I admired PMC-Sierra while working at Vitesse Semiconductor (VTSS). PMC has an excellent system/architecture based approach to design which sometimes worked in Telecom but is ideally suited to more commodity and standardized markets like storage. The first post I wrote for my blog was on PMC’s acquisition of Avago’s (Agilent component spinoff) storage semiconductor business. While the acquisition looked cheap from a financial perspective, I predicted it would create strategic and pricing challenges going forward.
In telecom, Lucent’s system architecture would be slightly different than Nortel’s box which would be slightly different than Huawei’s box. PMC’s approach required customers to adopt their architecture - if it didn’t match the customer’s legacy architecutral approach they couldn’t win the business. Some customers were wary to design in PMC because it locked them into PMC-Sierra as a vendor. The upside with the PMC approach was all of the components fit together nicely and worked. Here’s something I wrote for Lightreading in 2000 on semiconductor companies that lacked an open hardware architecture:
Convincing vendors to buy chipset solutions that don’t interoperate across standard interfaces is like asking them to drink the entire pitcher of Kool-Aid. Customers do not want to be locked into a Jonestown architecture.
Rule #1 for vendors: Most of the smart people out there are not working for the same company. Realize this, and make the best of the situation by allowing each company to customize its solutions in the areas where they add the most value. Forcing them to use your “All in One” solution is arrogant: It requires customers to believe that you are smarter than everyone else — and that you will continue to be.
Storage is built on common hardware architectures, so you don’t need to cater to 3-4 different architectural approaches. Using a proprietary approach would be pure suicide in this market. PMC’s system based approach should work out because there is no doubt about what hardware architecture to use. This also means that other semiconductor companies can easily enter the market, and many are.
The reality is, in storage, software is the only differentiator going forward. Semiconductor companies that do well going forward will be the ones that give away the best software to go with their silicon. There is no differentiation in the chips anymore. If PMC can accomplish in software what they accomplished in telecom hardware, they will do well. And I simply don’t know what their storage software expertise is like.
One interesting quote in the interview caught my eye:
IBD: How long have you been making storage chips?
Bailey: About three years ago, we entered the enterprise storage business. We wanted to diversify from pure telecom, which is a feast or famine business.
In telecom, there are about 15 carriers that purchase about 90% of capital equipment. They all buy or stop buying in symphony with one another, mainly due to economic cycles. It can whipsaw suppliers. It makes for a volatile stock.
Telecom was not always feast or famine. The reality is there was one huge feast 1999-2001 and one huge famine 2001-2005. It has yet to be seen that telecom is tightly tied to economic cycles. I think what Bob really wanted to say is his stocks performance was tightly tied to the cyclical interest of investors in Telecom, and he got tired of dealing with that.
Am I the only one who believes storage is more cyclical than telecom? What is more cyclical than Enterprise technology capex? And why would it be decoupled from the telecom capex cycle, if one now exists?