As mentioned in previous blogs that I have written, there really are only four players left in the semiconductor game outside of memory. It’s Intel, Samsung, Apple and Qualcomm. Of the four, Qualcomm has played the most risk-averse game of poker, not willing to make bets beyond a single penny ante. Qualcomm was satisfied for many years as TSMC’s largest customer, what could go wrong. Plenty. Like having to share the same leading edge factory capacity several times over with other sizeable fabless players (i.e. Altera, Xilinx, nVidia and Broadcom) including some who are your leading and future competitors.
TSMC can’t be faulted for tallying up every customers wafer forecast and dividing by three, four or even five to get to some reasonable expected market demand. But then nobody expected the “end of the world” economic situation in 2008-2009 followed by the Apple driven Mobile Tsunami of iPhones and iPADs that drove right through the downturn. Apple, though, had its supply chain covered with well-managed capacity build outs at Samsung and Toshiba. Vertical Integration is where we are at and Qualcomm is the only one who hasn’t figured it out.
Intel overbuilt on 22nm capacity knowing that a circuit breaker was going to trip with all their competitors tied into the same single Fab source called TSMC. Malcolm Penn of Future Horizons has a great pitch on this, which I highly recommend. The only way to avoid this trap is to return to the Jerry Sanders Real Men Have Fabs strategy. It is the way in which Qualcomm can break away from Broadcom, Marvell and Mediatek. It also is the only way Qualcomm has a shot of going mano-on-mano with Intel as the end game plays out these next 3-5 years.
Intel’s greatest leaps forward, as I witnessed in the 1990s , was when their competitors screwed up during the moment that they were making their own transition to a new process with a new product. The market jumped on the new product in a stepped function manor and demand went through the roof. Everything a day old was immediately obsolete rotting in the channels. I am thinking about the transition from Pentium to Pentium MMX in the mid 1990s as an example.
In the earnings conference call, there was a moment when Steven Mollenkopf, President and COO of Qualcomm said: “Now in some cases also, our OEM partners are, of course, working with us very closely to try to help us accelerate our own supply.” I take this to mean Apple is stepping in to open doors at Samsung in order for Qualcomm to tape out a part that will only go in the iPhone 5. It is a weak position to be in when your customer is needed to open the doors to new capacity. This is likely to be paid back with a pound of flesh.
For those who a year ago thought that the ARM camp was on its way to dethroning Intel and all the pieces were in place, it is time to adjust to the reality that having a Fab Matters, now more than ever. Qualcomm, at roughly half the sales of Intel needs to write a $5B check for a New Fab starting immediately.
FULL DISCLOSURE: I am long AAPL, INTC, QCOM and ALTR