ever Lower Prices
Going way back to the price of oil in the late 1800s in today’s money, in just 40 years the price hit extreme low and high values at $15, $115, $40, $80, $25, $55, $15, and $35. Two hundred years later, in the last 40 years, we have seen $100, $20, $100, $65, $125, and $50. That is the sort of price volatility for a commodity that can be stored … which depresses volatility. Cloud computing capacity is perishable, so much more like electricity, where some markets have experienced price spikes of 800%.
The reason there is a correlation (as opposed to a causal link) between commoditization and lower prices is because exponentially increasing demand tends to encourage large-scale investment in competing services, which drives down prices and further fuels demand.
3. Moore’s Law Is the Reason for Dropping Cloud Prices
If you deconstruct the historical pricing for the cloud provider with the longest history — Amazon Web Services (AWS) — into separate prices for CPU, RAM, and HDD storage, you find that the price per CPU has not changed materially since launch. Given that Moore’s Law focuses on the number of CPU transistors in an integrated circuit doubling every two years, with an implied impact on the cost of CPU, it is clear that if there are CPU-related cost reductions for cloud providers, they are not being passed on to cloud buyers in a consistent way.
4. Trading Cloud Is All About the Spot Price
Various markets and exchanges have been set up in an attempt to give transparency to the spot price (i.e., the price for immediate delivery of confirmed capacity). One common, and reasonable, objection is that it is not hard to look up competing cloud providers’ on-demand pricing, so a service offering to automate this data collection exercise across many providers is really useful only for those looking to switch providers frequently and at scale, which is difficult if you have a lot of data to migrate.
Now, just to deal with an unfortunate confusion, AWS calls its spot price “on-demand.” These are VMs that AWS will not deliberately turn off. AWS uses “spot instances” to mean VMs that it can choose to turn off if it decides to unilaterally increase the “spot price” above the threshold set by the user. These spot instances are often referred to as a “spot market,” but as AWS is the only seller and AWS alone sets the price, this is not a market. It is a capacity management tool for AWS that allows Amazon to insulate its on-demand price (the one that should have been called a spot price) from short-term price volatility caused by supply and demand.
A true example of a cloud market or exchange would be SpotCloud, which was the world’s first attempt at a proper spot market for VMs. There are current attempts to set up broadly similar services by the Deutsche Borse Cloud Exchange and by a collaboration between 6fusion and the Chicago Mercantile Exchange. These latter two both have an initial focus on the spot market because that is a prerequisite to futures trading, which is where the real value lies.
The main reason commodity trading exists is to allow companies to manage their exposure to future price changes for those commodities. Companies have budgets, which they try to stick to, and it tends to be unhelpful if prices then change, making those budgets inaccurate, even if costs come in underbudget, as this implies that the budget was padded. Budget padding is basically what CTOs are doing when they use the current on-demand price to budget for their future cost of cloud services. The excuse that is often given is that any price reduction will be eaten up by usage growth, but this is an admission that they are neither forecasting the future price nor their expected future usage.
The way to be insulated from future price changes, and hence hit your budget perfectly, is to lock in a fixed price (which, in the current “backwardated” cloud market, should be below the on-demand price). The old-fashioned way of doing this was simply to buy your own hardware. The “cloudified” way to do this is to make a future usage commitment, with or without prepayment, directly with your cloud provider. With AWS, this is called a “Reserved Instance”; with Microsoft Azure, it is called a “subscription.” Google’s “sustained use discounts” are a variation on on-demand pricing and should not be confused with Reserved Instances or subscriptions or any other transfer of pricing risk from the customer to the provider. The best way (in my biased opinion, as this is what my company, Strategic Blue, does with its Cloud Options service), is to buy cloud services through a financial broker/reseller that is both better able and more motivated to tailor the fixed-price contract to suit the buyer’s needs, and is much more likely to be willing to buy back fixed-price contracts should a buyer’s forecast usage change.
5. Wholesale Trading of Cloud Will Not Happen
With any product or service that is on the path to commoditization, there is a repeatable market evolution:
- Buyer buys from the only seller.
- Broker helps buyer choose between multiple sellers.
- Trader steps between buyer and seller, so that trader sells to buyer on buyer’s terms, and trader buys from seller on seller’s terms.
- Trader trades with other traders to offload part of a trade with a buyer/seller to help manage risk.
- Trader trades daily with other traders so as to hide when client trades are being offloaded onto the market.
- Exchange is set up to reduce the transaction costs of trader-trader deals by allowing true “commodity” trading of a “benchmark cloud.”
For those who believe that the cloud market is still immature, it is worth observing that almost every major IT distributor — nevermind major resellers — is now acting as a trader (Step 3). Two of the world’s major exchanges are trying to build out the early parts of Step 6. The market is now ready for Steps 4 and 5 to be implemented, and the traders who set this up will then choose whether or not to use an exchange to reduce transaction costs.
Commoditization of the cloud, and the maturation of the cloud market to lead to transparent wholesale trading, is a great thing for IT consumers, and also a good thing for cloud providers large and small, but that is the subject of another post.