Is Public Cloud a Natural Monopoly?

Allon Cohen October 18, 2017 blog, Data Storage, file services, Public Cloud

How To Reap the Benefits of Economies of Scale...and of Competition

Most of Silicon Valley’s engineers relocated to the region, and when doing so they were faced with many choices. They had to choose a home, choose a phone carrier, choose an internet provider, choose a car brand, etc., etc., etc. Indeed, the land of opportunity provides a myriad of choices to the newcomers arriving on its shores. But there is one dominant exception. When it comes to buying electricity and gas, the choosing bandwagon comes to a screeching halt. There’s but one company to choose from, Pacific Gas and Electric (PG&E), in essence a local monopoly, controlling the access of every home in the region to these vital resources.

A Natural Monopoly Close to Home

Many find this extremely surprising. What?! A monopoly? Here, at the center of technology innovations and at the heart of world’s strongest free economy? Aren’t monopolies a leftover from dictatorial regimes and late 19th century industrial barons? Are they not the enemy of economic efficiency? A tool for evil industrialists to create abnormal profits on the backs of the hardworking masses?

As modern economic theory would have it, the answer to these questions is, mostly, yes.

Yes, monopolies are evil. Though not formally, evil is defined by economists as something introducing inefficiency, as nearly all monopolies do. To maximize profits, they cause the overall economy to produce less and at higher prices. Mind you, maximizing profits is not in itself a sin. Under competitively fair conditions, with many small players, individual profit maximization actually leads to overall economic efficiencies. But monopolists are different, they can individually control prices and, left unchecked, they will do so to their advantage, at a high cost to the general public.

So, when you see a monopoly should you immediately step up to break it into pieces? Sadly no. Because there is an exception to the rule above. Under certain economic conditions, so called “natural monopolies” arise. Mind you, the positive connotation of the term “natural”, for this case, is somewhat misleading. “Natural monopolies” are just as evil as their non-natural brethren. What makes them different is that they are much harder to kill. While they introduce inefficiency, breaking up “natural monopolies” only creates even bigger inefficiencies (recall the brooms in Disney’s Fantasia?).  And so, they evolve into government regulated “necessary evils”. A sad fate for them, their customers, and the economy at large.  Which brings us to the subject of this blog: Is public cloud becoming a natural monopoly?

If you survived this far, know that the rest of this blog will be much less depressing. After reviewing the conditions for a natural monopoly to arise, we will show that we still can, right in the nick of time, divert public cloud from the dangers of transforming into one. 

So, What is a Natural Monopoly?

Natural monopolies are economic entities that suffer from an exceptionally extreme case of a conundrum called “economies of scale”. To understand why this occurs, and the danger it entails, we need to introduce the concept of average costs.

First, an Explanation of Average Costs

Consider a software start up. Producing software, its main costs revolve around the headcount (mainly of programmers producing the code)…while its main outputs are the lines of code.

When the team is very small, code production is somewhat inefficient. Perhaps the team is missing a networking or a storage expert. Perhaps the founders are good at one particular task, but not the other.  Increasing the headcount increases efficiency. Say at 20 or 50 programmers you get the highest efficiency (this is just an example of course and YMMV…).  However, grow the team further and inefficiencies of scale start to emerge.  You now need a hierarchy of managers, that do a great job, but don’t produce much code. Communication between a 100-member team becomes a new challenge, decreasing efficiency. Grow the team to a 1,000 and replication of effort begins to emerge along with things such as office politics, weak hiring choices, or team misalignments…all contributing to what economists call the law of diminishing returns. Don’t get me wrong, a 10,000-person software company produces much more code than a 1,000-person company, but it does so less efficiently.  The cost of producing yet another batch of new amazing, revolutionary, really valuable lines of code is higher for a 10,000-person company than it is for a smaller start up. Since it’s getting more expensive to produce new lines of code as the company grows, the overall average cost of producing a line of code begins to increase as well.

Now, when you plot the average cost as a function of the quantity produced, a typical company, such as the software company described above, will have the average cost curve looking as below:

Public cloud monopoly - chart 1.png

Now think of a company at a size producing the quantity Q2, as in the graph above. Break it up into two companies, each producing the quantity Q1, and the average production costs for each smaller company is lower. Aha! We just increased efficiency! Yes, this is the beauty of a competitive economy. Under perfect competitive conditions, customers and resources will flow between companies. Companies will constantly break up and converge in different sizes, until each and every company is producing at its most efficient point: the point of lowest average costs! At the same time, competition for customers will also insure that these cost savings are passed on to customers, so that no company is making an abnormal profit. With companies producing at maximum efficiency and customers receiving optimal prices, life would be beautiful. Ahhh…if only all companies had a typical average cost curve.

But, unfortunately, some don’t.

Back to Natural Monopolies

Consider a large public cloud provider. It just invested many billions on those shiny new cloud data centers. The more virtual servers it sells on the infrastructure, the more customers it can divide the initial cost between. Yes, its average cost per customer actually shrinks as the number of customers grows. All due to that large initial investment. Now, get additional customers on that public cloud, and it can leverage the fluctuations in use by its multiple users into an even more efficient use of its hardware resources. Grow that public cloud even further, and it can leverage its buying power to lower the price it’s paying the infrastructure manufacturers (such as servers, networking and storage devices), further diminishing its average costs. But wait, this is all going the wrong way. Natural monopolies break the law of diminishing returns. They grow more and more efficient with scale. Their average cost curve looks nothing like the one above. It looks much more like this:

Public cloud monopoly - chart 2.png

Think of a cloud provider producing at quantity Q3 as above. Its average costs are quite low. Try to break it up into a multitude of smaller clouds producing at quantity Q1 and you get much higher average costs! The smaller the cloud, the lower value we get from the economies of scale. (Is this inefficiency the reason many private clouds have had limited success?)

So, What’s the Problem?

If the average cost decreases with size, why not just let one very, very big cloud provider exist? Wouldn’t that be the most efficient economically?

No, no, no! Don’t give up on competition so quickly.

First, with only one vendor in existence, and no competition, there ceases to be any intrinsic incentive to pass on the cost savings to the end users. But that’s not the worst of it. Competition is also the father of invention (yes, OK, necessity is the mother). Sure, things may be sunny in the beginning, but ultimately, if only one large provider exists in a market, it has the interest to keep things in place and stamp out any cause for change.

Remember the days when there was only one big Mama Bell, controlling all telephone lines across the US? It decided that, “to avoid technical interference”, only “approved” phone handsets could be connected to its telephone lines. Coincidently, the only “approved” phone handset was manufactured by non-other than, surprise, surprise, a wholly owned subsidiary of Bell. Phones looked like the one below and hadn’t significantly changed for years. Imagine where we would be today, if someone had the power to decide that only Motorola flip phones can connect to its one all-encompassing cellular network. Yes, reach out for your smart phone and give it hug. You owe its existence to competition.

phone flipped horizontally - transparent.png

What we really want is to benefit from both economies of scale and competition.

Can we Have the Cake and Eat It Too?

Well, sort of. The thing about these diminishing average cost curves is that they tend to flatten out near the end. After all, average costs can’t go down forever. 

Consider a few companies producing at the quantity Q2 above. Their average costs aren’t significantly different than the company producing at level Q3. For all intents and purposes, as long as their sizes exceed a certain threshold, they are just as efficient as a large single provider.

So, can we have a set of large public cloud providers competing against each other, while we sit back and enjoy both the benefits of scale, and the fruits of competition?

It Depends

Remember, the key to making the competition engine work is the portability of customers and resources. It doesn’t really help me that there are multiple electric and gas companies (or baby Bells for that matter), if my house can only connect to one. I would still be tied to one provider with no real choice.

In the telecommunication world, competition was finally fully realized when universal number portability was introduced. Your phone number was a key asset keeping you tied to a provider, and once this stranglehold was lifted, true competition ensued.

And in the public cloud world? How can we ensure true competition between public cloud providers? Here too, portability is the key.  We need to make sure users aren’t tied to one provider by the shackles of data gravity. If users can’t efficiently move their data between public clouds, they are essentially tied to that single cloud provider they started off with.

Set our data free! Only with unified access to data across public clouds, enabled by an efficient cross cloud data fabric, would true competition rule.

It’s good for the users, it’s good for the economy, it’s good for the cloud providers.

Do you agree?