Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/whats-the-quickest-way-to-address-the-semiconductor-market-supply-crunch.14313/&_debug=1
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021370
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

What's the quickest way to address the semiconductor market supply crunch?

martinmason

New member
What's the fastest way to increase capacity and address the semiconductor supply crunch??

We all know that more FABs are slow and expensive.

Additional equipment takes many quarters to buy, build and commission (assuming you have the clean-room space)

Yield improvements help - but most processes are already far down the yield optimization curve.

Can we make more more chips on the same tools by being smarter and more efficient in the semiconductor process control?

I'd love to get the communities thoughts on smarter semiconductor processing with AI/ML and in-situ metrology to drive process control based throughput (and yield?) efficiencies?

A 3% improvement in critical path throughput on a 30K wfr/month fab could be an extra ~900-1,000 wafers a month. What is being done to improve throughput on the existing tools and processes? Where is the critical path pinch point for throughput in most fabs?
 
I think you'll find the leading foundries are already doing yield optimisation using "AI/ML and in-situ metrology" and have been for years, before these became trendy fixes-everything buzzwords... ;-)
 
I don't buy into the semiconductor shortage narrative. There is a COVID related correction happening which will resolve itself by the end of the year. Today there is panic buying all through the supply chain.

When COVID first hit my wife and I amassed a 30 day food an water supply. Not long after the grocery store shelves were empty because others did the same.

Automotive companies for example are going from Just in Time (JIT) to stocking chips. Distributors and brokers are stocking up as well hoping to profit from the shortage narrative with price increases for JIT deliveries.

"No matter how expensive, they'll buy what they can get," one auto industry analyst said.

TSMC Chip Coverage.jpeg


Only 4% of TSMCs chips are automotive. There is plenty of capacity for automotive once they fully embrace FinFETs. TSMC has already qualified their N5 process for automotive which is a big signal for the industry: Move to FinFETs if you want chips!


Bottom line: It's a correction. We have seen this before in the semiconductor industry and we will see it again, absolutely.
 
I'm with Daniel on this one. Industry was fine in 2019, there were no shortages.

2020 some companies cancelled orders, their capacity was reallocated.

2021 we have a demand catch up (and then some), now companies, especially those that cancelled orders before, can't get the chips they want.

The fabs are already growing capacity are a pretty fast rate, in the meantime prices will go up, demand will moderate again over the next year or two and people may well be talking about a semiconductor glut.
 
I think much of the auto chip shortage is caused by TI Texas freeze power outages. TI is number one world wide analog chip vendor. I do not know about fabs. Chemical plants took weeks to restart caused by the way the freeze caused outages.
 

What's the quickest way to address the semiconductor market supply crunch?​


Move your design to FinFETs:

 
The supply and demand of different nodes are different. 7nm could be in shortage at least to mid 2022 while 28nm 40nm shortage shortage will be resolved earlier.
 
I don't buy into the semiconductor shortage narrative. There is a COVID related correction happening which will resolve itself by the end of the year. Today there is panic buying all through the supply chain.

When COVID first hit my wife and I amassed a 30 day food an water supply. Not long after the grocery store shelves were empty because others did the same.

Automotive companies for example are going from Just in Time (JIT) to stocking chips. Distributors and brokers are stocking up as well hoping to profit from the shortage narrative with price increases for JIT deliveries.

"No matter how expensive, they'll buy what they can get," one auto industry analyst said.

View attachment 487

Only 4% of TSMCs chips are automotive. There is plenty of capacity for automotive once they fully embrace FinFETs. TSMC has already qualified their N5 process for automotive which is a big signal for the industry: Move to FinFETs if you want chips!


Bottom line: It's a correction. We have seen this before in the semiconductor industry and we will see it again, absolutely.
Looks at this chart and making the assumptions that:

a) we cannot quickly increase supply

b) there is a real shortage (rather than a temporary demand surge - and I'm rather with Dan on this one)

Then the only thing you can do is reduce demand.

So 48% of (chips ? wafers ? $sales ? doesn't seem to day) are "smartphone" and most of the smartphone market must now be the replacement and "upgrade" cycle.

Then we have only ourselves to blame for the ludicrously short lifespan of these devices. "Upgrade" every 3 years rather than 2 and that should drop demand enough ?

Of course, working in the industry, I have no desire to see demand drop. But what is good for us isn't always best for society as a whole.

One thing I don't understand here. Shouldn't a shortage cause price rises and those price rises cause a demand correction (drop) ? Or is the pricing mechanism too laggy - for example, if TSMC has long term pricing agreements and cannot raise prices to let the market do its work here ? As I believe they say in mining, "the cure for high prices is high prices".
 
TSMC is certainly catching up!

Taiwan Semiconductor reports revenue growth of 19.8% in May
TSMC's May revenue Y/Y rose 19.8% and 0.9% M/M to NT$112.360B.
YTD revenue advanced 17.1% to NT$586.085B.
 
Looks at this chart and making the assumptions that:

a) we cannot quickly increase supply

b) there is a real shortage (rather than a temporary demand surge - and I'm rather with Dan on this one)

Then the only thing you can do is reduce demand.

So 48% of (chips ? wafers ? $sales ? doesn't seem to day) are "smartphone" and most of the smartphone market must now be the replacement and "upgrade" cycle.

Then we have only ourselves to blame for the ludicrously short lifespan of these devices. "Upgrade" every 3 years rather than 2 and that should drop demand enough ?

Of course, working in the industry, I have no desire to see demand drop. But what is good for us isn't always best for society as a whole.

One thing I don't understand here. Shouldn't a shortage cause price rises and those price rises cause a demand correction (drop) ? Or is the pricing mechanism too laggy - for example, if TSMC has long term pricing agreements and cannot raise prices to let the market do its work here ? As I believe they say in mining, "the cure for high prices is high prices".
The economics of semiconductor manufacturing does not follow the standard demand supply logic of ECON101.
 
The economics of semiconductor manufacturing does not follow the standard demand supply logic of ECON101.
So what ? I said nothing at all about manufacturing and the supply side ! That has no bearing on anything I said.

My point is that a large proportion of chip demand is ultimately *discretionary*. We do not need to upgrade phones or computers or buy new cars (and more than we need to fly off on holidays - I've managed fine for almost 2 years without going anywhere near an airport and frankly I haven't missed that at all; we haven't been using our cars and will likely do so less in future - again I hope to waste less time travelling in future).

So the demand side is largely discretionary spend and therefore should be price sensitive. Do you disagree ?

My question was this : is TSMC passing on the increased prices fast enough for demand to adjust ? (Again, assuming there are real shortages).
 
I think much of the auto chip shortage is caused by TI Texas freeze power outages. TI is number one world wide analog chip vendor. I do not know about fabs. Chemical plants took weeks to restart caused by the way the freeze caused outages.
This is simply inaccurate. TI had ZERO shutdowns related to the freeze - only minor interruptions for a matter of hours, not even days. TI is #1 in analog, but maybe #4 or 5 in automotive, less than 20% of their revenues. Shortage is quite overly dramatic and self-inflicted, the supply-demand catch up will self-correct within 6-12 months.
 
Back
Top