3 ms·

You would still outsource running the fab to Intel, just have a government body overseeing it.

by rjsw 1y ago

chatmasta 1y ago
This won’t solve the biggest challenge/problem, which is lack of talent for staffing the fab. In Arizona they had to import a bunch of temp workers from Taiwan just to train the locals. We don’t have the skills. If you want to involve the government, maybe we should be training units of the military to make their own chips.
- jurking_hoff 1y ago
  [dead]
- ecshafer 1y ago
  How many workers do we need? TSMC has 80k workers all in. 15-24 there is about 3 million US NEETs. I think that we could have a 10 year plan to get 80k people with EE/CE/NanoTech Eng/ChemEng/etc degrees. Make college free, or hell even a stipend for people to pursue "Critical Careers", heavily fund Phds in the same areas for US Citizens, and we will see workers with the skills.
  - nemothekid 1y ago
    This is the wrong way to look at it. The people in the US that you would want working in semiconductors aren't NEETs. They are at Facebook/Google/Jane Street/Citadel/McKinsey. If I am a capable person working on delivering node improvements dealing with smaller and smaller challenges as the physics issues become quantum - I will eventually start to ask myself: why am I working on the hardest physics problems in the private sector for 150k/yr, when I can transition to Facebook or Jane Street, work equally (if not less) as hard and make 500k/yr? The US has plenty of smart people. I'd argue more that the wealth inequality gap makes it _incredibly_ difficult to justify working for less, even in a field you love, when you can make top 1-4% of income doing something else.
    - chatmasta 1y ago
      Is that true? I thought a lot of semi-conductor work is borderline blue-collar factory work and physical labor. What you’re describing is in the R&D area and also not physically dependent on being colocated in a fab. So we should have an easier time finding that talent, although we’re probably underpaying them now, as you point out.
      - nemothekid 1y ago
        The most salient issue with Intel in the past 10 year was their constant delay of the 10nm node process. While TMSC was constantly pushing down the Node size, Intel struggled and ceded a lot of ground to AMD & Apple. At the same time Intel struggled to develop a competitive 5G radio, and GPU. These are all downstream of R&D. If your fab cannot shrink it's node size, then you won't get the most profitable orders.
        chatmasta 1y ago
        I’m not sure it would have made a difference if they hit 10nm faster. Apple has always wanted to make its own chips. Some marginal efficiencies wouldn’t dissuade them from investing in themselves. And it’s not like Intel was going to start a consumer PC business…
        SlowTao 1y ago
        Yep, Apple buying PA Semi was an indicator of their intention all those years back. I suspect intel was just hoping they would fail at their ambitions.
        SlowTao 1y ago
        Hey... that Arc GPU is the one silver lining to what intel is doing recently. Is it perfect, far from it but it ain't bad.
        vel0city 1y ago
- luma 1y ago
  That’s the specific problem a well organized governmental body can address. Funding research through university and private R&D lab grants and spreading/licensing the knowledge to domestic companies. Fund grants for universities and scholarships to spread that further. Invest in multiple startups attacking the problem at different levels in the stack (semi manufacturing has a LOT of inputs). There are loads of ways you can effectively steer an economy to stimulate growth while also meeting geopolitical needs through effective and liberal application of the national purse. DOGE seems to have different ideas about the value of such programs.

You would still outsource running the fab to Intel, just have a government body overseeing it.

by rjsw 1y ago

You would still outsource running the fab to Intel, just have a government body overseeing it.

This won’t solve the biggest challenge/problem, which is lack of talent for staffing the fab. In Arizona they had to import a bunch of temp workers from Taiwan just to train the locals. We don’t have the skills. If you want to involve the government, maybe we should be training units of the military to make their own chips.

jurking_hoff 1y ago
[dead]
ecshafer 1y ago
How many workers do we need? TSMC has 80k workers all in. 15-24 there is about 3 million US NEETs. I think that we could have a 10 year plan to get 80k people with EE/CE/NanoTech Eng/ChemEng/etc degrees. Make college free, or hell even a stipend for people to pursue "Critical Careers", heavily fund Phds in the same areas for US Citizens, and we will see workers with the skills.
- nemothekid 1y ago
  This is the wrong way to look at it. The people in the US that you would want working in semiconductors aren't NEETs. They are at Facebook/Google/Jane Street/Citadel/McKinsey. If I am a capable person working on delivering node improvements dealing with smaller and smaller challenges as the physics issues become quantum - I will eventually start to ask myself: why am I working on the hardest physics problems in the private sector for 150k/yr, when I can transition to Facebook or Jane Street, work equally (if not less) as hard and make 500k/yr? The US has plenty of smart people. I'd argue more that the wealth inequality gap makes it _incredibly_ difficult to justify working for less, even in a field you love, when you can make top 1-4% of income doing something else.
  - chatmasta 1y ago
    Is that true? I thought a lot of semi-conductor work is borderline blue-collar factory work and physical labor. What you’re describing is in the R&D area and also not physically dependent on being colocated in a fab. So we should have an easier time finding that talent, although we’re probably underpaying them now, as you point out.
    - nemothekid 1y ago
      The most salient issue with Intel in the past 10 year was their constant delay of the 10nm node process. While TMSC was constantly pushing down the Node size, Intel struggled and ceded a lot of ground to AMD & Apple. At the same time Intel struggled to develop a competitive 5G radio, and GPU. These are all downstream of R&D. If your fab cannot shrink it's node size, then you won't get the most profitable orders.
      - chatmasta 1y ago
        I’m not sure it would have made a difference if they hit 10nm faster. Apple has always wanted to make its own chips. Some marginal efficiencies wouldn’t dissuade them from investing in themselves. And it’s not like Intel was going to start a consumer PC business…
        SlowTao 1y ago
        Yep, Apple buying PA Semi was an indicator of their intention all those years back. I suspect intel was just hoping they would fail at their ambitions.
      - SlowTao 1y ago
        Hey... that Arc GPU is the one silver lining to what intel is doing recently. Is it perfect, far from it but it ain't bad.
        vel0city 1y ago
luma 1y ago
That’s the specific problem a well organized governmental body can address. Funding research through university and private R&D lab grants and spreading/licensing the knowledge to domestic companies. Fund grants for universities and scholarships to spread that further. Invest in multiple startups attacking the problem at different levels in the stack (semi manufacturing has a LOT of inputs). There are loads of ways you can effectively steer an economy to stimulate growth while also meeting geopolitical needs through effective and liberal application of the national purse. DOGE seems to have different ideas about the value of such programs.