News TSMC's 2nm will reportedly receive a price hike once again — $30,000 per wafer

[Admin]

TSMC's 2nm wafers will cost two times more than 4nm and 5nm wafers, according to report.

TSMC's 2nm will reportedly receive a price hike once again — $30,000 per wafer : Read more

 

[TCA_ChinChin]

Its expensive, but it kinda makes sense for how advanced and complicated these newest nodes are. Theres lots of neat new and newish fab + design techniques being put out to push transistor count/density nowadays.

 

[joeer77]

Already need double patterning for EUV? Nuts.

 

[thestryker]

TSMC has no competition and even if Intel's 18A ends up being better than N2 most companies won't just jump ship. TSMC knows this so it makes sense to charge the premium and then adjust down the line as needed. I hope we see a future where Intel, TSMC and Samsung all have competitive nodes which not only should drive costs down but expose whether or not TSMC was leveraging their market position (they should be).

 

[Amdlova]

It's why people call me nuts when I say Grab your GPU today.
TSMC will get nuts overpriced wafer and will have a small fee outside of taiwan...
I hope the samsung improve their fabs to get new contracts.
AMD and nvidia maybe go to samsung on consumer market because gaming don't give money!
Intel in other hand will Drain more power from the Gov/Army, will have more time to improve their 10nm +++++ production wait to se the new 14999KSU "Uranium edition" with the multiplier free to 92.

 

[jabojabo]

Has anyone seen estimates of wafer cost for Intel 18A? Just curious. Intel 18A will be competitive, but as others have mentioned jumping ship is painful and Intel will have to slowly build up its customer base. Luckily the hyperscalers are making chips and MSFT and Amazon have chosen 18A. Like N2, Intel 18A has gate all-around transistor, but 18A also has backside power, which is a huge advantage, frees up frontside real esate and likely leads to higher transistor densities. Also, looking even further out, Intel has aleady purchased, installed, and began doing R&D with an ASML high NA EUV machine ($380M each). They have bought up ASML's entire 2024 capacity of these machines. Meanwhile TSMC says they don't want to use high NA. If this remains the case, Intel will have a huge advantage in 2026 when they launch 14A which uses high NA.

 

[thestryker]

18A also has backside power, which is a huge advantage,

BSPDN is an optional feature which leads me to believe it will increase costs a fair amount.

Meanwhile TSMC says they don't want to use high NA. If this remains the case, Intel will have a huge advantage in 2026 when they launch 14A which uses high NA.

The advantage of High-NA isn't guaranteed in the short term as the latest EUV machines can process wafers close to as fast which was the guaranteed advantage. Long term it will likely be the way to go barring one of the alternatives being worked on actually being viable. It is of course also a possibility that TSMC is wrong about the economic viability of continuing to use EUV for more advanced nodes, but I don't think there's enough information to be sure yet.

Has anyone seen estimates of wafer cost for Intel 18A?

Doubtful anyone that isn't internal would have any clues as Intel has never opened leading foundry to third parties until now. I'd imagine that as actual wafer orders are placed some clues may leak out.