Remarks on Marc Andreesen’s Twitter Essay, and the New Yorker’s Disruption Article by Jill Lepore
Mark Andreeson’s Twitter Storm/ Essay this morning has some unique insights. I’ve been thinking for some time now about the correlation between trends in Big Computing and Process Control Systems. Maybe it makes no sense, but as someone on the front lines of recruiting people who do these sorts of projects, I can see a tipping point.
Much of the DCS landscape of the past 20 years has been similar to large scale computing companies and their business models of the ’50s and the ’60s. Initially there were a larger variety of independent companies producing large architecture distributed systems, some focused on specific markets, some competitive based on smaller or larger systems. Over the course of time through natural marketplace consolidations, we are left with only a handful of DCS vendors, and some would argue that it’s now down to less than half a dozen or less. From where I sit, I see roughly only three or four viable competitors. I am not completely sure there won’t be less within the next few years.
Where the traditional offerings of hardware, software, and services were applied in the marketplace, each a profit center, the new standard is that hardware is a remarkably non profitable part of a project, with the profitable part of projects being the software and services. This may not be unlike Big Computing’s realization by IBM and others that providing the client front to back services in supporting the system is lucrative enough to offer the hardware nearly for free. Marc’s comments on bundling (as well as Ms Lepore’s arguments on disruption) seem to be analogous for the process controls industry since leading vendors such as Honeywell, or ABB, or Emerson would prefer to bundle a very proprietary product but make the hardware largely a gimme.
What I’ve seen over the last decade or so, is vendor consolidation and the growth of standards in network communication between devices (and a generally more simplistic approach to programming). The vast difference between the programming, maintenance, and upgrade support of these newer systems has meant that the threshold to competence on supporting these systems has meant that vendors no long hold end users at ransom. The vendors themselves have chosen some of this path by reducing costs (by using semi qualified programmers offshore). The result works, and it’s less expensive (even after the challenges of offshoring anything), but it also means that the holy priesthood of in house proprietary programmers has been replaced by lesser capable specialists who aren’t as loyal to an individual vendor.
The PLC applications industry is parallel and roughly 15 years ahead. In 1985, a user nearly required a vendor employee to write ladder logic programming on a controller for them. Today, the interfaces are easy enough that most instrument techs can read, understand and write control logic. I believe DCS isn’t far behind this trend.
Marc’s argument seems right. DEC unbundled IBM, only to later bundle (and decline), was followed by SUN who unbundled DEC only to bundle against Microsoft (and then decline), followed by Microsoft’s unbundling SUN/ IBM/ DEC, and then bundling (and then declining), and now firms like Google who are bundling, etc, etc. Professor Christiansen’s disruptive article in the New Yorker points to this same natural process.
If this holds for DCS, then as we’ve seen with the consolidation of the vendors, and then making hardware a commodity (unbundling), and later making software for the system a commodity (unbundling), the future hardly looks bright for the traditional full line control systems vendor. Instead, by virtue of driving down hardware prices, and the growing ease of software work, what little strength these vendors will have will be to create in house backend support as maintenance providers.