الخميس، 19 يناير 2023

NEWS TECHNOLOGIE

(Credit: TSMC)
The last we heard from TSMC, it was almost New Year’s Eve and the company had begun production on its cutting-edge 3nm process. It had promised to get to High Volume Manufacturing (HVM) of N3 in 2022, and it barely made the cutoff. That officially began on Dec. 29, so it was almost a photo finish. Now that the new year has arrived, TSMC has offered more information on its plans for 3nm in both 2023 and the years to come.

In its Q4 2022 earnings call transcribed by Seeking Alpha, the company went into more detail on what lies in store for its most advanced process. TSMC’s top brass covered a lot of ground wrapping up both the quarter and the year as a whole. However, when it comes to its burgeoning 3nm process, CEO C. C. Wei offered a lot of new details. For starters, he noted it had entered HVM for N3 as planned, and he said so far yields are good. He says the ramp for N3 will be “smooth.” In the latter half of the year though, he expects demand to overcome supply. That’s because, like other forecasts for the PC industry, he said it seems like the year will get off to a slow start, and pick up steam in the second half. When that happens, N3 will only contribute a single-digit percentage of the company’s wafer revenue for 2023. However, he says production will really ramp up in the second half of the year targeting HPC and smartphone applications.

TSMC’s roadmap to 2025. (Credit: TSMC)

Despite its infancy, he says he expects N3 to be more profitable than N5. Previously Mr. Wei had said he expects N3 to be worth more than $1.5 trillion once all the money has been counted in several years. So far the only customer we know of for this bleeding-edge process is Apple. That’s just because that’s always been the case in the past. Therefore, as Anandtech notes, Apple’s business will be the only way to measure the adoption of N3, at least in the short term. Apple is expected to utilize it for the upcoming A17 SoC in the iPhone 15 due in September. It was also thought to be using N3 for the M2 Pro and Max, but that didn’tcome to pass. It’s unclear why, but it could have been caused by delays from TSMC.

Going forward, N3 should be a blockbuster node for TSMC. It said it expects to earn more from it this year than it did when N5 was launched in 2020. Even at single-digit percentages, that’s still multiple billions of dollars. It’ll be starting with its baseline N3 product, then move quickly to N3E. That chip is not as dense but has a simpler design. While N3 uses 25 EUV layers, it only has 19. As such, it’ll be easier to produce, and thus less expensive for customers. It’s likely that only Apple will be onboard initially. However, when N3E comes online in the second half of the year, it’ll have customers lining up including AMD, Intel, and Qualcomm.

TSMC will be spending between $32 and 36 billion on capital expenditure this year to meet this demand. Over 70% of that wad of cash will be spent on its advanced nodes, which is 7nm and smaller. It’s not expected to move to 2nm Nanosheet Gate-All-Around (GAA) transistors until 2026 or so, meaning N3 will be its primary advanced node for the foreseeable future.

Now read:



from ExtremeTechExtremeTech https://ift.tt/d43XnN8

ليست هناك تعليقات:

إرسال تعليق