Paul Otellini, CEO of Intel (INTC) Corporation, made a presentation at the Bernstein Technology Conference on December 5, 2012. Since the conference comments were not transcribed, I have taken the time to transcribe about ten minutes of the conference. I would recommend listening to the full audio if you have 40 minutes to spend.
Intel: transcript from the Sanford C. Bernstein presentation, from 31:30 to 40:00
Q: You made a comment about…I think the sentence was to address the [inaudible] in the room, "our strategy won't change."
PO: I think I was pretty clear on what I said.
Q: Well Like, for example, being a foundry…being open.. for a strategic customer, would that be, in your mind, not a strategic change.
PO: Uh, No, we are doing that today. We are running a small foundry business. Um, we are building up our capability….we don't want to compete directly with TSMC, that's not our business model. But for the right types of products, and not to enable my competitors, I would certainly consider it. There is a lot of stuff in the pipeline. As you know, it takes a while to move your designs over… To design them onto our process for us to bring them in-house and so forth. And our foundry customers aren't going to let on that they have moved until they have moved because it would hurt them at their current suppliers. So for obvious reasons this has been a lower profile deal than the kinds of things you see with early announcement…No, I don't think that….I think it makes sense for us to have that kind of capability, particularly as our semiconductor lead gets wider over the industry. And that allows us to then consider and take business on a value pricing model vs. cost plus pricing model, which is what the foundry industry, in general, uses. I don't think a cost plus model makes sense for us. We want get paid for the value of the performance of the transistors.
Q: I just want to clarify that, on foundry, can you state, if you had a strategic customer to do foundry, would you do ARM architecture foundry...
PO: I don't want to answer that narrow of a question, I'm sorry. For the right strategic relationship, with the right customer, we would be open to a foundry arrangement. That's as narrow as I'd like to get right now.
Q: Paul, I've got another one that I'd like to ask. We've seen and talked a little today about the challenges of maintaining Moore's Law. Challenges increasing across the industry - we've seen rising CapEx - not just for you but for many players - difficulty managing product and process roadmaps, capacity constraints, yield problems. I wonder if you could talk about how you see some of the further challenges of Moore's Law impacting you over the next years, but more importantly how you think they will be impacting some of the other major players in [inaudible] the Samsungs of the world and, I guess, as a corollary, or as an add on to that, do you think Samsung's IDM model will give them an inherent advantage over the kind of pure foundry players in the Moore's Law march.
PO: well, I do think there are a number of transitions coming up in this decade. You've got the transition to 450 at some point, you've got the transition to EUV at some point. Both of those transitions are expensive and are going to require scale. In general, every time there has been a wafer size change, in our industry, there has been half as many players on the other side of those transitions, half as many people manufacturing, half as many people designing. Because the scale drives this stuff. So, I think you will likely see the structure of the industry evolve pretty dramatically over the next 4-5 years. It's clear to me that Intel will be viable on the other side of that chasm, it's clear to me that Samsung is very likely to be viable, certainly as a memory supplier, on the other side of that chasm, although they will have to deal with EUV a generation before we do because of the memory cell. Um, how far the foundries go d across that chasm is really a function of how good they are in term of technology development. And that gets back to my view of the increasing lead we have on the industry. You talked about strained silicon before, yeah, we did that, what six years ago? Now it is pretty much common place. We did High K four years ago at 45(nm), ah.. people are barely figuring that out today. Trigate? We are on our second generation no one has figured that out, they aren't going to figure that out. Part of the thing is that…the industry..if you wind the clock way back, it was all Integrated Device Manufacturers, and it was because there was no common recipe for semiconductor process technology. Everybody kind of invented their own, therefore you would have to design your own product for your own silicon technology and, to some extent, your technology represented the product you were building. And then by the time you got to the 80-90s the recipes were pretty common, we all bought the same equipment, Intel might have been early on the transition, but it was pretty much the same. And reverse engineering through destructive analysis was very much possible, You could sort of figure it out. You get to the dimensions that we are at today, and reverse engineering destroys the part in a way that you can't analyze it. So, you know there's Hafnium in there, but you have no idea how we use it, what the proportions are. So, it's becoming more difficult for everybody to, sort of, ride the coattails, either as equipment manufacturers or as device manufacturers. So, I think that, therefore, you're likely to see much more churn in the classic fabless world than we've seen in, certainly, the last couple of decades. And that churn is manifesting itself with increased costs, you've seen a lot of the noise around that, and in poor yields, and in pricing models that are changing, where people have to pay for wafers as opposed to good die. So now the fabless guy eats the yield loss vs. the foundry eating the yield loss. All that is changing, and it's happened already, that's new. Fast forward another 2-3 years, and I think it becomes more difficult. So, our position looks very good in that model, because we will be able to bring to our shareholders the profit margins of running a foundry for our own products and the architectural margins we get already. So, we get paid twice for our chips is the way I would describe that vs. Qualcomm that gets paid once for the chips, maybe once for the IP, but they certainly have to pay someone else the foundry margins.
As this stuff gets harder, and as the benefits of the technology becomes more manifest in particular, mobile devices where form factor and power are critical, I think our lead extends there.
I think Samsung will be across the chasm as a survivor, certainly in memory, maybe in logic. The rest of it is subject to these transitions of EUV and wafer scale. When fabs cost $10 billion, which they will, it takes $20 billion to fill them at 50% margin, who's got that kind of revenue?
My thoughts on Mr. Otellini's above comments:
"And our foundry customers aren't going to let on that they have moved until they have moved because it would hurt them at their current suppliers."
This is a common sense observation that is often lost in the conversations about Intel and foundry work. The only two fabless semiconductor companies that meet the above requirements would be Xilinx (XLNX) and Altera (ALTR), since both companies make Field Programmable Gate Arrays and, therefore, are not competitive with Intel. Both companies would benefit greatly from access to Intel's 22 and 14nm processes.
After the above two companies, virtually all other large fabless semiconductor companies compete with Intel in one way or another and would, therefore, not be candidates for Intel foundry work. The Intel foundry business will, of necessity, serve primarily OEM equipment makers such as Cisco (CSCO) and Apple (AAPL). I think that Mr. Otellini was giving an intended, or unintended, hint that Apple would be moving their business to Intel and we would not have a clue about it until Apple was ready to completely cut off Samsung.
Cisco is no big secret, since the rumors seem to involve new business that is not being fabbed anywhere at this time. Altera and Xilinx are not particular secrets since the industry knows that both companies use more than one foundry and move their business around. Apple is the only potential Intel foundry customer that has a great deal to lose by early disclosure of their intentions.
"Trigate? We are on our second generation no one has figured that out, they aren't going to figure that out."
This is a very important comment for Mr. Otellini to make. He is saying that the foundries such as TSMC (TSM) will never make the transition to Trigate. Never. Perhaps this is strictly due to finances. Why would TSMC invest $10-15 billion in technology development and CapEx only to remain 1-2 nodes behind Intel at the end of the day?
I have written an article on just this subject. TSMC could just give up on the technology race. Texas Instruments (TXN) did just that, they gave up at the 65nm node and will never again be a factor in any technology below 65nm. This is the company that is the co-inventor of the integrated circuit; they simply gave up to Intel, and now have left the mobile base band and application processor business.
Any leading edge competitor of Intel (there are only two, Samsung and TSMC) would have to consider the immense un-utilized capacity for 14nm Trigate that Intel is expected to have by mid-2013. That capacity will not remain unfilled. That fact has to keep management at TSMC and Samsung awake at night. Intel could change direction without notice and make billions of dollars' worth of production capacity available to any customer with a check. That would be a devastating blow to TSMC or Samsung.
As a short-timer, Mr. Otellini's credibility could be considered higher than if he were not retiring soon. His comments in the transcribed part of the conference clearly indicate that he feels Intel has already won the end game in the semiconductor business.
I agree with him.