Of course all the big ones are all chasing the same next big thing and along comes a little guy who a vision, who will listen to him and pivot or do something different, counter and maybe totally against the current, the Giant ?
President Wei Zhejia (CC Wei) shows confidence: "There is no way to compete with TSMC"
- Thread starter Daniel Nenni
- Start date
Little guys mostly can't afford leading edge processes or packaging.
If it takes $60M cost for one product development in advanced nodes, and $billions to build a fab, that little guy must be rich and hard to be low profile.
I've heard that's in the range of the correct number, and it doesn't include the digital architecture, design, design verification by emulation, and implementation. Or the software or firmware. Or QA. Or system-level testing.
Taiwanese people and leadership are not naive that most of them understand the importance of diversification. However, there are many ways to achieve diversification and we can observe Japan and US taking very different approaches. Just like Aesop's Fables <<The North Wind and the Sun>>
For US, Raimondo, Secretary of Commerce keep emphasized Taiwan is a dangerous place and the need of chip diversification. Trump used even stronger words about stealing jobs. On the other hand, Japan shows warmer welcome to tsmc and the support can be seen from central government/local government/local people and even television.
In the end, we know the huge difference between an assignment and something we are really willing to do. US definitely has absolute bargain power to interact with Taiwan and tsmc, but it would be much better if all people feel it's a win-win opportunity.
As I've already said in another post, I think Japan is the role model for attracting fab construction right now. No question.
The problem the Biden administration has in the US is that his party, the Democrats, have constituents (voters) who are skeptical of all big companies, especially foreign companies, and they don't want to be perceived by their constituents as giving no-strings-attached funding to rich and powerful corporations. So there are inefficient rules about construction labor union membership, "supporting families", buying domestic materials, fiscal controls (restrictions on stock buybacks), profit sharing with the USG, and refusing to bypass turgid US federal, state, and local permitting processes and requirements. These are some among many other reasons why I think the CHIPS Act is so poorly conceived. These issues beg the question: Do we want companies to build in the US or not? I don't think we're serious.
US politicians think the security guarantee that keeps China from forcefully taking Taiwan is risky and expensive for the US, and literally priceless to Taiwan (and TSMC). From the perspective of the US government, the win-win is that Taiwan gets to remain an independent democracy, probably only because of the US defense umbrella. Frankly, I see their point, but I think the USG could do a lot of things quickly to make fabs far easier and cheaper to build in the US. The government just isn't motivated enough to risk the ire of their voters. It's pathetic, actually.
Last edited:
Building a fab is messy and complicated.
What and how TSMC does it is very unique everything from buildings, chemicals, to final hookup and the details are very different than any other company.
You would think such a veteran and experienced company that has built so many in Taiwan and with so much successful and such a dedicated a work force could get it right in the US. Everything they did started wrong and no wonder the schedule and stories we hear are coming out.
From the start who they picked, how they planned and trained was done naively and from a position of arrogrance. I am sure Morris has re emphasized his opinion and Mark clearly knows. A pitty to the crazy hard working stiffs on the bottom, but they are always the ones that pay when the generals and dictators got it wrong. Sure more money will offset the blood letting but never buy back the schedule or missed product alignment and the most painful thing the loss of face, morale, and confidence of the larger workforce.
I don’t think there will ever be the six phases that they planned after the nightmare experience of the first one. Especially now that the Germany one will be distraction number 2 sucking valuable resources and money too.
What and how TSMC does it is very unique everything from buildings, chemicals, to final hookup and the details are very different than any other company.
You would think such a veteran and experienced company that has built so many in Taiwan and with so much successful and such a dedicated a work force could get it right in the US. Everything they did started wrong and no wonder the schedule and stories we hear are coming out.
From the start who they picked, how they planned and trained was done naively and from a position of arrogrance. I am sure Morris has re emphasized his opinion and Mark clearly knows. A pitty to the crazy hard working stiffs on the bottom, but they are always the ones that pay when the generals and dictators got it wrong. Sure more money will offset the blood letting but never buy back the schedule or missed product alignment and the most painful thing the loss of face, morale, and confidence of the larger workforce.
I don’t think there will ever be the six phases that they planned after the nightmare experience of the first one. Especially now that the Germany one will be distraction number 2 sucking valuable resources and money too.
Last edited:
The only "non change" is it is changing with time and tsmc is not naive after decades experience in Wafertech (Now tsmc Washington), but still got burned in Arizona. tsmc is good at fab-to-fab benchmarking and practices it for decades. The jasm is a very good start/initiation of benchmarking cycle. Let's see the progress and response from US and Germany governments soon. Carrots has been there, the goal has been demonstrated, let see how the "invisible stick" works.
TSMC Washington is an illogical facility. It is in the small residential town of Camas, which when the facility was built was a town of a little over 20,000. The population is now 27,000. Housing is limited and surprisingly expensive. It's about a two hour commute each way in the mornings and late afternoon from Hillsboro, OR, where most of the people with fab experience in the region live. (As you probably know, Intel has over 20,000 employees in Hillsboro.) Construction expenses must have been significantly higher because of its relatively remote location, especially when it was built. If WaferTech was an example of TSMC's best thinking at the time, I'd certainly find their reasoning fascinating.
Naive, arrogant, or stupid ?
Given schedule slip, surprise not public that happened to the team all of the above is my assessment!
I would say surprised to see how inefficient US government still be, even for strategic projects.
Last edited:
I completely agree. And the CHIPS Act is far from the worst current example of our dysfunction.
TSMC to Win More Than $5 Billion in Grants for a US Chip Plant
REUTERS FILE PHOTO: Taiwan Semiconductor Manufacturing Company (TSMC) logo is seen near computer motherboard in this illustration taken January 8, 2024. REUTERS/Dado Ruvic/Illustration/File Photo (Reuters) - Taiwan Semiconductor Manufacturing Co (TSMC), the world's largest contract chipmaker...
semiwiki.com
Back to the original topic, I do not see how IDM foundries or any other foundries can compete with TSMC today. I have worked inside the foundry ecosystem for 30+ years and I just do not see it. GF is a great example. They splashed around billions of dollars inside the ecosystem for 15+ years which got them the #3 slot, no where near TSMC. Samsung has been at it for 20+ years and has spent hundreds of billions of dollars and is no where near TSMC. Intel can certainly try and they will no doubt be #2 but they are a very far cry from TSMC. It is just not going to happen. You cannot compete with a hyper focused company with 30+ years of ecosystem investment. Something catastrophic would have to happen to TSMC which is why GF used to play up hurricane and earthquake FUD and now Intel is using China FUD. It didn't work back then and it will not work now, my opinion. It will also alienate the Taiwanese and other people like myself who do not like dirty marketing tactics.
Mooredaddy
Well-known member
Agreed. I want to see competition in the market because it will make everyone better, TSMC included. But TSMC is not getting replaced or unseated as #1 in our lifetimes (my opinion). The economies of scale is just unfathomably large for them. Never mind the open ecosystem, institutional knowledge etc. Micron for its part actually has a tech lead on Samsung and will prove to be annoying to samsung for years to come I firmly believe. Samsung has also proven themselves to be untrustworthy with I.P. That leaves Intel and they just do not have the financial strength to go blow for blow with TSMC. TSMC has been expertly run for decades and C.C inspires all the confidence in the world (my opinion). Intel will get some wins kind of like IBM releasing press releases on their tech but they will not threaten TSMC in any way. It's difficult because I believe Pat is making the right moves in terms of trying to make money on their fab equipment (other then badmouthing TSMC and nvidia, amd, and many others) but its just a decade late and $200 billion short. SMIC is also not a threat just do to the fact people trust china even less then samsung to make their stuff.
Last edited:
About more than 10 years ago, foundry still worked on more variety of products and less volume in each product, and CPU/memory (intel/Samsung) world were less products and higher volume. We speculated that the foundry world would move to less products and higher volume years after. It was easy to expect the trend will favor intel/Samsung's working model, and tsmc/GF, especially tsmc might take times to adjust their manufacturing methodology. Upon the outcome 10 years later, I am surprised and would say tsmc adjust himself to the new normal and becomes even better than the potential winners now. It takes more than 10 years with clear strategies and executions. As we all know, the old days is you need to understand the technologies in RD thoroughly, ramp in scale and proliferate to high volume manufacturing site by "Copy Exactly". But now, it becomes " Doing before Understanding Thoroughly" be more efficient and find directions by proper "trial and error" work faster. This needs more talented engineers in RD and Fab site (not all scientists) by continuous improvement. tsmc works in the model for years and it seems better fit for the competition. Besides, foundry needs more team works even you have very good technologies. If you do not have mature field tech engineers, seasoned customer service manager to lead the tapeout, product engineers, mask house/litho engineers, reliability engineers..... as a team to plan proper tapeout and new product validation process, it could fail in the first silicon result. Then you lose the cycle (time to market) for at least 3-6 months. In my opinion, to catch up for building the seasoned team, it takes years and might cut short upon using artificial intelligent (like buzz word digital twin) but not that easy.
No one will ever beat TSMC to become the #1 foundry on equal terms of course...
But there are unequal ways to try to defeat TSMC:
1. Stay ahead of them on process technology for a while as a foundry - it's going to sap TSMC's strength over time when TSMC isn't "the cutting edge"
2. Play/Win the global subsidies game (unequal economics)
3. Develop and patent some technologies that make current chip manufacturing largely obsolete (maybe fantasy, maybe not)
There is no guarantee that TSMC will stay in the lead or 'strongest' forever. Once founders leave, companies can lose focus. TSMC could always face an IBM moment, decide to diversify in the wrong way, or simply move into stagnation.
(And it was really annoying when Samsung got hit by Earthquakes and we had NAND shortages in 2017..)
Mooredaddy
Well-known member
Morris hasn’t been involved for some time and TSMC is doing better then ever. Also the subsidy game is irrelevant. The CHIPS act is $39 billion for all chip manufactures spread over 10 years and is a once in a generation initiative. TSMC spends $30+ billion plus in capex each year by itself. It’s possible Intel gets ahead but I’ve been hearing very good things about TSMC 2nm so who knows. At the very least TSMC has the huge design wins on 2NM which will put it further in the lead financially at the very least. Dan would know better then most as to how 2NM PDK is being received. Also the North American tech symposium is next month so we will know alot more then. Exciting!!!
Last edited:
To be clear I'm in the camp there's only a 5% chance that TSMC is not #1 in foundry over the next 10 years, but I don't think it's guaranteed they remain #1 forever. There's plenty of history of companies finding ways to fail over time to disregard the possibility.
I have very high respect for DN and not just because I don't want my account banned here
. The amount of knowledge shared here freely is amazing..I agree on the excitement - honestly every decade gets more exciting than the one before when it comes to semi tech to me at least; the amount of innovation is staggering.
I think the CHIPs Act money is a bragging point or badge of honor so to speak. It will also help with the TSMC AZ fab costs. I have had quite a few conversations with the CHIPs Act people. Here is the latest one:
semiwiki.com
Podcast EP211:A Look at the Inner Workings of the CHIPS and Science Act with Mike O’Brien - Semiwiki
Dan is joined by Mike O'Brien. Mike was recently the vice president of aerospace and government at Synopsys, He has 40 years of experience in the semiconductor, software and computer industries. In his 27 years in EDA and IP at Synopsys and Cadence, Mike helped build new lines of business...
semiwiki.com
Ongoing development amidst production seems consistent with many process versions within same node family.