Cloud Price Wars: The Joke is on Us

The current popular perception of the 'Cloud Price Wars' glosses over a key nuance that suggests public cloud provider margins remain high and each large provider is in fact incentivized to compete on everything except price.  While related non-price competition has fueled tremendous product development and spurred innovation, this misperception continues to result in marginal workloads prematurely moving to the cloud with a persistent disconnect between assumed and realized savings.

By upeslases (http://www.flickr.com/photos/upeslases/160434597/) [CC BY-SA 2.0 (http://creativecommons.org/licenses/by-sa/2.0)], via Wikimedia Commons

By upeslases (http://www.flickr.com/photos/upeslases/160434597/) [CC BY-SA 2.0 (http://creativecommons.org/licenses/by-sa/2.0)], via Wikimedia Commons

Background

The idea of the public cloud computing 'Price Wars', sometimes even referred to as the 'Race to Zero', seems to be nearly universally accepted and seldom questioned.  A Google search of 'cloud price wars' yields a healthy sample of reputable sources propagating this idea.  It sounds like a classic Coke vs. Pepsi type battle fought not with advertising, but rather with supposed cloud consumer cost savings.  This perception has permeated industry media coverage, the popular press, and vendor white papers and it continues to reverberate from boardrooms and C-suites down through IT departments and the greater developer community.  The untested assumption that the Cloud Price War story is real has helped set the stage for a rare situation where technology adoption is being pushed down from the very top strategic levels as no company can risk being "left behind."

A closer examination of the facts, however, indicates that the Cloud Price Wars might not be what they seem.  Rather, it appears highly likely that to this day the result is a wholesale transfer of profits from cloud customers to providers and vendors...

Recent Price Moves

The recent dramatic price moves that have propelled the Cloud Price War story forward are summarized in this excellent chart produced by Mark Mahaney at RBC Capital:

Source:  Mark Mahaney, RBC Capital

The Cloud Pricing Puzzle

It is well understood that the public cloud should be the low cost option as the large providers benefit from massive economies of scale, aggregate variable workloads with differing timing, and work down broad experience curves at a pace unattainable by most enterprises.  The open question that very few seem to be asking publicly is: given that the large cloud providers will end up with the low cost position, to what extent will they share those savings with customers?

Slide from Urs Hölzle’s keynote at Google Cloud Live, March 25, 2014, via Greg O'Connor of AppZero's astute post on gigaom: https://gigaom.com/2014/04/19/moores-law-gives-way-to-bezoss-law/

Taken at face value, the Cloud Price War story would indicate that the large providers are sharing all of the savings with customers.  While not obvious why they would do this, the best implied theory I can offer would be that there is an assumed land grab for customers in this crucial period of initial adoption as well as a boost ahead of competitors in scale and learning/experience curves.  

However, anecdotal evidence and industry best practices suggest that, while true value and savings are currently being realized for highly variable workloads (think US Open tennis website/mobile app) and workloads where servers can be turned off (think turning development/testing servers off at night and/or weekends and some Disaster Recovery configurations, e.g. a 'pilot light'/auto-scaling strategy), in many cases running stable, consistent workloads in the cloud leads only to head scratching as to why the bill is not lower.

Seeing the Matrix: The Real Price War

On July 31, 2013, the curtain got pulled back completely as the Cloud Price War among the major players was exposed as a quasi-theater PR exercise and at the same time a true price war broke out in earnest.  It was as though someone just showed up at a professional wrestling match and fired a rocket launcher across the ring.  Suddenly guesswork was no longer required, the missing piece of the puzzle was provided and the real situation was fully exposed and well articulated for all to see.  Finally there truly was a price war.  The most amazing part of the whole thing was that no one seemed to pay much attention.

The mid-2013 event was the change in strategy by ProfitBricks, a then relatively upstart provider with a premium cloud product and more limited geographic reach and overall scale than the larger players.  In one fell swoop, ProfitBricks defected from the cloud club by slashing prices to roughly 50% lower than large competitors and started a one company crusade to expose the true nature of margins and pricing in the industry.  It was as though they flipped a switch and became a vocal and articulate equivalent of Warby Parker, which exposed the close-to-monopoly pricing in the eyeglasses market by their very existence. 

The obvious question here is: Why did the popular Cloud Price War story not evolve to accommodate this seemingly game changing data point, especially when most necessary interpretation was already literally typed out for anyone to read?  My current thinking is that:

1. The source is inherently biased, so someone would have to be highly motivated to expend precious attention and effort to become convinced.

2. There never emerged an unbiased first follower to provide enough credibility to justify updating the collective theory of Cloud Pricing.

Cloud Market Pricing: Working Hypothesis

1. Large scale Public Cloud Computing is a natural oligopoly and will continue to be so for the foreseeable future:

  • Favorable cost position gained from economies of scale provide a significant 'moat' for incumbent large scale cloud providers, representing significant barriers to entry and putting a natural limit on the number of big players in this elite club.
  • There is no incentive for any company to try to take the entire market because the prize in a winner-takes-all scenario would have to be utility-style government regulation, which does not seem like the end game any of the top players would want to pursue from either a profit or culture perspective.
  • Further, as cloud adoption ramps up and more enterprise workloads are housed outside of the corporate data center, sound risk policy will dictate that there must be more than one provider.  Can you imagine going to the board of a large enterprise with a plan that would make survival as a going concern completely dependent on amazon.com not only staying in business but withstanding major security threats, natural disasters, etc.?

2. Market behavior observed to date is consistent with each large participant individually facing a kinked demand curve.  A key feature of this dynamic is that the current price takes on an almost magical 'sticky' quality in the short term and each firm is incentivized to stay with the pack and only move prices once significant changes in costs accumulate.

By Perfect_competition_in_the_short_run.svg: *Perfect_competition_in_the_short_run.png:Sheitan at en.wikipedia Costcurve_-_Combined.svg: *Costcurve_-_Combined.png:Trampled at en.wikipedia derivative work: Jarry1250 (Perfect_competition_in_the_short_run.svg) [CC BY-SA 3.0 (http://creativecommons.org/licenses/by-sa/3.0)], from Wikimedia Commons

  • Around since 1939, the kinked demand model implies that market forces alone are such that each individual firm maximizes profit by keeping price constant until such time that costs have fallen substantially (i.e. the 'MC3' curve in graph) and then a new 'sticky' price is set and all move together.  Important Note: No collusion is implied here, it is not necessary.
  • We know that for the core compute product, costs must be falling rapidly given the exponential nature of Moore's Law compounded by economies or scale, increased pooling of offset workload timing, etc.

Implications:

  1. Margins are larger than currently assumed and growing steadily between rounds of price cuts.  The lack of publicly available transparency to cloud margins is 100% consistent with this disparity between the implied PR impression of costs and reality.
  2. We can expect more periodic rounds of cuts with one firm initiating then the rest moving with a pack.
  3.  In the kinked demand model, price is the one lever on which it does not pay to compete, since in most cases moving price leads to less total profit.  The model would suggest that we would continue to see substantial non-price competition in the form of more and more wonderful services being layered on top of the core offering, which should remain great for cloud customers and innovation in general.

Key Takeaways:

  • The cloud will ultimately result in increasing savings over time, but customers should be aware that the savings may remain elusive until market forces encourage providers to share.  Also, beware anyone standing in between you and the provider, e.g. a reseller, as this presents one hurdle to savings reaching the end customer.
  • Boards and management should carefully examine any initiatives currently underway based on assumed savings because 'cloud is cheaper'.
  • Cloud stocks could briefly become attractive as the next round of price cut news is digested by the market, but this will likely be short-lived as the margin bonanza should erode as the nascent cloud market matures and public perception gets closer to reality.