9 ms·

Did chip espionage, IP theft give Samsung its 14nm manufacturing lead?

68 points by luu 12y ago

ChuckMcM 12y ago
Money quote: "According to Liang himself, he left TSMC after he was passed over for promotion and felt his work was under-appreciated by his former employer." Which apparently he was, as TSMC argues they are now "losing" millions in revenue because this guy helped Samsung get their 14nm node up. And yet at the time they were unwilling to give him a promotion and/or pay raise. What this argues is that TSMC could have doubled the guys salary and come out ahead by millions. Think about that employers when you start "nickel and diming"[1] the engineers that work for you :-) [1] American slang for charging (or being charged) in many small and ways which add up to a much larger total later on.
- SolarNet 12y ago
  That's a fair assessment, on the other hand, down this road is corporate dystopias and more powerful multinationals. When corporations have no recourse about their knowledge workers leaving and working for a competitor and giving up their IP/Trade Secrets, well companies are going to start to take more drastic measures. It's like the cold war, with scientists defecting back and forth, at some point these companies are going to start have spy/counterintel services if they can't rely on governments to deal with it.
  - bsder 12y ago
    > When corporations have no recourse about their knowledge workers leaving and working for a competitor and giving up their IP/Trade Secrets Companies do have recourse. It's called treating your employees well enough that they don't leave. If your company isn't smart enough to know who is important and how to keep them happy, then your company is going to have problems. Technical staff, especially, tend not to jump for small amounts of money or minor problems. As for technical workers, you simply can't prevent knowledge from leaving when they walk out the door. Especially for things like semiconductor manufacturing research, the knowledge of what was tried and didn't work is often more valuable than the knowledge as to what did work. How in the world would you stop that? Finally, he left in 2011. You're going to have a pretty difficult time convincing me that Samsung's current 14nm node is solely base upon TSMC manufacturing that was only at 28nm at the time. Overall, it sounds like TSMC had an internal political battle over what direction the semiconductor process research should go. Liang lost the battle and left. It seems that Liang was correct from an engineering standpoint, and the folks who stayed behind at TSMC were wrong. Sucks to be TSMC.
    - gress 12y ago
      You say 'treating their employees well enough that they don't leave' but that ignores the fact that a small number of employees will be able to blackmail the company for millions with the IP they have access to, while regular employees pay the price.
      - wombatpm 12y ago
        you are right, only CEO's should be able to command those kinds of dollars
      - 6d0debc071 12y ago
        Because big-wigs would totally pay their regular employees more if their expenses were lower rather than celebrating slightly more profit for themselves?
    - abawany 12y ago
      They unfortunately have an anti-pattern recourse as well: non-compete agreements. I was presented one today after a few months on the job and it is a doozy. Being in Texas, I am sure the courts will look much more kindly upon its shackles than California has so far. Such is the price of "living free" :).
      - hga 12y ago
        Side note which doesn't affect your point: without "consideration", such as more money, or when you start a job, the job itself, a non-compete is worthless, it's a contract where each side has to gain something. And "keeping your job" isn't one of those. As for California vs. Texas, well, you do get some serious considerations by making a move in either direction. Greater note: I've been convinced for some time that the non-enforcablility of non-competes is "Silicon Valley's" unique advantage. Given the steadily increasing difficulty of doing business in the state, it's almost certainly been critical in keeping that Golden Goose alive and somewhat healthy.
    - rgbrenner 12y ago
      Companies do have recourse. It's called treating your employees well enough that they don't leave. Except one person is part of a team that produced the IP. So they would always have access to IP worth more than their salary (since salary x NTeamMembers must be <= 100% of the IP value). So even if the team members were paid every dime of the value they've created, a single person would always be able to earn more by stealing IP/trade secrets.
  - x0x0 12y ago
    Yet somehow intel, nvidia, amd, and a thousand other companies manage to do just fine in CA, where non-competes are largely unenforceable.
    - ChuckMcM 12y ago
      Funny how that works.
    - gaius 12y ago
      They do it illegally, the news has been full of it. Eric Schmidt and Steve Jobs were the masterminds who set it up.
    - rm445 12y ago
      Carrying expert knowledge to another company is one thing, but disclosure of trade secrets? Not a lawyer, but I gathered that when trade secrets are properly handled, you have legal recourse against an employee (or ex-employee) who discloses them.
      - hga 12y ago
        Indeed. The only legal basis for non-competes being valid is that they prevent the "first bite of the apple", the damage caused by the first such disclosure of IP, which then requires the harmed company to sue to obtain recourse.
  - gcb0 12y ago
    wrong, they should patent it. let the hobbyist learn how they do, and prevent competitors from copying. they tried to have the cake (don't patent, don't publish) and eat it too (not pay people and processes enough to protect your trade secrets)
    - chrisbennet 12y ago
      Patents don't work if they can't be enforced. How would you enforce a manufacturing process if you weren't allowed into your competitors factories? "We have essentially no patents in SpaceX. Our primary long-term competition is in China," said Musk in the interview. "If we published patents, it would be farcical, because the Chinese would just use them as a recipe book."
- gcb0 12y ago
  but even before that, the issue was the company either tried to be smarter and game the system by not patenting their methods and calling them trade secrets... or maybe it's a new necessity because of China? i don't know. what matter is that either they had patents and with a former employer or not, they can crush Samsung. or if they tried to be opportunistic and threat it as trade secrets to game patents expirations or need to publish methods... than they should just shut up and lament not paying that guy more.
yellowapple 12y ago
I'm just going to assume that Betteridge's Law of Headlines is in effect here. If you're going to make some strong claim like "Samsung stole IP", phrase it as a damn declarative statement. Phrasing it as a question instead only suggests to me that the claim resulted from rectal extraction and is entirely unfounded.
ghshephard 12y ago
Interesting that there is an entire article on potential (emphasis on potential) theft of intellectual property, but not a single mention of patents. Don't the various semiconductor companies have massive stocks of patents, for just this sort of situation?
- sliverstorm 12y ago
  FWIW neither is a US company; TSMC is Taiwanese and Samsung is Korean. I have no idea how that plays out in patent-land, but patents are per-country. E.g., "U.S. Patent #123"
- rayiner 12y ago
  Trade secrets are also IP, and favored in the semiconductor industry 1) because you don't have to make the designs public; 2) because process generations are short and patents take too long to get.
- WildUtah 12y ago
  It takes about three years to get a patent (in both the US and Europe, the two largest semiconductor markets) and a valid patent (one that is genuinely original and not obvious) is obsolete in cutting edge semiconductors within three years, so patents play almost no role in semiconductors except to delay new entrants, obstruct compatibility with installed base software, and troll fabs with invalid scam patents. Trade secrets are essential or else nobody would innovate -- they'd just steal each others' designs until nobody could afford to invent anything. Fortunately a short non-compete (during which engineers should be paid compensation for taking time off between employers) is enough to exhaust a lot of cutting edge insider knowledge. The accusation here is that an engineer cheated his non-compete.
- rhino369 12y ago
  In something like semiconductors, the main competitors have tens of thousands of patents and each infringe each other 500 times over. It is essentially a wash. Competitors suits are only a good idea when you have a clear patent advantage. That is why Apple and Google got blitzed during the smartphone wars. They didn't have a phone patent portfolio like everyone else, since they were new players. It's the reason Google bought Motorola. Also, semiconductor processing is hard to reverse engineer so there are legit trade secrets that are kept.
bhc 12y ago
The fact that the article used by ExtremeTech as the source (http://english.cw.com.tw/article.do?action=show&id=14895 http://english.cw.com.tw/article.do?action=show&id=14895) refers to Liang, the former TSMC employee, as a "traitor" over and again casts doubts in my mind about the accusations presented.
bsder 12y ago
This is nothing new. If you piss off the people important to your company, you're gonna have a bad time. cf. The original Traitorous Eight from Shockley Semiconductor in 1957 https://en.wikipedia.org/wiki/Traitorous_Eight https://en.wikipedia.org/wiki/Traitorous_Eight
- raintrees 12y ago
  Or it can be done legally above board to get around trade secrets for competition, as in the Phoenix BIOS. https://en.wikipedia.org/wiki/Phoenix_Technologies https://en.wikipedia.org/wiki/Phoenix_Technologies
  - hga 12y ago
    Not trade secrets, by definition, because all was exposed. In fact, if I remember correctly, IBM even published on paper the assembly for their (first) BIOS. This is a copyright vs. patent sort of thing. What Phoenix copied using the "Chinese Wall" technique is the sort of thing that might be covered by patents, that is, ideas. The IBM BIOS was a creative implementation/expression of those ideas, and that code was copyright protected, as was Phoenix's independent version of the code that implemented the same ideas.
    - raintrees 12y ago
      You are correct about the paper copy, I still have the IBM Technical Reference Manual binder that has the full code listings.
ripberge 12y ago
Doesn't surprise me. My old boss (a professor at Cal Tech) has a story of coming into his labs one afternoon when an assistant greeted him with "Oh, the Koreans came by to take the photos of the hardware." To which he replied, "What Koreans?"
- hkmurakami 12y ago
  Out of curiosity, what era was this in? Recent or decades ago? (genuinely curious if this style of blatant espionage still goes on today -- I have some stories from the 80's)
  - ripberge 12y ago
    He told me this story around 2005. I would guess it may have been about 5-10 years prior to that. Could have been more as he was there for a long time.

68 points by luu 12y ago

Money quote: "According to Liang himself, he left TSMC after he was passed over for promotion and felt his work was under-appreciated by his former employer." Which apparently he was, as TSMC argues they are now "losing" millions in revenue because this guy helped Samsung get their 14nm node up. And yet at the time they were unwilling to give him a promotion and/or pay raise. What this argues is that TSMC could have doubled the guys salary and come out ahead by millions. Think about that employers when you start "nickel and diming"[1] the engineers that work for you :-) [1] American slang for charging (or being charged) in many small and ways which add up to a much larger total later on.

SolarNet 12y ago
That's a fair assessment, on the other hand, down this road is corporate dystopias and more powerful multinationals. When corporations have no recourse about their knowledge workers leaving and working for a competitor and giving up their IP/Trade Secrets, well companies are going to start to take more drastic measures. It's like the cold war, with scientists defecting back and forth, at some point these companies are going to start have spy/counterintel services if they can't rely on governments to deal with it.
- bsder 12y ago
  > When corporations have no recourse about their knowledge workers leaving and working for a competitor and giving up their IP/Trade Secrets Companies do have recourse. It's called treating your employees well enough that they don't leave. If your company isn't smart enough to know who is important and how to keep them happy, then your company is going to have problems. Technical staff, especially, tend not to jump for small amounts of money or minor problems. As for technical workers, you simply can't prevent knowledge from leaving when they walk out the door. Especially for things like semiconductor manufacturing research, the knowledge of what was tried and didn't work is often more valuable than the knowledge as to what did work. How in the world would you stop that? Finally, he left in 2011. You're going to have a pretty difficult time convincing me that Samsung's current 14nm node is solely base upon TSMC manufacturing that was only at 28nm at the time. Overall, it sounds like TSMC had an internal political battle over what direction the semiconductor process research should go. Liang lost the battle and left. It seems that Liang was correct from an engineering standpoint, and the folks who stayed behind at TSMC were wrong. Sucks to be TSMC.
  - gress 12y ago
    You say 'treating their employees well enough that they don't leave' but that ignores the fact that a small number of employees will be able to blackmail the company for millions with the IP they have access to, while regular employees pay the price.
    - wombatpm 12y ago
      you are right, only CEO's should be able to command those kinds of dollars
    - 6d0debc071 12y ago
      Because big-wigs would totally pay their regular employees more if their expenses were lower rather than celebrating slightly more profit for themselves?
  - abawany 12y ago
    They unfortunately have an anti-pattern recourse as well: non-compete agreements. I was presented one today after a few months on the job and it is a doozy. Being in Texas, I am sure the courts will look much more kindly upon its shackles than California has so far. Such is the price of "living free" :).
    - hga 12y ago
      Side note which doesn't affect your point: without "consideration", such as more money, or when you start a job, the job itself, a non-compete is worthless, it's a contract where each side has to gain something. And "keeping your job" isn't one of those. As for California vs. Texas, well, you do get some serious considerations by making a move in either direction. Greater note: I've been convinced for some time that the non-enforcablility of non-competes is "Silicon Valley's" unique advantage. Given the steadily increasing difficulty of doing business in the state, it's almost certainly been critical in keeping that Golden Goose alive and somewhat healthy.
  - rgbrenner 12y ago
    Companies do have recourse. It's called treating your employees well enough that they don't leave. Except one person is part of a team that produced the IP. So they would always have access to IP worth more than their salary (since salary x NTeamMembers must be <= 100% of the IP value). So even if the team members were paid every dime of the value they've created, a single person would always be able to earn more by stealing IP/trade secrets.
- x0x0 12y ago
  Yet somehow intel, nvidia, amd, and a thousand other companies manage to do just fine in CA, where non-competes are largely unenforceable.
  - ChuckMcM 12y ago
    Funny how that works.
  - gaius 12y ago
    They do it illegally, the news has been full of it. Eric Schmidt and Steve Jobs were the masterminds who set it up.
  - rm445 12y ago
    Carrying expert knowledge to another company is one thing, but disclosure of trade secrets? Not a lawyer, but I gathered that when trade secrets are properly handled, you have legal recourse against an employee (or ex-employee) who discloses them.
    - hga 12y ago
      Indeed. The only legal basis for non-competes being valid is that they prevent the "first bite of the apple", the damage caused by the first such disclosure of IP, which then requires the harmed company to sue to obtain recourse.
- gcb0 12y ago
  wrong, they should patent it. let the hobbyist learn how they do, and prevent competitors from copying. they tried to have the cake (don't patent, don't publish) and eat it too (not pay people and processes enough to protect your trade secrets)
  - chrisbennet 12y ago
    Patents don't work if they can't be enforced. How would you enforce a manufacturing process if you weren't allowed into your competitors factories? "We have essentially no patents in SpaceX. Our primary long-term competition is in China," said Musk in the interview. "If we published patents, it would be farcical, because the Chinese would just use them as a recipe book."
gcb0 12y ago
but even before that, the issue was the company either tried to be smarter and game the system by not patenting their methods and calling them trade secrets... or maybe it's a new necessity because of China? i don't know. what matter is that either they had patents and with a former employer or not, they can crush Samsung. or if they tried to be opportunistic and threat it as trade secrets to game patents expirations or need to publish methods... than they should just shut up and lament not paying that guy more.

yellowapple 12y ago

I'm just going to assume that Betteridge's Law of Headlines is in effect here. If you're going to make some strong claim like "Samsung stole IP", phrase it as a damn declarative statement. Phrasing it as a question instead only suggests to me that the claim resulted from rectal extraction and is entirely unfounded.

ghshephard 12y ago

Interesting that there is an entire article on potential (emphasis on potential) theft of intellectual property, but not a single mention of patents. Don't the various semiconductor companies have massive stocks of patents, for just this sort of situation?

sliverstorm 12y ago
FWIW neither is a US company; TSMC is Taiwanese and Samsung is Korean. I have no idea how that plays out in patent-land, but patents are per-country. E.g., "U.S. Patent #123"
rayiner 12y ago
Trade secrets are also IP, and favored in the semiconductor industry 1) because you don't have to make the designs public; 2) because process generations are short and patents take too long to get.
WildUtah 12y ago
It takes about three years to get a patent (in both the US and Europe, the two largest semiconductor markets) and a valid patent (one that is genuinely original and not obvious) is obsolete in cutting edge semiconductors within three years, so patents play almost no role in semiconductors except to delay new entrants, obstruct compatibility with installed base software, and troll fabs with invalid scam patents. Trade secrets are essential or else nobody would innovate -- they'd just steal each others' designs until nobody could afford to invent anything. Fortunately a short non-compete (during which engineers should be paid compensation for taking time off between employers) is enough to exhaust a lot of cutting edge insider knowledge. The accusation here is that an engineer cheated his non-compete.
rhino369 12y ago
In something like semiconductors, the main competitors have tens of thousands of patents and each infringe each other 500 times over. It is essentially a wash. Competitors suits are only a good idea when you have a clear patent advantage. That is why Apple and Google got blitzed during the smartphone wars. They didn't have a phone patent portfolio like everyone else, since they were new players. It's the reason Google bought Motorola. Also, semiconductor processing is hard to reverse engineer so there are legit trade secrets that are kept.

bhc 12y ago

The fact that the article used by ExtremeTech as the source (http://english.cw.com.tw/article.do?action=show&id=14895 http://english.cw.com.tw/article.do?action=show&id=14895) refers to Liang, the former TSMC employee, as a "traitor" over and again casts doubts in my mind about the accusations presented.

bsder 12y ago

This is nothing new. If you piss off the people important to your company, you're gonna have a bad time. cf. The original Traitorous Eight from Shockley Semiconductor in 1957 https://en.wikipedia.org/wiki/Traitorous_Eight https://en.wikipedia.org/wiki/Traitorous_Eight

raintrees 12y ago
Or it can be done legally above board to get around trade secrets for competition, as in the Phoenix BIOS. https://en.wikipedia.org/wiki/Phoenix_Technologies https://en.wikipedia.org/wiki/Phoenix_Technologies
- hga 12y ago
  Not trade secrets, by definition, because all was exposed. In fact, if I remember correctly, IBM even published on paper the assembly for their (first) BIOS. This is a copyright vs. patent sort of thing. What Phoenix copied using the "Chinese Wall" technique is the sort of thing that might be covered by patents, that is, ideas. The IBM BIOS was a creative implementation/expression of those ideas, and that code was copyright protected, as was Phoenix's independent version of the code that implemented the same ideas.
  - raintrees 12y ago
    You are correct about the paper copy, I still have the IBM Technical Reference Manual binder that has the full code listings.

ripberge 12y ago

Doesn't surprise me. My old boss (a professor at Cal Tech) has a story of coming into his labs one afternoon when an assistant greeted him with "Oh, the Koreans came by to take the photos of the hardware." To which he replied, "What Koreans?"

hkmurakami 12y ago
Out of curiosity, what era was this in? Recent or decades ago? (genuinely curious if this style of blatant espionage still goes on today -- I have some stories from the 80's)
- ripberge 12y ago
  He told me this story around 2005. I would guess it may have been about 5-10 years prior to that. Could have been more as he was there for a long time.