You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
I was at SPIE as well and I did see Chris. He was always the first up to the microphone for questions but rarely did I hear them answer him directly. And after the talks most speakers outright avoided him. Probably because he knows too much and has no filter.
Years ago these conferences provided a wealth of information. Lately however they seem more of marketing vehicles or forums for competitive disinformation. The same thing goes for the briefings we attend. You really have to ask some pointed questions to get to the heart of the matter. Rarely do I see traditional journalists ask these questions. Maybe because they don't want to offend and get shut out or maybe because they don't have the first hand experience to stand behind the questions. This really is a complicated business full of highly intelligent people.
I find that private discussions over drinks or lunch are much more informative but of course you cannot reference those discussions directly in print. I now get more than 100 emails per day, mostly to tell me when I'm wrong about something but sometimes when I am right or tips on things I should look into. I also make a point to let people know where I will be so we can meet in person because that is where the best information comes from.
Given all of that, based on what I know publicly and privately, I have ZERO confidence that EUV will make it into HVM for the foundries during this decade, if at all. Just my opinion of course.
Given all of that, based on what I know publicly and privately, I have ZERO confidence that EUV will make it into HVM for the foundries during this decade, if at all. Just my opinion of course.
Given all of that, based on what I know publicly and privately, I have ZERO confidence that EUV will make it into HVM for the foundries during this decade, if at all. Just my opinion of course.
In 2012, Samasug, Intel, and TSMC all invested in ASML through ASML's co-investment program. In 2015, TSMC sold all ASML shares they owned and made about US$660 million.
TSMC’s stake was part of a co-investment program ASML started in 2012. The objective of the program was to accelerate the development of ASML’s extreme ultraviolet lithography technology, which enables customers to produce smaller chips while increasing capacity and speed. Samsung Electronics Co. Ltd. and Intel Corp. are the two other companies participating in the co-investment program.
TSMC linked their investment in ASML to other financial vehicle to hedge their risk and they claimed it had come to the point they can (or they must) sell those shares. Looking back, I think lacking sufficient reason for TSMC to hold ASML's shares was probably the major reason for such stock sale.
Can we say TSMC has strong confidence on the prospect of ASML's EUV? Probably not, according to this stock sale.
I purposely left out Intel so they would not do it just to spite me! But since you asked.... Intel is the only one who could possibly use EUV for 10nm or 7nm but I highly doubt it since they are under CAPEX pressure. Besides, how can ASML do HVM themselves with only one customer that is in no hurry to introduce it? (No Apple pressure at Intel)
The most interesting piece of news is that 5nm may not require EUV either. In fact, if 5nm uses stacked nanowire technology current lithography technology will more than suffice. Just my opinion of course.
In 2012, Samasug, Intel, and TSMC all invested in ASML through ASML's co-investment program. In 2015, TSMC sold all ASML shares they owned and made about US$660 million.
The program started with Renee James making a $4.5 billion investment in ASML in the form of research payments and blocked shares. Intel was followed by TSMC investing about $1.5 billion to ensure, in my view at the time, that Intel did not gain exclusive control.
What then cracked me up was what ASML did with Intel's money. Within three months they bought Cymer for $3bn creating their own monopoly over EUV. At the time both Intel and TSMC (Morris Chang) were saying publicly that Moore's Law beyond 14nm depended on EUV. In Intel's case this was how they justified the $4.5bn spend.
I called to mind the 17th Century ditty that, "the fault of the Dutch, is offering too little and asking too much." In her own terms Renee James had given control of Moore's Law to a supplier.
In in the event, of course, technology moved on and the industry invented multi patterning. But if, and it's an improbable if, EUV is the only way to 5nm or below, ASML may have the levers to decide how that is done, which is my only reason for following them.
The story was one of the reasons I decided that the Emperor did not have any clothes.
That means so far Intel invested in ASML $4.5 billion while TSMC already walked away with all its original investment in ASML plus $660 million profit.
BTW, TSMC didn't invest $1.5 billion in ASML. $1.5 billion was the total amount when TSMC sold their ASML shares in 2015.
Samsung also made an investment in ASML. Does anybody know that status of that? I think it was similar to the TSMC investment. If Samsung still has their equity in ASML maybe they are still on the EUV train? Clearly TSMC is not.
Samsung also made an investment in ASML. Does anybody know that status of that? I think it was similar to the TSMC investment. If Samsung still has their equity in ASML maybe they are still on the EUV train? Clearly TSMC is not.
As ASML's President Martin vd Brink once labeled this program: shared pain, shared gain....
to me it sounds like a reasonable strategy for a company that is fully dependent on these top players to get the new technology introduced into manufacturing.
TSMC got out already of the finances (not the technology I think). I think they thought they did `their duty` and as they consider themselves not to be an investor, they got rid of the shares as soon as allowed and from the profit they made they are buying lots of ASML (immersion) tools it seems...
Shares Issued in Customer Co-Investment Program
In connection with the CCIP, on September 12, 2012, we issued 62,977,877 ordinary shares to the Stichting that holds shares on
behalf of Intel and 12,595,575 ordinary shares to the Stichting that holds shares on behalf of Samsung and on October 31, 2012,
ASML issued 20,992,625 ordinary shares to the Stichting that holds shares on behalf of TSMC. We received an amount of EUR
3,853.9 million in relation to the shares issued under the CCIP. The Stichting that held TSMC's shares in the CCIP has informed
ASML that TSMC has sold all of those shares. For further details on our CCIP see Note 34.
------------------------- Participating customers in our Customer Co-Investment Program together own a significant amount of our ordinary shares and their interests may not coincide with the interests of our other shareholders In the CCIP, the Participating Customers, being Intel, Samsung and TSMC, through certain wholly-owned subsidiaries, acquired in aggregate 96,566,077 ASML shares, which represented 23% of our outstanding shares at that time. In the CCIP, all of the Participating Customers agreed to a lock-up arrangement with us which expired in the first half of 2015. As the lock-up has now expired, the Participating Customers are permitted to sell their shares and the Stichting that held TSMC's shares in the CCIP has informed ASML that all of the ASML shares acquired by TSMC have been sold (20,992,625 ASML shares). Intel and Samsung are now presumed to own a total of 75,573,452 ASML shares based on the number of ASML shares initially acquired. Any sales by the Participating Customers are subject to the following limitations: any market sales are limited in any 6 month period to a total of 4% or 1.5% of our disclosed outstanding shares, in the case of Intel and Samsung, respectively, but such limitations do not apply to underwritten sales or block trades. The sale of a large number of these shares, or the perception that such sales may occur, could have an adverse effect on the trading price of our shares.
You are right of course. TSMC's investment was, from memory now, about $0.9bn, the difference in the reconciliation with your numbers being probably due to FX.
Samsung put in a smaller number. It might well have been about $0.5bn.
Intel announced $4.5 bn representing $3.3 bn invested in shares, which I found in its financials, so this is the figure to be compared with TSMC and Samsung investment. The balance of about $1.2 bn I could not find.
ASML's public announcement did not mention Cymer in connection with the coinvestment program. It is difficult to believe, however, it was not in the frame.
The SPIE papers are now available for download. I checked some key ones by GlobalFoundries (9776-1R) and SK Hynix (9776-1Q). EUV has problems for 7nm and even 10nm design rules. 10nm SADP is already possible by 193i single exposure.
Another paper, also from GlobalFoundries (977616), about 5 nm node patterning by EUV. Two options were considered: SAQP+EUV cut, and EUV single exposure with anamorphic 0.52 NA. Both had severe pattern shift problems (24 nm and 32 nm pitches). Suggest EUV change the acronym to DOA, DNR.
I have an article coming out here tomorrow on how logic may be able to do 5nm economically without EUV. Its just one possible scenario but some really smart people I know are pretty high on it.
I have an article coming out here tomorrow on how logic may be able to do 5nm economically without EUV. Its just one possible scenario but some really smart people I know are pretty high on it.
I think that given the semiconductor industry's uncanny ability to extend existing process that the idea of extending multi-patterning to 5nm may be the option with the least risk as it has the most "knowns". The industry hates risk as you don't want to risk a $5B -$10B fab on unproven process or "bet on the come" which is what EUV is. Although it seems somewhat clear that EUV would be the preferred option if all options were available ..... they are not. Given the horrible track record of delays and inability to get closure it would be safe to assume that the risk is similarly high for EUV going forward as ASML has been unable to show otherwise.
I think Intel is taking a rational approach saying they will use EUV when its ready. Given the lead time needed to nail down a process and tool set combined with EUV missing every milestone ever set will likely make it non starter until its proven out 100%.
It seems that the industry has contributed to the EUV delay issue by waiting for significant progress from ASML before working on the "ecosystem" issues of resist, pellicles, blank defects, mask inspection, etc; etc; etc;. Obviously all this should have been started years ago and worked on in parallel rather than serially as we have seen.
The other half of the equation component is economics....the long held belief is that EUV has to be cheaper given fewer process steps etc; but I wouldn't be so quick to "assume" thats the case currently given the progress on multi-patterning and lack of progress on EUV. Perhaps what was once a "slam dunk" economic decision is a bit closer today (I would be very interested to see Scotten Jones analysis...)
Like everything else in business we are talking about risk versus reward.....right now the variables are very high making any decision on EUV implementation today, almost worthless....there is less variability on multipatterning and if the economics improve then....
"A bird in the hand is worth two in the bush......"
Here is a link to a Wall Street Journal article from almost exactly 15 years ago talking about EUV and future prospects......its both funny and sad at the same time...
It seems that the industry has contributed to the EUV delay issue by waiting for significant progress from ASML before working on the "ecosystem" issues of resist, pellicles, blank defects, mask inspection, etc; etc; etc;. Obviously all this should have been started years ago and worked on in parallel rather than serially as we have seen.
Obviously you only seem to be aware of the marketing side of things. These things have been already going on for years, the EUV program was running full steam @ imec when I left litho 6 years ago; all the things you mention were looked at except that maybe at that time people thought it would still be possible to do EUV without pellicles.
Obviously you only seem to be aware of the marketing side of things. These things have been already going on for years, the EUV program was running full steam @ imec when I left litho 6 years ago; all the things you mention were looked at except that maybe at that time people thought it would still be possible to do EUV without pellicles.
Since ASML is the sole owner of EUV "all the things" mentioned will come about if, and only if, it commits itself and ponies up to make them happen. That is economics not marketing. No one else has a reason to take any kind of major risk. Were Intel or TSMC to take the plunge they would be handing half the benefit as a present to the other.
ASML have already pulled the "coinvestment" trick. Quite an object lesson in what is wrong with monopoly!
Brianhayes: ASML has the effective monopoly of immersion ArF as well (they sell about 70 a year, Nikon a handful if they are lucky) and the immersion monopoly is just accepted well in the industry. Nobody is complaining about ASML as far as such complaints would reach SemiWiki's forum. The barrier to entry in immersion-litho is simply infinity on the life time of a Semi-manager...
In my opinion it has nothing to do with monopoly. Simply tool availability with sufficient source power for the industry to work on the other EUV issues. The whole other 'EUV-infrastructure' was waiting for this to happen, and that kick-off started about 1.5 year ago (Summer 2014) when Robert Maire called the IBM EUV results with 40 Watt source power bogus: EUV Results Bogus, Says Analyst | EE Times
The industry `registered` that verdict of Mr. Maire and since then the industry stepped up the efforts on pellicles, resist, mask blanks etc........ Nothing to do with monopoly....
Now we are in `the stochastics land of LER`, much more interesting to see how that get's `solved`....
