Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/advanced-node-equipment-shortage-cited.16251/
        )

    [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
)

Advanced node equipment shortage cited

Fred Chen

Moderator

(autotranslated excerpts)

The Wall Street Journal reported on the 9th that some TSMC (2330) customers have been warned that the company's production expansion in tomorrow (2023) and 2024 may not be as fast as expected, and the difficulty in obtaining manufacturing equipment is the main reason. Affected by the shortage of mature process wafers, the delivery schedule of chip manufacturing equipment has been delayed. In some cases, the lead time of new orders has even reached 2 to 3 years.

IBS CEO Handel Jones warned that 3nm and 2nm will face high demand and equipment shortages may be more serious. He estimates that in 2024-2025, the supply gap of 2-3nm chips may reach 10-20%.
 

(autotranslated excerpts)

The Wall Street Journal reported on the 9th that some TSMC (2330) customers have been warned that the company's production expansion in tomorrow (2023) and 2024 may not be as fast as expected, and the difficulty in obtaining manufacturing equipment is the main reason. Affected by the shortage of mature process wafers, the delivery schedule of chip manufacturing equipment has been delayed. In some cases, the lead time of new orders has even reached 2 to 3 years.

IBS CEO Handel Jones warned that 3nm and 2nm will face high demand and equipment shortages may be more serious. He estimates that in 2024-2025, the supply gap of 2-3nm chips may reach 10-20%.
2nm will not face high demand anytime soon. It is a yield learning node and will be mostly skipped. 3nm however will be the biggest node in modern times. I do know that future 3nm versions will have less EUV layers and maybe this is why. TSMC has more than half of the EUV systems in production so I don’t see a problem there for N3.
 
2nm will not face high demand anytime soon. It is a yield learning node and will be mostly skipped. 3nm however will be the biggest node in modern times. I do know that future 3nm versions will have less EUV layers and maybe this is why. TSMC has more than half of the EUV systems in production so I don’t see a problem there for N3.
But will they repurpose N5 EUV systems for N3 if demand for N5 is still high. I understand that yield for N5 is still bad but by the time N3 is into HVM, N5 yield would be much better.
 
But will they repurpose N5 EUV systems for N3 if demand for N5 is still high. I understand that yield for N5 is still bad but by the time N3 is into HVM, N5 yield would be much better.

Why would you say N5 yield is bad? I have heard the opposite from customers. N5/4 has more than 150 chips in process. Samsung 5nm yield was bad, I have not heard otherwise as of yet.
 
But will they repurpose N5 EUV systems for N3 if demand for N5 is still high. I understand that yield for N5 is still bad but by the time N3 is into HVM, N5 yield would be much better.

Further, it has been reported that Samsung has had such poor yields on their 5nm, 4nm, and 3nm, that Samsung management decided that falsifying their yields was necessary, as the truth would be too damaging.
 
The original WSJ article here: https://www.wsj.com/articles/supply...ting-edge-chips-11654777800?mod=djemalertNEWS

Chip Shortage Threatens Cutting-Edge Tech Needed for Next-Generation Smartphones

TSMC and Samsung are grappling with technological hurdles and a shortfall in manufacturing equipment

By Asa Fitch and Jiyoung Sohn


June 9, 2022 8:30 am ET

SEOUL—The two-year global semiconductor shortage is threatening to spread to some of the most advanced chips needed for next-generation smartphones and the data centers that power apps.

Chips with the tiniest transistors and highest performance had largely escaped the drought that has hit the auto industry and other electronics. Now, problems ranging from production hitches to a shortage of manufacturing equipment have raised concerns over the ability of the world’s two highest-end chip manufacturers to meet delivery promises to customers.

The challenges could ripple through the electronics supply chain as soon as next year, with one analyst warning of shortfalls as high as 20% for the most advanced chips by 2024 and beyond. Without improved chips, technologies such as high-performance computing, artificial intelligence and more evolved forms of autonomous driving might see a slowdown in deployment, industry analysts say.

Part of the problem is that just two companies—Taiwan Semiconductor Manufacturing Co. and Samsung SSNHZ 0.00%▲ Electronics Co.—are capable of building the industry’s most cutting-edge chips because of the high costs and technical barriers. Both have ambitious road maps in the coming months.

Some of TSMC’s TSM 0.57%▲ customers, however, received warnings that the company might not be able to increase production next year and in 2024 as quickly as hoped because of issues with acquiring manufacturing equipment, according to a person familiar with the situation. The company is making efforts to head off trouble, the person said.

Link: There Aren’t Enough Chips

Chip-manufacturing equipment is increasingly arriving later than expected, and lead times on new orders have stretched to in some cases two or three years, largely due to a dearth of less-advanced chips.

Then there are the technical issues. The contract-manufacturing unit of Samsung, the world’s second-largest contract chip maker, has experienced some capacity constraints. The Suwon, South Korea-based company saw slower-than-expected improvements in yields of chips made using the 4-nanometer process—the measurement used in the semiconductor world that loosely refers to the size of the transistors used in production.

Due to the low yields, Samsung was unable to supply as many chips as promised this year, prompting key customers including Qualcomm Inc. and Nvidia Corp. to place orders for their next-generation products with rival TSMC instead, people familiar with the matter said.

One nanometer, or one-billionth of a meter, is about a hundred-thousandth of the width of a strand of hair. The smaller the transistor, the newer and more advanced the chip, and the greater the number of chips that can be made on a single silicon wafer.

Both TSMC and Samsung say they are making progress on efforts to avoid any disruption.

Asked about TSMC’s production of its latest 3-nanometer chips in a call with analysts in April, Chief Executive C.C. Wei said that the company had issues with manufacturing-tool deliveries that it was working through.

“We’re working on 2023 right now, and we hope that we won’t have any big issue,” he said.

Samsung experienced delays in ramping up yields of its 4-nm processes, but the company is now “back on the expected yield improvement curve,” Kang Moon-soo, executive vice president of Samsung’s foundry business, said in a call with analysts last month. Samsung has said that it is on schedule to start mass production of the world’s first 3-nm chips using a novel transistor architecture by this month.

Any market concerns regarding the foundry business are excessive and unfounded, Mr. Kang said during the call.

Qualcomm has for years had a strategy of sourcing its chips from multiple manufacturers, and Samsung and TSMC are both important partners, a company spokeswoman said. Nvidia declined to comment.

Much of the equipment TSMC needs is also used for manufacturing older-style chips and has been in high demand, including from China. Some chip makers want equipment manufacturers to de-prioritize Chinese customers to more quickly meet their needs, according to people familiar with the matter, although equipment makers have pushed back against that effort.

TSMC sent executives to negotiate with equipment manufacturers to fend off any future impact to its growth plans, according to people familiar with the matter. Earlier this year, the company had discussions about obtaining more equipment from ASML Holding NV, one of the people said. ASML makes a range of crucial machinery for manufacturing the most advanced chips.

An ASML spokesman said demand for the company’s systems currently outstrips its ability to fulfill orders, which it is trying to address by helping customers get more output from existing tooling and other measures.

There is a mismatch in the amount of money advanced chip companies hope to spend expanding their production and the projected sales of the manufacturing-equipment industry. Chip equipment, which comprises most of the cost of setting up new chip factories, is expected to generate around $107 billion of sales globally this year, according to industry group SEMI. But planned capital expenditures by chip makers is projected to be more than that, at $180 billion, according to chip consulting firm International Business Strategies Inc.

The impact of high demand and equipment shortages on even more advanced 3-nanometer and 2-nanometer production will be significant, said Handel Jones, chief executive of IBS. He estimates a potential 10% to 20% supply shortage in that area in 2024 and 2025.

Chip-design companies that depend on contract chip makers have warned of technology and manufacturing-related risks that could affect their business down the road. Qualcomm said in its recent quarterly filing that developing or maintaining leading process technologies, including transitions to smaller geometries, could bring down manufacturing yields and reliability.

“Certain of our suppliers have in the past attempted, and may in the future attempt, to unilaterally reduce their capacity commitments to us,” Qualcomm said in its latest quarterly filing.

Jensen Huang, the chief executive of Nvidia Corp., the U.S.’s largest chip maker by market capitalization, said the supply chain would continue to be complicated and constrained, but his company had paid for longer-term commitments from its leading-edge suppliers.

“An agreement is an agreement, and we’re fairly confident with ours,” he said.

U.S.-based Intel Corp. aims to build a business making chips for others, but those plans are still at an early stage, and Intel can’t yet serve as an alternative to Samsung and TSMC.

Yang Jie contributed to this article.
 
Last edited:
I think there is a mismatch of semiconductor CAPEX over the next 5 years to what will actually be needed. Remember, TSMC builds fabs to order based on customer wafer agreements and pre pays from the big customers. The IDM foundries build capacity on what they expect to sell which is a much less predictable number. My prediction is that TSMC will spend the CAPEX they have forecasted but Intel and/or Samsung may not. This could lead to empty fab shells like we have seen in the past. And as I have said before, 2nm will be a yield learning node so no big equipment surge is needed yet. N7, N6, N5, N4 and N3 will live on for years to come.

I know the media wants controversy, fear, uncertainty, and doubt but it is just not there in this case.
 
I think there is a mismatch of semiconductor CAPEX over the next 5 years to what will actually be needed. Remember, TSMC builds fabs to order based on customer wafer agreements and pre pays from the big customers. The IDM foundries build capacity on what they expect to sell which is a much less predictable number. My prediction is that TSMC will spend the CAPEX they have forecasted but Intel and/or Samsung may not. This could lead to empty fab shells like we have seen in the past. And as I have said before, 2nm will be a yield learning node so no big equipment surge is needed yet. N7, N6, N5, N4 and N3 will live on for years to come.

I know the media wants controversy, fear, uncertainty, and doubt but it is just not there in this case.
The situation with ASML was already predicted here at SemiWiki by Scotten Jones, but it seems other equipment makers could be similarly affected.
 
And as I have said before, 2nm will be a yield learning node so no big equipment surge is needed yet. N7, N6, N5, N4 and N3 will live on for years to come.
That's why TSMC CEO C.C. Wei said N3 will be an important and long lasting node.
 
I think there is a mismatch of semiconductor CAPEX over the next 5 years to what will actually be needed. Remember, TSMC builds fabs to order based on customer wafer agreements and pre pays from the big customers. The IDM foundries build capacity on what they expect to sell which is a much less predictable number. My prediction is that TSMC will spend the CAPEX they have forecasted but Intel and/or Samsung may not. This could lead to empty fab shells like we have seen in the past. And as I have said before, 2nm will be a yield learning node so no big equipment surge is needed yet. N7, N6, N5, N4 and N3 will live on for years to come.

I know the media wants controversy, fear, uncertainty, and doubt but it is just not there in this case.

That seems spot-on, given the following CAPEX forecasts vs. wafer agreements and prepays:

TSMC - $120B over the next 3 years ($40B/yr.)
Samsung - $355B over the next 5 years ($71B/yr.)
Intel - $27B in ’22

https://www.androidheadlines.com/20...-dollar-semiconductor-expansion-new-fabs.html
 
OK, we have a dangerous situation ahead of us.

A. Because the excess inventory, capacity overbuilding, recession, and war, semiconductor industry will face a difficult market to sell their products.

B. Because the shortage of equipment and materials to quickly build new fabs or to add capacity to the existing fabs, semiconductor industry won't be able to meet the ever-growing market demand.

Can both be true?
 
OK, we have a dangerous situation ahead of us.

A. Because the excess inventory, capacity overbuilding, recession, and war, semiconductor industry will face a difficult market to sell their products.

B. Because the shortage of equipment and materials to quickly build new fabs or to add capacity to the existing fabs, semiconductor industry won't be able to meet the ever-growing market demand.

Can both be true?
I believe it could be both.
Difference nodes have their own supply and demand.
 
The situation with ASML was already predicted here at SemiWiki by Scotten Jones, but it seems other equipment makers could be similarly affected.
"IBS CEO Handel Jones warned that 3nm and 2nm will face high demand and equipment shortages may be more serious. He estimates that in 2024-2025, the supply gap of 2-3nm chips may reach 10-20%."

Handel Jones is just trying to stay relevant. 2nm will be scarce in 2024-2025 because it is a node most companies will skip. TSMC has a huge amount of N3 on tap then there is Samsung 3nm and Intel 3 coming soon.

Deposition and etch equipment is not a problem (LMRC, AMAT), or wafer inspection (KLAC). ASML is working on getting more lenses from Zeiss, paying more money, ASML is also improving throughput for existing systems per their presentation at SPIE. There will be more detail on this later from Scott Jones.

Bottom line: Over capacity will now come before lack of tools, absolutely.
 
I think there is a mismatch of semiconductor CAPEX over the next 5 years to what will actually be needed. Remember, TSMC builds fabs to order based on customer wafer agreements and pre pays from the big customers. The IDM foundries build capacity on what they expect to sell which is a much less predictable number. My prediction is that TSMC will spend the CAPEX they have forecasted but Intel and/or Samsung may not. This could lead to empty fab shells like we have seen in the past. And as I have said before, 2nm will be a yield learning node so no big equipment surge is needed yet. N7, N6, N5, N4 and N3 will live on for years to come.

I know the media wants controversy, fear, uncertainty, and doubt but it is just not there in this case.
I thought TSMC pre-payments were for adding capacity beyond TSMC's plan. TSMC still has a basic plan that is a forecast, not a deterministic indicator of demand. Is this a misconception on my part?

I also wonder... how much is supply chain diversification actually worth to big chip customers? If the answer is a lot, doesn't that make TSMC's plan significantly riskier as new non-TSMC capacity comes online?
 
Last edited:
How would advance nodes be over capacity in 2023 and 2024?

As I have mentioned before, Intel and Samsung are building fabs without customer commitments so there is a lot of overlap there with TSMC. Also, reshoring semiconductor manufacturing will add CAPEX and capacity without direct customer demand. That coupled with a drop in consumer spending in 2023 could create the perfect storm for a wafer glut.
 
  • Like
Reactions: VCT
Back
Top