John West has a good article on a phenomenon in HPC. Generally speaking there is a disconnect between chip vendors and final customers. This disconnect often means that high performance computing solutions vendor (like my day job) often have to deal with … well … interesting and exciting problems, in supply, quality, and so on.
John talks about this with respect to chip vendors, but his analysis also applies to motherboard vendors, ram makers, disk vendors, and so on.
Their business model in many cases is to get as many units into the channel as possible. They are relatively isolated from margin pressure, as they don’t directly sell to the consumer. They have far more freedom to set prices, and hold the line.
Which is part of the reason that smaller vendors who integrate are squeezed between rocks and hard places. This is why most of the system vendors are not dealing with more than single digit gross margins.
Ever try to run a company with high costs, and support costs on single digit hardware margins? It ain’t fun.
But back to the point. We do bear the brunt for crappy hardware. We test and test and test. Units come to our lab, we do the build, and we burn it in more intensely than we think our customers will. Because we cannot be sure that what we get will work, and frankly, it costs us (the smaller shops) less to fix the problems upfront than it does to ship another unit out to replace the malfunctioning one (like the large shops do). The manufacturers don’t care, they can keep shipping stuff. They are designed to do this.
We have seen the larger shops ship out full non-working clusters, and fix them by replacing every single unit. We prefer to avoid that scenario. It isn’t time or cost effective for anyone, apart from the big shops.
As noted John’s article is really a must-read.