Yesterday we wrote about AMD’s acquisition of Xilinx. And in walking through the history of AMD we noted that after a tumultuous history, AMD’s fortunes really turned when they spun off their manufacturing fabs into Global Foundries, a separate company. This was somewhat of a throwaway comment, but it has stuck with us, and got us to wondering. Maybe this strategy could work for another major CPU maker encountering hard times – Intel.
Intel’s problems largely stem from falling behind in the race to advance to the next chip manufacturing process. For almost its entire existence, Intel has been the leader in advancing semiconductor manufacturing technology. After all, Moore’s Law is named for an Intel founder and former CEO. But in recent years they have been struggling.
So maybe it is time to split the company in two. To be clear this is not easy, Intel’s core CPU business is deeply enmeshed with its fab operations team. But a growing share of Intel’s revenue comes from other sources, a trend only accelerated by its acquisition history. Our rough estimate is that 20%-30% of Intel’s revenue is now produced in someone else’s fabs, as they have never really been able to port acquired products like Altera’s FPGAs or Infineon’s modems into their own fabs.
So why not take the next step and split the two? This is what AMD did a decade ago. Since then AMD, the fabless chip design company, has done incredibly well. The foundry side, Global Foundries, has struggled somewhat, but is still a solid producer of trailing edge chips.
The benefits would be clear – providing coherent strategies for both sides of the split. This is American Capitalism 101, align employees’ rewards (in the form of stock compensation) with things they can actually influence, and allow management to build independent competitive advantages.
The drawbacks are less obvious, but are nonetheless formidable. A big part of Intel’s processor product strategy rests in having access to advanced production capabilities. Even having fallen behind as they are now, their products are highly capable as a result. Severing the link with the production side would prove challenging just from the product strategy side. Moreover, to really take advantage of the split they would have to learn how to port their production to other people’s fabs, a process which take years and billions of dollars.
It would also lead to huge challenges for the foundry side. This would have essentially one customer to start. They would have to learn customer service, a skill which is completely alien to them today. Every attempt they have made to open up their fabs to third party customers has failed. So again, it would be years before they could get up and running.
One benefit of this approach would be to allow for Intel’s Foundry unit to acquire Global Foundries. This would consolidate the American foundry industry into a National Champion, a term reminiscent of China’s state planning, which may be the whole point of all this. The US government, looking to respond to China’s state-organized chip development program has clearly been thinking about how to promote the US’s domestic advanced fab capacity. A stand-alone Intel fab unit would presumably be on the receiving end of that funding funnel.
To be clear, this post is entirely idle speculation on our part. Intel is clearly in trouble, not existential trouble, but far from its happy place. And we have a history of fantastical ideas of what Intel could, but won’t, do. This kind of split would be time consuming and hard. There is no reason to think Intel is actually thinking about this, but they should at least consider it.