Qualcomm/MediaTek Will Beat Apple To Shipping TSMC 3nm Based Smartphone Chips
Apple is well known for being TSMC's number one partner. They have been the first to move to every newest node in recent memory. Additionally, Apple helps with designing process development kit and even influences the standard cell design libraries. This may be starting to change as evidenced by the 3nm node and TSMC's comments surrounding it. It is looking increasingly likely that the first smartphone chips that are based on the N3 node will be coming out in early 2023, from Qualcomm or MediaTek, not Apple.
With the N10 node, TSMC went into high volume manufacturing in April of 2017. Apple subsequently released the iPhone, 7nm, and 5nm nodes, TSMC began volume production in April or May. This is the timeline needed to meet a fall release schedule for the iPhone. The complexity involved with manufacturing a leading-edge silicon wafer takes about 3 months. This is because there are hundreds of steps ranging from lithography, deposition, etch, atomic layer deposition, and more. Once you add in slicing, packaging it, and producing a consumer product around the chip, that means that there must be about 6 months from the start of high-volume manufacturing to products shipping to consumers.
Furthermore, Apple has typically followed a 3-year cadence with TSMC and the fabs they utilize. No one will ever publicly comment about Apple’s deals with TSMC, but it seems this is something that this could be related to their partnership. Leading edge silicon fabs cost tens of billions of dollars, and TSMC needs commitments on long term volume before they can make such a huge investment. Fab 14 was used for 20nm and 16nm products from the A8 through the A10. Fab 15 was used for 10nm and 7nm products from A11 through A13. It appears increasingly likely that Fab 18 will be used for 5nm and 4nm products from A14 through A16. This is further evidence that 2022 iPhones will be equipped with an A16 which uses a N5 or N4 node rather than N3 for 2022.
Risk production is scheduled in 2021, and production will start in second half of 2022. Our 3-nanometer technology will be the most advanced foundry technology in both PPA and transistor technology when it is introduced.
C. C. Wei – TSMC CEO
TSMC shared their timelines on their leading N3 process. N3 is the first time in recent history TSMC has had more than a 2 year gap between leading edge nodes. In response to this statement TSMC fielded many questions.
Andrew's first question, let me summarize, is asking about our 3-nanometer ramp. He notes that 5-nanometer and 7-nanometer in the past few years basically ramped in the middle of the year. N3, we said the ramp will be in second half of next year. So what is the reason behind this?
Question
Yes, about 3 to 4 months is a delay as compared with 5-nanometer. Yes, 3-nanometer technology actually is very complicated and in both processing technology and also the customers' product design. So we work with a customer, and finally, we decided to ramp up in the second half of next year.
C. C. Wei – TSMC CEO
C. C. Wei essentially is confirming Apple has chosen to stay on N5/N4 for 2022. Despite this, he also says that smartphone is still the first thing to ramp on this node.
Charles' first question is about the drivers of leading node adoption. He notes in the past traditionally it's mainly come from smartphones, but HPC also seems to becoming more important. So his question is on N3, do we expect HPC to play a bigger role in the ramp of N3, particularly in the first year? And could N3 become the first adopter or primary adopter?
Question
The N3's first year is ramping up, still smartphone plays the biggest role.
C. C. Wei – TSMC CEO
It seems very clear based on the CEO comments that a smartphone SOC or multiple will ramp first on 3nm just like with previous nodes, but this time it is not Apple. This only leaves 2 other companies. Qualcomm and MediaTek.
Qualcomm has been coy trying to use Samsung to optimize costs as much as possible, but this has bitten them in the butt as MediaTek has been gaining share. MediaTek has even surpassed them as the highest volume smartphone SOC vendor.
Samsung’s has a tremendous disadvantage in performance and power for their leading-edge nodes versus TSMC. Due to these issues, Qualcomm is rumored to be transitioning back to TSMC. Their next flagship smartphone chip which we will call Snapdragon 895, is rumored to be Samsung 4nm, but there is also allegedly + version of this chip which moves over to TSMC nodes.
People may be surprised to hear that MediaTek would be considered for aggressively progressing to TSMC 3nm, but they are also one of TSMC’s closes partners. Over the last year they have garnered the largest increase in 7nm wafer supply. Furthermore, they are looking like they may even beat AMD to becoming a $20B a year revenue semiconductor company. MediaTek ships over 1.5 billion Arm based SOCs a year and has a fledgling networking and custom ASIC business.
Since we made this call last year, the stock has more than doubled. They have been the biggest benefactor of the new semiconductor cold war. Chinese smartphone players are heavily prioritizing them and asking them to develop higher end SOCs. They were the first firm to utilize the N6 node. Furthermore, MediaTek’s next generation Dimensity 2000 SOC will be based on the TSMC N5 node family.
SemiAnalysis has found that MediaTek has been hiring at a rampant pace and even hired a number of engineers who work on TSMC’s process development kits (PDK). All signs point to MediaTek becoming more aggressive on node transitions and vying for the high end smartphone SOC market.
Whether it is Qualcomm who will try to stop the bleeding to MediaTek, or MediaTek trying to move into a new territory, it seems almost certain Apple will not be the first company to ship a 3nm based smartphone. This is a shock to the industry as Apple has been the closest TSMC partner for years. It is possible Apple ships a Mac based chip before these two firms ship smartphone chips, but that is an edge case. Furthermore, as AMD gets more aggressive with node transitions as well, it seems the long-term prioritization of TSMC favoring Apple is starting to break down. We are certainly in a renaissance for leading edge fabless silicon providers powered by the might of TSMC.
This article was originally published on SemiAnalysis on July 26th 2021.
Clients and employees of SemiAnalysis may hold positions in companies referenced in this article.