SAN DIEGO - If you believe what Gartner CEO Michael Fleisher had to say in an ice-breaking keynote here at Gartner Symposium/ITxpo 2003, then the balance of market power will shift from buyers to vendors by the end of this year.
For the past two years, Fleisher has talked about how IT buyers should take advantage of the tremendous over-capacity in the market to negotiate favorable terms. "The end of this over-capacity is now within sight," predicted Fleisher on Monday. . "The rationalization of supply through continuing vendor cutbacks and consolidation will reach its crescendo by the end of this year. At that point, pricing power will begin to return to the vendors."
I'm not so sure I agree. If I'm a technology buyer, I'm not about to give up my highly leveragable position anytime soon. I doubt that an oversupply correction is about to take that position away from any buyer.
Scanning the entire technology landscape (hardware, software, networking, etc.), one can see a single trend rise above the rest: commodization.
If the technology market behaved like other typical commodity markets, I'd be inclined to agree with Fleisher. But I'm positing that technology commodities behave unlike commodities in any other market. Other natural resource-constrained commodities are not subject to the technology economies of scale that improve almost annually and often do so by orders of magnitude.
Take the application service provider (ASP) idea. Contrary to popular belief, the ASP approach to application delivery survived the dot com bust. While it hit a road bump that had to do with some ill-conceived applications of the ASP approach as well as some network reliability issues, the dust has now settled. Several ASPs with good business fundamentals and a commitment to availability are turning the entire software licensing and hardware buying models on their ears. The mere idea that companies of any size --- from small businesses to enterprises --- can roll out a mission critical application to as few or as many users as needed at a fraction of the cost of conventional software approaches and with little more than some Web browser technology and a good service level agreement (SLA) is mind-altering for many IT managers.
Some would argue that outsourcing application hosting is nothing new and that the only part of the model that has changed is the delivery vehicle (the browser). As an aside, I would argue that even the presence of a standard client that can connect to multiple applications (instead of multiple proprietary clients) is in itself a revolution in economies of scale. But what sets the ASP model apart, and what gives it the economies of scale that previous hosting models typically didn't have, is the notion of what RightNow Technologies CEO Greg Gianforte calls multiple-tenancy.
Multiple-tenancy refers to the ease with which an application host can virtually partition a single physical instance of an application infrastructure in a way that creates a private experience for each customer (the "tenant").
According to Gianforte's white paper on multiple-tenancy, "[Multiple-tenancy based] applications allow multiple customers to be hosted securely on a unified host, while individual clients still maintain control in a scalable atmosphere. Its resiliency in these tough economic times derives from the extraordinary value proposition of hosted applications."
Describing the chief benefits of the multi-tenancy hosted applications model, Gianforte says, "Compared to conventional environments, a multi-tenancy hosted application environment accelerates significantly the time-to-benefit from years to a few months or even weeks; decreases investment dramatically; and relieves the customer burden of steep investment costs related to infrastructure, licensing, deployment and integration complexities, shifting any headaches to the vendor."
With respect to supply and demand, the multiple-tenancy ASP model by itself wipes out our traditional notions of supply and pits a new world against the old world in a way that old world supplies of just about any technology are rendered entirely irrelevant. If conventional software providers refuse to embrace this change, they'll end up in the same junkyard as their predecessors, and it will be the customers (the ones who are supposedly losing their leverage) --- motivated by time-to-benefit, cost, and simplicity --- that drive them to extinction.
I could end my argument here, but there are several layers to this onion. A notion of a standards-based services oriented architecture (SOA), otherwise known as Web services --- that drives down the cost of integrating dissimilar systems while at the same time promising (a business process-oriented development environment that even a CEO can access)--could have a similar commoditizing effect on the industry. Prior to the idea of Web services, the constraining factor that resulted in a supply problem was the huge undertaking needed in order to integrate any two dissimilar systems.
Imagine, for example, that eBay makes a Web service-based auction widget --- XML schemas and all --- available to camera sellers and buyers. The resulting marketplace could scale from no participants to an infinite number virtually overnight regardless of what systems they're using. A little drag here. A drop there. The marketplace scales virtually unimpeded by anything that remotely resembles a supply issue. About the only lever eBay would have to keep an eye on would be the way its systems respond to peak loads as its marketplaces grow--which brings us to the next layer of the onion.
Today, most companies that run applications --- be they the primary user, the ASP to which that user outsources, or your business partner to whose Web services you connect --- generally choose to use their own hardware. Even when making their own choices about systems, the idea of Web services once again comes along, abstracts the underlying hardware, and blows away any supply issues because the Java and .Net middleware layers level the playing field between operating systems and the hardware they depend on. If Web services didn't come along and we were stuck forever in our stovepipes, I might be inclined to say that hardware vendors like Intel, Sun, IBM and Hewlett-Packard. might be in a position to work supply to their advantage. But now that they're competing with each other (thanks to the abstracting middleware), all bets are off.
But let's just say we were stuck with our stovepipes. Could there really be a supply problem when it comes to CPUs or storage, or might the day come when we dynamically outsource that as well (according to demand)? The world is on the verge of a major disruption, thanks to technologies like clusters and grids. As a side note, Gartner analyst John Enck predicted here at the event that cluster and grid technologies will merge into one technology by the end of this decade. Almost a year ago, I opined that MIPS (millions of instructions per second) would one day become a commodity thanks to grid technologies, and I'm still a believer. Not only are vendors beginning to talk up the notion of on-demand computing as though it were electricity (and could be billed like a kilowatt), but the management infrastructure for that sort of paradigm is slowly coming to fruition.
Just this month, HP rolled out an automated metering technology that measures the CPU utilization on its SuperDome servers down to the percentage level, and then bills the user accordingly. This sparked a few thoughts. Considering that HP is on a strategic path to base all of its systems on Intel's IA-64 technology (a part of which HP contributed to), it's only a matter of time before the idea of paying to play on heavily commoditized Intel hardware becomes reality. Second, if the application emphasis truly shifts to the middleware layer and no one cares what hardware is running underneath, it may not matter where you pay-to-play. Conceivably, not much would have to change for a Java-based server application (or, perhaps an OS-agnostic .Net) to get some CPU-cycles in on-demand style from any computer on the network should it suddenly overload its primary host. Third, if HP can come up with a kilowatt-like billable infrastructure for Superdome servers, that technology is surely extensible to grids where, for the benefit of middleware as well as native applications, we will see virtually limitless pooling of networked-based processing resources. Are you beginning to get the picture? If there was about to be a supply problem with MIPS, that problem is on the verge of amelioration thanks to several disruptive technologies that are currently in play.
Fleisher is probably right when he says the vendors are correcting their supply issues. They have no choice. They are probably hoping to gain back some of that leverage that the overcapacity gave you in the first place. But, by the time that happens, too many industry-altering business models will already be at your disposal. I believe you'll have your leverage for some time to come--perhaps forever, or until there's only one vendor left. Whichever one comes first.
