back to article Pumping billions into data centres won't guarantee you an empire

It's not a surprise for us, in economics, to find that we've got two (or more) different processes going on, each working in opposite directions. The final result will come from the interaction of the two and we're never really sure which is going to win out. Sometimes, in a certain society, one will beat the other, and at …

  1. Pascal Monett Silver badge

    Economics is a difficult question

    If it wasn't, we'd be able to forecast it properly, instead of only being able to sometimes find a reason after the fact.

    I find this article quite interesting, but I can't help having a bit of an issue with the premise of billions not guaranteeing a return.

    Amazon, Google and others have thrown billions into making data centers and the result is showing : Amazon is apparently better off than Microsoft when it comes to cloud services.

    Cloud services are not summarized by the services you offer to your customers. You can have the greatest list of features on the market, if your cloud is down regularly, people will go for alternatives that may offer less features, but are more stable.

    Stability is what costs billions. The backend requirements for ensuring proper response times and failover is something that is quite beyond me, but I'm quite sure that, in terms of hardware, bandwidth and server monitoring, it must absolutely mind-boggling. Experience in this domain is capitalized upon, and cannot be just bought.

    So I do not see newcomers having the slightest chance at upsetting the establishment. Very, very deep pockets are required to enter this market, and when you're in, you still have to learn how to make it work the hard way - meaning losing customers every time you fail to maintain stability.

    There are three big names in the enterprise cloud space at this time : Amazon, Microsoft and IBM. I'm pretty sure we won't be adding many more to that list.

    1. ecofeco Silver badge

      Re: Economics is a difficult question


  2. thames

    Network Effects

    I don't think that economies of scale involved in operating a data centre will explain the dominance of a few cloud providers. Rather it will be the network effects of all the third party software, developers, consultants, and services that will be built around the proprietary features of those clouds. A challenger would have to reproduce all of those in order to effectively challenge an incumbent.

    It's like how Microsoft is dominant on the desktop, and Android is dominant in mobile. Apple can be a distant runner up in either of those spaces, but anybody else is a marginal player. It's the third party effects though which are the sources of that dominance.

    Cloud providers will concentrate on increasing the network effects, and making it difficult for customers to leave once they're inside. They will also work on creaming off as much of the potential profits for themselves as possible, including by carving out promising lines of business to be exclusively for their own products.

    If you believe that a free market is a good thing when it comes to economic performance, then this is a dystopian future. Economic growth will be weighed down by sub-optimal allocation of capital and poor use of resources. All in all, this is not a good thing.

  3. Ian Easson

    The Laughing Curve?

    You're actually basing this pseudo-analysis on the "Laughing Curve" (aka the Laffer Curve):

    - Which doesn't exist, because it is just a bunch of randomly scattered data points

    - Was actually created in 1974 by Dick Cheney, Donald Rumsfeld, and his deputy press secretary Grace-Marie Arnett (reference:


    1. Roj Blake

      Re: The Laughing Curve?

      I stopped reading after he claimed the Laffer Curve was true. I don't know of any serious economists who consider it a valid thing.

      1. Tim Worstal

        Re: The Laughing Curve?

        That's a fascinating statement.

        Diamond and Saez.

        That's Peter Diamond (Nobel Laureate) and Emmanuel Saez (Likely future Nobel, John Bates Clarke Gold Medal he's already got, sometime collaborator with Piketty) and in that paper they calculate the peak of the Laffer Curve (for both income and capital taxation) for the US economy.

        Which is a pretty odd thing for serious economists to do over something that no serious economist considers to be a valid thing.

        Might I gently suggest that you might want to revisit your preconceptions in this area?

        We could also look into the EU's report into the financial transactions tax (aka Robin Hood Tax). Where they show that the FTT would lower the size of the economy as compared to a future without the tax. To such an extent that total tax revenue would be lower than in the absence of the FTT. Which is exactly what the Laffer contention is: lower tax rates can, sometimes, lead to more revenue.

    2. Tony Haines

      Re: The Laughing Curve?


      It's pretty obviously a frowning curve.

    3. Naselus

      Re: The Laughing Curve?

      Much as I hate to agree with Tim, I'm not sure there's many serious economists (outside of some MMT guys, since they conceive of money and taxation entirely differently) who don't accept the Laffer curve; it's not really open to debate. There IS huge debate on where the data points actually fall (anything from 30%-70% is credible - Picketty headed up some research which points to the latter, for example), but the idea that the optimum tax rate for revenue is less than 100% isn't controversial.

  4. Anonymous Coward
    Anonymous Coward

    Computation consumes time, distance, and electricity

    There's a video out there of Grace Hopper demonstrating the length of wire that is traversed in one nanosecond - she makes the important point of how we have to take the speed of light into account when designing computers. Our current computer science training glosses over how exactly the computation gets rendered onto the silicon - since current dogma assumes a von Neumann processor, there's no point in thinking about exactly where a bit of information is when the process calls for it..

    Turns out when you scale out to the cloud this is a very very big deal.. assumptions that only cost you a few seconds or a few watts on your desktop quickly scale to hours and megawatts at cloud scale. Consider the architectural design of Facebook, and what a bit has to go through when someone interacts with the web site.. merely pressing the 'Like' button calls up at least a few roundtrips for their API and then to refresh the UI.. even though all you are transmitting is a single bit of information, you've engaged (and someone has paid for) hundreds of miles of wiring and millions of bits as this messaging works its way through the internet, to their datacenter, and through the layers of virtual machines and interpreters involved in their total process.

    The reduction-ad-absurdum of this kind of thinking is the Yo app - a power and information efficiency analysis of that application sending a single bit of information would be most interesting.

    The future will probably have higher energy costs and investors will demand that companies account for the bytes and watts that their business processes consume, and lay these out in their financial statements alongside the profit these are expected to produce. Companies will not be able to hand-wave their income statement to some mythical situation where billions of eyeballs somehow translate to billions of dollars.

  5. Salamander

    Interesting article.

    However, there is an elephant in the room which is the politics of the nation state. If the nations of the world start to demand that their citizens data always be stored in their country, then it become harder for global centralised cloud infrastructure to dominate. In essence, the Googles and Amazons of the world will have to build data centres in individual countries which will increase costs. Smaller countries will be left out in the cold as the big player would simply not want to invest in a such small countries, though they might form partnerships with local companies, providing the tech but leaving the operational details to the partner.

    This is pretty much the situation in ERP software. Due to the plethora of GAAP and local legislation, ERP software vendors have to make a choice on which countries to support and which countries to make use of local business partnerships.

  6. Micha Roon

    The money is not only in the infrastructure

    It might be true that AWS, Azure, Google, IBM Bluemix are hard to beat on infrastructure. This does not mean it is impossible. For example Digitalocean is trying and has some success.

    There are also a myriad of smaller players which offer great value for money or localized services.

    But the real money is in ease of use. Most developers don't want to struggle with infrastructure and scaling and administration. They want a CLI wich offers a single command to deploy applications. The likes of Nodejitsu and heroku are the front end to the cloud servers of AWS and others. They hedge their bets by being able to deploy to all the clouds and if one gets more expensive, it might be shunned by users.

    It would be interesting to know what part of revenue for the cloud providers comes from those Platform providers, from big enterprises and then from small companies.

  7. Anonymous Coward
    Anonymous Coward

    The reason they are spending their billions...

    .. is because by dumping cash profits into infrastructure projects means they don't pay tax on it.

    It's the same reason a shop-by-post outfit, Amazon, have gone into cloud computing. It's not because they are an IT company - it's just another way for them to pay no tax by showing no profits.

    1. Dan Paul

      Re: The reason they are spending their billions...

      Is that by putting money into a "Cloud Infrastructure" to run their shopping business, Amazon found an economy of scale that worked best if they resold excess cloud computing capacity to the public. That became a business model by itself.

      Now add a favorable tax incentive and they'd be stupid not to build even more capacity.

  8. P. Lee Silver badge

    Contestable Monopoly

    The problem is that monopoly is not contestable by the little guy.

    You could do things better and cheaper, but the incumbent sitting on a large cash pile will simply drop the price until you go away. They can even afford to pay customers to take their product, as intel was doing in the mobile market. Even if they don't engage in such practises, the threat of being able to do it is enough to stifle the competition. Intel missed out because (a) ARM was already operating where Intel wasn't and (b) the completely missed the mobile thing and the ecosystem built before it realised what was going on. If ARM's only business was phones & tablets, Intel would have squashed them. Do you think Google would be able to compete in most of its markets, such as internet provision vs Comcast, if it wasn't cross-funding from Search? There is no way to build from scratch just by being better, as Netscape found out.

  9. crayon

    "Intel wasn't and (b) the completely missed the mobile thing and the ecosystem built before it realised what was going on."

    Intel didn't miss it in the sense that they didn't know what was going on - they had been trying to get into the mobile space but missed because had no suitable products to compete with ARM. Same with Microsoft, they had been trying for years to get "Windows" (or some abomination thereof) onto mobile devices but have failed miserably.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Biting the hand that feeds IT © 1998–2021