One of the clearest trends over the past few years has been Qualcomm’s dominance of the LTE modem market. I have written about it a bit in the past, and I know there is an ocean of ink being written about it on the Street coming out of the show. Rather than dwell on the details which are being covered elsewhere, I want to make a somewhat crazy suggestion.
Intel has spent the past twenty years and probably $20 billion trying to get into the mobile market. For all of that, they still have very little to show for their effort. As far as I could tell at MWC, their only design wins are in low volume SKUs at a few global OEMs or equally low volume designs at China design houses. And they continue to heavily subsidize these devices. One vendor had an Intel-powered phone in their booth that the vendor claimed would retail for $60. The large screen on that device probably costs $40 alone. If I had to guess, I would say that Intel is paying the ODM $20 or $30 per device in ‘contra-revenue’ aka marketing subsidies. Intel can afford to subsidize a lot of these phones (for now), but eventually Mediatek and Spreadtrum will get to volume in LTE (probably mid next year) and at that point, even the subsidies will not keep Intel chips competitive.
So here is my idea. Intel should exit the mobile business. They have looked to mobile as a way to fill their fabs, so abandoning mobile will leave a hole to be filled. As many have pointed out, the best way to fill that hole, would to open up their fabs to outsiders and become a 3rd party foundry. So when they announce their mobile exit, they should also announce their first major foundry customer – Qualcomm. Qualcomm needs a new foundry supplier, given their difficulty with TSMC in getting their latest Snapdragon 810 product out the door. So they could use the help of a leading foundry. In turn, Qualcomm could also abandon its nascent ARM server business. This relationship would rock the industry.
This idea is odd enough that I hesitated to publish it. But during the Qualcomm investor Q&A I heard an interesting comment. In response to a question about Moore’s Law, Qualcomm pointed out that they have a choice of multiple foundry partners for the 10nm process node. The way they said it made me think multiple meant more than three, but there are really not that many options left – TSMC and UMC in Taiwan, Global Foundries and Samsung are pretty much it. My basic premise is that Qualcomm wants to diversify (not abandon, just diversify) away from TSMC because of the 810 snafu. That leaves three obvious ones. Maybe I am reading too much into an off-hand comment, but I have to wonder if they had maybe considered Intel.
I will be the first to admit that such a deal would be so complex that it may not be operationally feasible. Technically it does not make a lot of sense. It may not even be legal. But it would certainly shake things up. Maybe its time to think the unthinkable.