Intel: New Products must generate 50% gross margin to get the green light

[KevinK]

The iPhone decision was most about the necessary CAPEX for the fabs, magnified by Job's low unit price demand. Even in retrospect, fabricating A-series chips never looked like a winning plan for Intel. Intel had the wrong cost structure for anything but high-margin (to Intel) chips.

That's the Innovator's Dilemma at work. In retrospect, taking the iPhone deal and subsequent business is pretty much the only way Intel could have kept TSMC from outflanking them in volume at the leading edge. Pretty sure that Otellini would have taken the deal if he got to see how things played out for Intel, Apple and TSMC in the future. Miss a disruption and get left behind.

The lack of a discrete GPU strategy was not a gross margin issue. It was a "we don't need them because we're going to win with integrated graphics" for clients, and datacenter GPUs were a niche, and small one at that, with annoying SIMD systolic array architectures that made finding a killer app daunting until LLM training came along. I remember when GPU-accelerated database management software was thought to be next big datacenter thing. That turned out to be a small niche. Supercomputers, which used to be the realm of systolic arrays, went to scale-out x86, even rolling over Intel's own i960 in the process. If it wasn't an x86 CPU, it wasn't a good plan. By the time datacenter GPUs got hot in the market, Intel was (IMO) too far behind to develop the product maturity to catch up. And Xe HPC is not an easy architecture to evolve. Have you ever looked at it closely? This is one complicated beast of a product line.

True, it wasn't really a margin decision per se. It was take enough of the market doing the easy-to-do, x86-adjacent high-margin stuff, that the discrete GPU remains a niche. But the problem is that the niche GPU evolved into the AI beast that ravaged customer spend on Intel's data center products.

In both cases, Intel failed to make a lower margin bet in an area that eventually disrupted their core product lines.

 

[blueone]

Kevin, I seem to almost always agree with you, but not on these two points.

That's the Innovator's Dilemma at work. In retrospect, taking the iPhone deal and subsequent business is pretty much the only way Intel could have kept TSMC from outflanking them in volume at the leading edge. Pretty sure that Otellini would have taken the deal if he got to see how things played out for Intel, Apple and TSMC in the future. Miss a disruption and get left behind.

The Apple chip decision was not the Innovator's Dilemma at work, it's that the proposition made no financial sense for Intel. Like it or not, Intel is a high-cost producer with fabs in high-cost geos. If I were PSO I probably would have made the same decision he did, to say no .

And the strategy to compete in the mobile world with the x86 Atom was just silly and uncompetitive.

True, it wasn't really a margin decision per se. It was take enough of the market doing the easy-to-do, x86-adjacent high-margin stuff, that the discrete GPU remains a niche. But the problem is that the niche GPU evolved into the AI beast that ravaged customer spend on Intel's data center products.

It was impossible to foresee the LLM training market in 2011. LLMs didn't exist yet.

In both cases, Intel failed to make a lower margin bet in an area that eventually disrupted their core product lines.

Not just lower margin, probably negative margin to be Apple's foundry at their price. I don't know what TSMC's and Samsung's pricing were to Apple at the time, but I have always suspected it was higher than the price demanded of Intel. Intel people used to tell themselves the BS story that their wafers were the cheapest on the planet. I don't think that was ever true.

 

[KevinK]

If I were PSO I probably would have made the same decision he did, to say no .

And the strategy to compete in the mobile world with the x86 Atom was just silly and uncompetitive.

Do you think that if PSO knew the full arc of smartphones and app processors as a disruptive technology, he (or you) would have still walked away from the Apple business ?

Reminder/ Christensen defines disruptive innovation as a technology that initially offers a lower-performance product or service to a niche market, but over time, it improves and eventually displaces established technologies in the mainstream market.

 

[blueone]

Do you think that if PSO knew the full arc of smartphones and app processors as a disruptive technology, he (or you) would have still walked away from the Apple business ?

I knew PSO pretty well. I had a few one-on-ones with him. I know he knew that there were different usage models for smartphones, tablets, laptops, and desktop PCs. Paul was mostly about finances and markets, but he didn't think smartphones or tablets would replace PC laptops, and of course he was right. He was, however, only comfortable with Intel being a merchant chip vendor. He never expressed interest I heard in going vertical like Apple and doing systems at any level (such as phones or tablets), and Intel never was comfortable selling software, so figuring out the right merchant chip markets was the name of the game.

I think he and I could see the potential for smartphones and tablets, but to me, and I think to him, these were obvious vertical product opportunities, and the large majority of the margins would go to Apple and its ilk. Intel was always a performance product play, no question. I always thought Intel should have a foundry division, because the proliferation of merchant fabless chip companies was making it obvious there was a growing foundry market opportunity. The foundry option was where I think Innovator's Dilemma came into play. Back then, in terms of logic chip revenue, nothing beat PCs. Nothing. And that got the fabs focused on high GHz, big chips, leading processes, and a cost be damned mentality. As a TSMC customer in the Intel product group I was a member of, it was easy to see how much easier development was with TSMC and industry IP available. (Back then, CPUs did not use industry IP, unless they acquired it. I don't know if that's still true.) In previous jobs I was a foundry customer with LSI and IBM Microelectronics. IMO, life is better with a foundry if you're in product group.

So I do think Paul had a good grasp of the emerging technologies and markets, and in retrospect so did I, but I would have made the same decision he did. It takes years and zillions of dollars to become a good foundry, then or now, and without a good financial proposition, which wasn't there with Apple back then, it would have made no financial sense. And Intel manufacturing had to cut costs and change cultures. Both difficult tasks when you've been so successful for decades.

Reminder/ Christensen defines disruptive innovation as a technology that initially offers a lower-performance product or service to a niche market, but over time, it improves and eventually displaces established technologies in the mainstream market.

I've read the book, and lived both sides of disruption. My first product development leadership assignment was on a software product for which you could see the disruption coming like a freight train, yet I was still forced to develop the disrupted version. I never forgave the idiots that made that decision, and left as soon as practical (though I did finish the product first). I was in my late 20s at the time, and I felt like a dinosaur. Years later, I was hired by Intel to be the disrupter, but in a different part of the industry. I know which one I'd rather be.

 

[Artificer60]

As a TSMC customer in the Intel product group I was a member of, it was easy to see how much easier development was with TSMC and industry IP available. (Back then, CPUs did not use industry IP, unless they acquired it. I don't know if that's still true.) In previous jobs I was a foundry customer with LSI and IBM Microelectronics. IMO, life is better with a foundry if you're in product group.

I'm curious, do you think that Intel Foundry will be able to offer Intel Products a good enough experience in the next 5-10 years to win their business? Assuming of course they survive that long. If the headcount reduction includes process engineers and equipment technicians I figure they may well be cooked.

My son works as an equipment technician for Intel in AZ. The other night with sick calls and vacations there were all of 3 techs for his toolset on site. Considering OSHA requires 2 people to perform many tasks, they could only work on a single tool at a time much of the shift since with lunch breaks and other mandated breaks there were several hours where there were not enough people to do any more than that. At least in that case, there is no room to make additional cuts. After all the voluntary severance I'm told that many groups in his fab are in the same boat.

 

[blueone]

I'm curious, do you think that Intel Foundry will be able to offer Intel Products a good enough experience in the next 5-10 years to win their business? Assuming of course they survive that long. If the headcount reduction includes process engineers and equipment technicians I figure they may well be cooked.

Perhaps I'm being unrealistic, but I expected Intel Foundry to make faster progress than they appear to be making. Or perhaps there is more actual progress towards being a TSMC competitor than I'm seeing. I just don't have a good feeling about what I'm seeing and hearing. I hope I'm wrong.

My son works as an equipment technician for Intel in AZ. The other night with sick calls and vacations there were all of 3 techs for his toolset on site. Considering OSHA requires 2 people to perform many tasks, they could only work on a single tool at a time much of the shift since with lunch breaks and other mandated breaks there were several hours where there were not enough people to do any more than that. At least in that case, there is no room to make additional cuts. After all the voluntary severance I'm told that many groups in his fab are in the same boat.

Depressing. Like I've posted before, I think voluntary severance programs are pretty much always a bad idea. What you're saying supports that fear.

 

[siliconbruh999]

Perhaps I'm being unrealistic, but I expected Intel Foundry to make faster progress than they appear to be making. Or perhaps there is more actual progress towards being a TSMC competitor than I'm seeing. I just don't have a good feeling about what I'm seeing and hearing. I hope I'm wrong.

I am hearing mixed bag as well some people say that 18A is pretty good some consider it worth not looking at. Some people say it barely competes with N3E let alone N3P .

Depressing. Like I've posted before, I think voluntary severance programs are pretty much always a bad idea. What you're saying supports that fear.

Agreed you this makes the most talented people leave many times.

 

[blueone]

I am hearing mixed bag as well some people say that 18A is pretty good some consider it worth not looking at. Some people say it barely competes with N3E let alone N3P .

18A status isn't my primary concern, it's the rest of the readiness factors that I wonder more about. The PDKs, IP availability, tools support, how efficient and easy to use is the MPW process, turn-around times, the customer IP trust policies... TSMC has been tuning these factors for a long time now. Perhaps my visibility expectations are too high too soon.