The thinking of Pat Gelsinger

[hist78]

"The Path Forward: Digital Innovation with Intel CEO Pat Gelsinger - The Washington Post" https://www.washingtonpost.com/wash...ital-innovation-with-intel-ceo-pat-gelsinger/

Full interview video:

My thoughts:

1. Mr. Gelsinger said when Covid-19 stopped automakers making cars, semiconductor factories consequently stopped too. This is far from the truth.

Semiconductor manufacturers did not stop their business during the pandemic. Actually they are doing very well. Some of the fabs experiencing interruption were caused by fire or weather related power outage.

2. Mr. Gelsinger said building a new fab in "Asia" is 30% cheaper than in US and 50% cheaper in China than in US.

I'm not sure his estimate is close to the real situation. First, the super majority of cost of building a fab are equipments cost. And most of those equipments are coming from the same US, Japanese, and European manufacturers. How can a new Intel US fab pays a lot more for the same equipment than Samsung in Korea and TSMC in Taiwan? Unless those fab equipment manufacturers such as Applied Materials (US), LAM Research (US), KLA Tencor (US), ASML (Dutch), and Tokyo Electron (Japan) treat Intel as a second tier customer and charge Intel higher prices? If that's the case, no matter where Intel is going to build a new fab, Intel's equipment procurement cost will be higher.

Then there are the land, building materials, water, electricity, and other utility cost. Again US is very often cheaper if not the cheapest one compare to Taiwan and South Korea. On the other hand, due to PRC's heavy subsidies, the fab building construction cost might be much cheaper in China. But can that lead to a 50% total cost reduction to build a new leading edge fab in China? I doubt it.

Additionally, Intel has easy access to super cheap financial resources in US and around the world. There is no way to cost Intel paying 30% more to build a state of the art fab than TSMC and Samsung do.

It's a different subject if Mr. Gelsinger really meant to "operate" a leading edge fab in US is 30% to 50% more expensive than other places in the world. But does that mean Intel needs long term and continuous subsidies to make up the difference? It makes sense for Washington to do so if it's for national security related products. But for consumer products, is it sustainable?

3. In the interview Mr. Gelsinger stated Intel is looking for a US site to build a leading edge fab. With the two new Arizona fabs, a new European fab, and now one more US fab, there are four fabs on the horizon. Does Intel build their capacity based on customers' demand or on speculation? If it's based on speculations, it can become a very expensive Las Vegas exercise.

 

[Andy1299]

Good comments, I think the obvious answer (and I think you were alluding to it) is that pat is deliberately misconstruing the situation in order to simultaneously gain government funding and make it harder for TSMC and Samsung gain said funding/ procure American customers.. a lot of companies do this, it’s ironic that TSMC isn’t one of them.

I personally like the globalized view of TSMC vs. Pat reigniting national industrial policy but little to be done about it. I also don’t like the blurring of China and Asia.

 

[debasis]

Pat is people's person. He is selling his ideas to US Govt. Since US Govt is giving incentives/funding to US companies, why not. He is not directly selling his ideas from nationalistic views. Its wrapped. One interesting point he made is: 'manufacturing cost vs. semicond supplychain resiliency' and it is true (we have seen that under Covid-situation). I Community college program is with nationalistic interest. Overall, he has convincing points.

 

[benb]

Interesting listening thanks hist78.
He’s talking to a young journalist but really addressing Gina Raimondo, the Secretary of Commerce, the awarder of money.
The Biden Administration is concerned about China, so he puts it on the table that Intel thinks China costs are 50% lower. He’ll take it off the table in return for, oh, $12B ($3B for 11X, $3B for Ocotillo, $3B for Portland).
30-50% lower? Like hist78 I’m skeptical, but I can tell you the Chinese do big facility infrastructure better than anyone; faster, cheaper (as a consequence of faster), and high quality. Taiwan is close to China and you can build with Chinese contractors in Taiwan, and I highly recommend it. My impression is they move from job to job to job, learning how to do things better and faster, and it shows. There is potentially billions in savings if you can use Chinese contractors. Obviously, they’re not union, but the real savings is they can do in 10 days what union guys take 100 to do. If the unions were as skillful and hard working, it would be even.

I see no reason why Chinese contractors can’t supplement union labor in the USA. It would help close the cost gap. It would help the unions, if they could work with these amazing Chinese dudes, they would learn a lot. But the money is on the table and with it, I’m certain, US union labor requirement.

OK, the facility costs are determinative of overall cost in this discussion, despite the fact the fab equipment is very specialized and costly. That is true, to a point. It takes a year or more for the ArF crystal optics to be grown in a furnace. But that is photo, and the fab is essentially photo + support equipment. The support equipment is expensive but produced quickly on an assembly line, and flown around the world in 747 freighters. It is therefore available when your facility is ready for it, just call them and they’ll deliver like Amazon (I’m exaggerating slightly; there is haggling, mind-blowing vicious haggling, over fait accompli stuff).

Each production tool requires 20 different basic facilities, before you get to the actual production materials, the chemicals that make the machine do something you want it to do. It takes years to prepare a facility to be ready to drop a tool into a spot, connect the facilities, and start doing something. Tools are an easy expense, because you can amortize immediately. Facilities are the costly cost. It takes 3-4 years of cost before you start to even start to obtain a return on the facility investment.

So that’s my theory, how $7B of equipment cost can be meh because the time to payback is short while $7B of facility cost is super painful hard costs, which China minimizes because they’re just terrifically fast.

 

[Barnsley]

Has there been any environmental reports done on the impact the new FABS in Arizona will have on the water situation there?

Western USA still looks to be burning up , where will the water be coming from for use in the FABS?

Or is there little concern for the Environmental impact?

 

[benb]

It’s not good news for Mexico or Tucson that’s for sure.
Besides Las Vegas, and Phoenix, Albuquerque has stuck a straw in the Colorado, and that affects F11X expansion plans. ABQ is basically dependent on wells now which isn’t sustainable.
The scale of the semiconductor water use is considerable, but farming use is still much larger. Semiconductor mfg don’t want to do this, but they could pay the farmers for their water and just shut down the farming. Which is what has happened in NM, directly or indirectly. That move is logical, economic, but politically unsustainable.
Nothing about Western water use has ever been sustainable. Fab expansions just highlight this, then it’s forgotten for a while.

 

[chipsntexas]

It’s not good news for Mexico or Tucson that’s for sure.
Besides Las Vegas, and Phoenix, Albuquerque has stuck a straw in the Colorado, and that affects F11X expansion plans. ABQ is basically dependent on wells now which isn’t sustainable.
The scale of the semiconductor water use is considerable, but farming use is still much larger. Semiconductor mfg don’t want to do this, but they could pay the farmers for their water and just shut down the farming. Which is what has happened in NM, directly or indirectly. That move is logical, economic, but politically unsustainable.
Nothing about Western water use has ever been sustainable. Fab expansions just highlight this, then it’s forgotten for a while.

Yes, these are legitimate concerns in AZ - also in Taiwan (& anywhere you locate a fab!). But Intel already has water recycling in AZ fabs, with timeline to achieve 100% reuse within the fabs. Still obviously a huge impact and I'm sure that new fabs will be required to reuse some or all water. But let's not act like this is exclusive to AZ or even the US - it's all over and frankly a small part of the impact.

 

[count]

I think water use concerns are overblown. Fracking in West Texas and New Mexico use way more water in similarly dry environments and people were talking about how early on it wouldn't work because of water... but the oil companies set up their own water plants and got their recycling rates up and the fears of water stress have so far been totally unfounded. Fabs have even higher recycling rates (close to 100%) and use relatively less water. This stuff about how you can't build a fab in Arizona is just plain FUD.

Arizona is the perfect place for a fab with low seismic activity, low air pollution, and less moisture in the air actually helps a lot with HVAC costs, low power costs thanks to incredibly low cost solar, plus there is a good talent ecosystem around them. It's almost the ideal spot for a fab to be perfectly honest.

 

[hist78]

Pat is people's person. He is selling his ideas to US Govt. Since US Govt is giving incentives/funding to US companies, why not. He is not directly selling his ideas from nationalistic views. Its wrapped. One interesting point he made is: 'manufacturing cost vs. semicond supplychain resiliency' and it is true (we have seen that under Covid-situation). I Community college program is with nationalistic interest. Overall, he has convincing points.

If Pat is truly a people's person, I really hope he can:

1. Stop any attack ads against Apple and make a sincere apology to Apple. Apple is a customer of Intel and it seems really childish for Intel to attack their own customers. Additionally I thought the new IFS division needs more customers not less customers.

2. Stop attacking TSMC. TSMC is a longtime and important Intel's partner. I see 100 reasons for Intel to cherish this unusual relationship started many years ago by Intel's founders and Morris Chang and I see no reason to do anything otherwise.

 

[Portland]

His own company has tuned him out.