Supercomputing 2010’s keynote was Clayton Christensen, the author of The Innovator’s Dilemma (and its numerous follow-ons), whose research, studying the demise of companies over time, is utterly fascinating (to me) and I can’t believe I hadn’t read before.
He has studied the demise of companies over time — think Digital Equipment Corporation (cut up into bite-sized chunks in 1998, the rest of which was eaten by Compaq, which was eaten by HP), SGI (a darling when Jurassic Park came out, but ultimately liquidated in 2009, assets bought by Rackable systems), and Sun (bought by Oracle). He’s observed recurring patterns of market disruption. For example, in the case of DEC, their rise was predicated on being cheaper than mainframes. As they gained more experience, the minicomputers became more reliable, but were still in the $100k range. DEC itself was subject to PCs meeting a nascent market need willing to tolerate PCs being utter crap (early-on), but at a far different price point. As PC technology evolved, it became compelling enough that few were buying VAXen.
He recounted the steel industry being a multi-tiered market with rebar (crap steel used to reinforce concrete) at the low-end and sheet steel at the high-end. Large, integrated mills, characterized by huge capital and needing to produce at least 2 million tons per year to be viable, had traditionally serviced each of these areas. In the 80s, there were fifty of these in the U.S. At the low-end end, minimills sprung up to deal with scrap steel from autos and manufacturing. Though they were more efficient, they were considered a secondary supplier but benefited by lower overhead (less skilled labor) and, capital overhead and being closer to the end consumer of their products.
In the 80s, mini-mills began expanding into rebar production, which they were able to service for 20% less than the integrated mills. The integrated mills were willing to concede the rebar business because it was only 7% margin and 4% of their total tonnage. This would let them focus on the higher-end products. For a while, the mini-mills were doing pretty well, but then because there was no higher-priced competition, the rebar became commoditized. Its price eventually decreased 20%. As Christensen says, “The minimills’ reward for victory was that none of them could make money.”
As mini-mills wanted to make money, they sought out higher-margin markets and improved their processes and capacity. In each step, many of the integrated mills were willing to exit the product tiers to concentrate on more profitable product tiers. Each time, when the higher-priced competition exited the market, competition at the mini-mill level soon commoditized that tier, reducing its price to minimal margin. Eventually, there were very few places for integrated mills to make money, and they dwindled from 50 in the 80s to two.
At each stage of the minimills’ climb up-market, an asymmetry of motivation was at work. For the minimills, the need to enter a more profitable market provided the motivation to solve the technological hurdles preventing them from producing higher-quality steel. The integrated mills were happy to leave these markets because the lower tiers in their product mix were always less profitable than products targeting higher-end customers. Eventually, of course, the integrated mills ran out of markets to flee to.
He ended by discussing a case from pp 264-266 of his newest book, “The Innovator’s Prescription,” on how Dell’s been displaced by ASUSTeK. The story begins as thus: Dell had presence in multiple levels of the market:
- Component manufacturing – small circuit boards used in a computer, usually supplementing a motherboard’s functionality. These could be things like USB boards (when not part of the motherboard), network cards, modems. They’re low-end, commodity items.
- Motherboards – the big, honking board in a computer or laptop. It houses the CPU and memory, but more sophisticated ones can include a network, USB ports and integrated video boards. When I interviewed with Dell in 1991, they still had actual engineers designing their own motherboards.
- Assembly of computers – collecting all of the sub parts (but not peripherals) that make a computer.
- Supply chain management – He didn’t specify much about this, but my recollection was reading of Dell’s legendary “just-in-time” sourcing of third-party components. One thing they also excelled at was having manufacturers drop-ship things. For example, when I ordered a desktop with a monitor, the monitor was shipped to me directly from the manufacturer while the computer came (at the same time) from Dell. Because Dell didn’t incur the additional shipping charge (getting it to Austin), or have to deal with housing inventory, it was able to offer lower prices.
- Design of systems – piecing together computers and peripherals that will result in a complete system for end-users. This includes things like knowing the 2560×1600 monitor ordered for me would not be compatible with my Dell Precision 9600M laptop because the latter has a “fake” Dual-DVI plug and not wasting this end-user’s time. Just saying.
Early on, ASUSTeK was a small company that Dell outsourced some of its components to. They approached Dell with the proposition, “Hey, we’ve done good work on components for you. Motherboards are really a bunch of components, so whattaya say we build these for you at a 20% savings.” Dell’s business folks, having attended any business school around, run the numbers in Excel and realize, yep, this makes sense. Motherboards aren’t a high-margin item. By outsourcing these for 20% less, revenues are unaffected, but Dell can increase its effective profit margin by concentrating on higher-margin items. Even better, they can get assets off the books. It’s win-win!
As time progresses, this cycle repeats itself down through and including the Design of Systems. Now, Dell just has its brand left, outsourcing everything else to ASUSTeK. ASUSTeK, having grown larger and more efficient, can now proposition Best Buy. “Hey, we have experience building reliable computers from components up through system design. If you contract with us, we can make any brand you want — including your own — at 20% savings.”
Christensen points out that, as in the case of the steel industry, at no time in all of this can you say that Dell’s management is making an obviously bad decision.