News TMSC Is Reportedly Terminating Discounts and Increasing Prices

[Admin]

As demand for chips grows, foundries increase quotes.

TMSC Is Reportedly Terminating Discounts and Increasing Prices : Read more

 

[Makaveli]

They have so much demand this was expected leaving money on the table with those discounts.

 

[jonathan1683]

hopefully this will drive bigger companies to use the own fabrication processes especially apple.

 

[watzupken]

hopefully this will drive bigger companies to use the own fabrication processes especially apple.

This is quite unlikely, especially for the likes of Apple. If you look at Intel and AMD, there are good reasons why other big chip makers don't want to run the fab business. It may seem cheaper to run your own, but there are complications and from as far as I can tell reading some of the articles out there, it is not cheap to own one as well.

 

[Shadowclash10]

This is quite unlikely, especially for the likes of Apple. If you look at Intel and AMD, there are good reasons why other big chip makers don't want to run the fab business. It may seem cheaper to run your own, but there are complications and from as far as I can tell reading some of the articles out there, it is not cheap to own one as well.

Yep. IIRC, back in the day (someone please correct me if I'm wrong), AMD tried their hand at owning their fabs and that was when they started spiralling downhill until recently, or something like that.

 

[OriginFree]

hopefully this will drive bigger companies to use the own fabrication processes especially apple.

Long term this may be cost effective, but from a cash flow point of view it usually isn't.
To me it's like a lease vs buy decision.

Lease: You can pay less and get the new model when it comes out (new fab mode)
Buy: Pay more overall and have the equipment still working 10-20 years down the road. It may not be producing CPU chips but you may be able to have a profit stream from other chips.

Does Apple really want to drop 10+ Billion on a fab to make their own CPUs for 5 years and then be in the fab business after that? Or is it easier to just pay a premium to use the best node out there? It's not like Apple know anything about charging people a premium to recoup costs.

 

[Jim90]

Seems eminently reasonable, considering the current status of TSMC. I'm sure profits all round will still be 'acceptable'.

 

[cryoburner]

Yep. IIRC, back in the day (someone please correct me if I'm wrong), AMD tried their hand at owning their fabs and that was when they started spiralling downhill until recently, or something like that.

That's not quite accurate. AMD was fabricating their own chips since their start in 1969. Only in 2009 did they spin off their fabrication business, which became GlobalFoundries, as they were finding it was becoming significantly more expensive to build fabs for smaller process nodes.

TSMC's discount prices fell within 3%, but the world's largest contract maker of semiconductors had decided to cease discounts starting next year, reports Taiwanese Central News Agency.

If the discount amounts to no more than 3%, then that in itself isn't going to affect much. The actual fabrication cost of each Ryzen chiplet, for example, is estimated to be quite cheap actually, only a small fraction of the total selling price of the processors. So ending a 3% discount isn't likely to raise the cost to manufacture them by more than a dollar. Even additional increases to fabrication costs probably won't mean much to the end-user, as these processors have rather high profit-margins. Large graphics chips would naturally be affected a bit more, but still not likely enough to affect pricing or profit margins much.

 

[escksu]

Only way is to build your own fab so you wont rely on tsmc. Right now they are the only ones with 5nm.

 

[InvalidError]

Yep. IIRC, back in the day (someone please correct me if I'm wrong), AMD tried their hand at owning their fabs and that was when they started spiralling downhill until recently, or something like that.

As cryo said, AMD was a chip foundry from the beginning.

What changed in 2009 is that its chip sales spiraled downward due to lack of competitive products and fabs are very expensive to own when you cannot run them at 100% capacity from lack of demand.

 

[Arrion]

As cryo said, AMD was a chip foundry from the beginning.

What changed in 2009 is that its chip sales spiraled downward due to lack of competitive products and fabs are very expensive to own when you cannot run them at 100% capacity from lack of demand.

AMD got its start in the CPU business as a foundry for Intel. Back in the 70s and 80s, chipmakers were required to provide a second source for their parts to ensure supply for a big contract. Intel licensed their early designs to AMD as that second source. AMD didn't get into the design side until Intel started making the Pentium.

That requirement for a second source shows how much things have changed.

 

[JamesSneed]

hopefully this will drive bigger companies to use the own fabrication processes especially apple.

No way that expenditure is 10+ billion. What you will likely see is Samsung pick up more business.

 

[InvalidError]

No way that expenditure is 10+ billion. What you will likely see is Samsung pick up more business.

The up-front build cost isn't an issue as long as you operate the fab long enough to make a profit relative to the cost of outsourcing fab. The main difficulty with owning your own fabs when you make chips that require high density is that you only get a few years useful life out of the equipment before you need to upgrade again. Fabs-for-hire on the other hand can pitch their more coarse nodes for things that don't require bleeding-edge process such as analog and high-power stuff. While everyone is talking about Intel's 7-14nm today, Intel still has fabs up to 45nm spitting out support components and long-term-support parts for things like military and aerospace applications.

 

[jkflipflop98]

Apple is about the only company with the resources needed to break into the fabrication industry. The barrier to entry is incredibly high.

Intel D1X was built on an existing site with everything you need to build a fab right there, by people who have built and started up fabs multiple times before. The best case scenario.

Just the building itself has cost over 10 Billion dollars. Then the tools needed to fill said building are about another 10 Billion dollars. That's just a building and machinery. Then you need (highly paid) engineers to actually figure out a process line consisting of over 40,000 individual steps to create the parts. In a brand-new factory with no point to start from, this is going to take 5 years at best. Then you need thousands of (well paid) technicians to actually operate the machinery after the engineers have figured out what to do with it all. After that, you're going to need a million wafers to burn through to do your learning.

Then you have to pay all those highly skilled people for at least 5 years with no return on investment at all before you start cranking out anything you could possibly sell.

It's difficult at best.

 

[JamesSneed]

The up-front build cost isn't an issue as long as you operate the fab long enough to make a profit relative to the cost of outsourcing fab. The main difficulty with owning your own fabs when you make chips that require high density is that you only get a few years useful life out of the equipment before you need to upgrade again. Fabs-for-hire on the other hand can pitch their more coarse nodes for things that don't require bleeding-edge process such as analog and high-power stuff. While everyone is talking about Intel's 7-14nm today, Intel still has fabs up to 45nm spitting out support components and long-term-support parts for things like military and aerospace applications.

While your point is valid the expenditure for a company that has no fab's is massive. Not only do they have to create all the infrastructure like buildings, load it with tech like ASML machines etc but then you also have to do all the R&D which is time, people, and money. I'm certain my $10B off the cuff was very low especially to start up a new cutting edge fab. Honestly I don't see any new fab entrants, at lest not cutting edge ones, as the price for entry is so absurdly high and the risk of that not paying off is just as high.

 

[jkflipflop98]

Funnily enough, Intel still makes 8088 processors. They produce a few lots of them in Israel every year.

 

[mikeebb]

I'm going with the ones who say losing a relatively small volume discount shouldn't affect the retail price of products. The CPU/GPU/SoC is a small part of the price of the finished product. That said, it'll certainly be used as an excuse to raise retail prices by some increment. Just because they can (people still buy at the scalpers' prices, right?).

A bigger long-term problem is that, to a first approximation, every fab in the world that's capable of running at or near the bleeding edge is in Taiwan and, perhaps a step or 2 behind, China. Pretty soon, if the saber-rattling continues, Taiwan will become physically a part of China not just politico-screaming. So if China so desires, its enemies will no longer have access to 5nm and the like. Think about that, as both a user and as a strategic issue. The fabs need to diversify production and corporate support locations.

 

[Acxe]

Now i see why nvidia 3000 series and amd cpus are always stock out.

CPUs and GPUs from nvda and amd are out of stock for months.
At first, i thought it was a coincidence.

Now i see the real picture




To all those who say Intel should go fabless, fabless is not as rosy as it seems.

With so much demand (especially after apple has entered the chat), it basically becomes a money game---the more money you pay, the more chips you get.

The "fairness" quotient disappears.



P.S.-- this money game is skewed heavily towards apple with its earnings

https://www.extremetech.com/computing/315186-apple-books-tsmcs-entire-5nm-production-capability