How ASML, TSMC And Intel Dominate The Chip Market

[soAsian]

its a pretty good report.

 

[benb]

It’s not bad.

What’s missing is semiconductor economics, the thing driving ASML, Intel and TSMC. Why they spend the money and build the fabs, at such a huge expense.

Who can answer such a big, basic question though? Who has the expertise?

 

[hist78]

"Who can answer such a big, basic question though? Who has the expertise?"

I guess everyone on this Semiwiki?

Seriously, I think there are multiple reasons for it.

 

[Daniel Nenni]

It’s not bad.
What’s missing is semiconductor economics, the thing driving ASML, Intel and TSMC. Why they spend the money and build the fabs, at such a huge expense.
Who can answer such a big, basic question though? Who has the expertise?

I think part of it is ego. Remember the "Real men have fabs" era? Intel and Samsung build fabs based on expectations. TSMC builds fabs based on customer demand which includes LTAs and prepays. Times have really changed but how do you catch a dominant market leader like TSMC? You build fabs. Thankfully IDM foundries can do this because they have big egos and hordes of cash. The rest of the pure-play fabs have neither.

The chip shortage really boosted semiconductor moral and the onshoring narrative continues. At some point in time however the semiconductor bubble will pop and it will be back to business as usual which is single digit growth and survival of the fittest. Just my opinion of course.

 

[blueone]

I think part of it is ego. Remember the "Real men have fabs" era?

Now it's "Real countries have fabs".

 

[Fred Chen]

Samsung was not covered?

 

[Daniel Nenni]

Samsung was not covered?

Clearly it should have been. Questionable report in all.

 

[benb]

"Who can answer such a big, basic question though? Who has the expertise?"

I guess everyone on this Semiwiki?

Seriously, I think there are multiple reasons for it.

So this will be partly in jest, partly in earnest:
If I was a semiconductor economist, I would start with Malcolm Penn’s 4 indicators
1. The overall economy
2. Unit sales
3. ASP
4. Capacity
So the model would be, build fabs when ASPs are high, capacity is low, unit sales are predicted to rise, and the economy is stable.

The complication is, it takes 1-2 years for investments to bear fruit, and in 1-2 years circumstances will be different. So a long term contract is an insurance policy.

Long term contracts must be plentiful for there to be such a massive uptick in fab construction? Or else there is massive risk appetite at Intel, all of a sudden?

 

[hist78]

So this will be partly in jest, partly in earnest:
If I was a semiconductor economist, I would start with Malcolm Penn’s 4 indicators
1. The overall economy
2. Unit sales
3. ASP
4. Capacity
So the model would be, build fabs when ASPs are high, capacity is low, unit sales are predicted to rise, and the economy is stable.

The complication is, it takes 1-2 years for investments to bear fruit, and in 1-2 years circumstances will be different. So a long term contract is an insurance policy.

Long term contracts must be plentiful for there to be such a massive uptick in fab construction? Or else there is massive risk appetite at Intel, all of a sudden?

Intel's current business model is to build a new fab upon market research is done or government subsidies is promised. I don't think HP, Dell, Leveno, Asus, and Acer are willing to sign a long term contract with Intel at this point.

 

[blueone]

Intel's current business model is to build a new fab upon market research is done or government subsidies is promised. I don't think HP, Dell, Leveno, Asus, and Acer are willing to sign a long term contract with Intel at this point.

It won't be this way now that they have a foundry business, but back in the day the common wisdom in Intel was that they figured out how many fabs they thought they would need for the new N process and rounded down to build or outfit one less than the marketing projection. True or not? I don't know, but Intel always did seem to be capacity limited.