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!
That's nearly $43 billion for FY 24, for 3nm & 5nm.
Well, there's no denying that TSMC has been leading in the industry in terms of the adoption of its processes and the revenue the Taiwan giant has been gaining in the past few quarters. With the AI hype kicking in and companies such as Apple preparing its next-gen lineup of products, TSMC has seen a massive increase in demand, especially for its 3nm and 5nm processes, which is why we previously reported on how the Taiwan giant is considering a large-scale price increase on 3nm/5nm, to capitalize on the market demand.
It is different this time. Everybody is using TSMC N3 since there was no trusted alternative. It looks like the same will happen with N2. 2025 and 2026 will be the years of N3 dominance, absolutely.
It is different this time. Everybody is using TSMC N3 since there was no trusted alternative. It looks like the same will happen with N2. 2025 and 2026 will be the years of N3 dominance, absolutely.
Should Samsung and Intel just give up and sell their tools to TSMC in exchange of getting preferred pricing? Both companies are spending all this money on their 2 nm class nodes and if they can’t get any customers, then they are flushing money down the toilet. ASML has a monopoly and there is still technological progress so what’s wrong with a monopoly on advanced logic technology?
Should Samsung and Intel just give up and sell their tools to TSMC in exchange of getting preferred pricing? Both companies are spending all this money on their 2 nm class nodes and if they can’t get any customers, then they are flushing money down the toilet. ASML has a monopoly and there is still technological progress so what’s wrong with a monopoly on advanced logic technology?
Technology moves faster when there is competition. Prices are lower when there is competition. Given the importance of semiconductors, I doubt Intel or Samsung will give up the effort. IDM foundries also have crossover technologies. When the memory business drops Samsung cuts investment. Intel is doing the same. The question I have is will Intel's business come back? This is a problem with IDM foundries. When the other side of the business tanks the foundry business is impacted.
I see the foundries coming after their suppliers to drop prices , when they themselves are maximising , not sure how thats working out so well for suppliers who are forced into race to bottom.
It seems everyone happy when the big boys are making their pile, but like all empires its built on a foundation of exploiting the vulnerable "in an economic sense".
I dont understand why its not possible to throw their suppliers a little break now and again.
Prices for who are lower?
I see the foundries coming after their suppliers to drop prices , when they themselves are maximising , not sure how thats working out so well for suppliers who are forced into race to bottom.
It seems everyone happy when the big boys are making their pile, but like all empires its built on a foundation of exploiting the vulnerable "in an economic sense".
I dont understand why its not possible to throw their suppliers a little break now and again.
Samsung sells on price. At 7nm I saw a 20% discount for one customer. Samsung 5nm is discounted as well. Intel will have to be aggressive on price since they are behind TSMC a bit. I doubt that TSMC will lower prices anytime soon. In fact, quite the opposite since there is no real competition for N3 and N2.
I see. It would be tragic if Intel succeeded in their ambitious node plans only to fail to win customers and sputter out of gas in the face of the ever increasing capex requirements of leading edge.
I see. It would be tragic if Intel succeeded in their ambitious node plans only to fail to win customers and sputter out of gas in the face of the ever increasing capex requirements of leading edge.
It would, but apart from having a competitive silicon technology to TSMC in PPA -- which they may well succeed at -- they have several other massive hills to climb, not just capex.
The first is cost competitiveness against a much bigger competitor pushing far more different products through the fabs, which is what you need to succeed in the foundry business because it pushes yield up. I believe Intel's yields -- like Samsung's -- have historically been lower than TSMCs, which is fine for an IDM selling high-margin Xeon products, but not good for a high-volume customer-driven foundry.
The second is the TSMC ecosystem -- not just things like tools and PDKs designed to support a massive customer base, but also the huge array of IP developed by all TSMCs customers and partners. The cost and effort to develop this IP is very high in nodes like N2/A16, and the likely ROI is much better with TSMC than Intel because there are far more potential customers for it.
The third may well be the arrogant Intel "we're the best!" attitude meaning they're less willing to do what customers want them to do rather than what they want to do to support their inhouse x86 CPU customer -- and that they make process choices targeted at this particular market (e.g. higher power/speed/voltage/cost) rather than the much bigger market of "everything else".
I agree it would be tragic for Intel (and the US semi industry, and perhaps the world...) if Intel fall and TSMC end up with a de facto monopoly, but it's difficult to see another realistic outcome... :-(
I see. It would be tragic if Intel succeeded in their ambitious node plans only to fail to win customers and sputter out of gas in the face of the ever increasing capex requirements of leading edge.
Who knows. It may turn out to be a blessing in disguise if Intel is forced to spin off its manufacturing division. AMD and Globalfoundries' split created a very successful AMD and a not-too-bad Globalfoundries. Most important thing is that both of them survived.
Intel does face a crisis but it can use it as a turning point for changing to a better and viable business model.
Who knows. It may turn out to be a blessing in disguise if Intel is forced to spin off its manufacturing division. AMD and Globalfoundries' split created a very successful AMD and a not-too-bad Globalfoundries. Most important thing is that both of them survived.
Intel does face a crisis but it can use it as a turning point for changing to a better and viable business model.
Global Foundries had to give up leading edge, losing AMD. The only reason why there even still are fabs in America chasing leading edge is because of Intel and their margins from both design and manufacturing chips. If there’s a split, then they too won’t be able to keep up and TSMC will have a complete monopoly.
Is it a reasonable estimate to assume that TSMC still makes 50%+ GM on their leading nodes. People have said they make tons of money on mature nodes but Leading nodes, not so much. I plugged that into my spreadsheets and it doesnt work. Apple is special case since they are development co-traveler, But I expect margins on N3,4,5 to be 50% plus? is that correct?
I see. It would be tragic if Intel succeeded in their ambitious node plans only to fail to win customers and sputter out of gas in the face of the ever increasing capex requirements of leading edge.
We should know by 2025 whether Intel has truly "only failed to win customers."
One advantage of being an IDM is the ability to showcase manufacturing prowess with their own products. For example, Samsung still hasn't used their 3nm GAA for Exynos processors, which raises questions about external customers' confidence.
If Intel’s 18A process achieves its PPAC goals on time and the design team delivers, we should see Clearwater Forest and Panther Lake launching in 2025, potentially outperforming AMD’s products made on TSMC’s N3/N4 nodes. If not, it’s no surprise that external customers would be hesitant to adopt 18A.
We should know by 2025 whether Intel has truly "only failed to win customers."
One advantage of being an IDM is the ability to showcase manufacturing prowess with their own products. For example, Samsung still hasn't used their 3nm GAA for Exynos processors, which raises questions about external customers' confidence.
If Intel’s 18A process achieves its PPAC goals on time and the design team delivers, we should see Clearwater Forest and Panther Lake launching in 2025, potentially outperforming AMD’s products made on TSMC’s N3/N4 nodes. If not, it’s no surprise that external customers would be hesitant to adopt 18A.
So the billion-dollar question is -- will Intel actually deliver both the 18A process and the products designed into it *in volume* in 2025?
Not the kind of "hey' we've got this fabulous new process (but it doesn't yield and we can only make a few chips)" launch they famously did in 10nm.
If they do then they'll have a lead on AMD. If they don't, they won't, because as sure as eggs is eggs TSMC will ramp up both N2 and A16 pretty damn quick...
And given Intel's recent process history, I don't think anyone other than Intel's product group will bet the farm on them delivering 18A as planned until they've actually shown they can do it, in volume, with good yield and competitive pricing.
Once Intel have done that -- assuming they do it! -- I'm sure they'll pull in more customers. But I can't see that happening until 2026 or later, maybe in the next process node -- which they'll also have to deliver on.
Their basic problem is that customers believe TSMC when they say "we'll deliver this process in volume with good yield in 202x" because they have a track record of doing exactly that over multiple process nodes in recent years -- and Intel don't, their recent record is one of dismally failing to do this. It's going be be a difficult reputation to shake off...
Global Foundries had to give up leading edge, losing AMD. The only reason why there even still are fabs in America chasing leading edge is because of Intel and their margins from both design and manufacturing chips. If there’s a split, then they too won’t be able to keep up and TSMC will have a complete monopoly.
wccftech.com
