Apple is expected to buy chips from TSMC’s upcoming fab in Arizona to reduce its reliance on manufacturing in Asia, but the company won’t truly rely on Taiwanese chip manufacturing. So, our colleague bloomberg We believe the decision, which has yet to be formally announced, is a marketing gimmick rather than a need-based decision.
However, Apple has yet to confirm any plans to use TSMC’s fab in Arizona, which is set to begin operations in 2024, but what Apple needs in two years, and what TSMC will need in a few years. Let’s analyze what this powerful company has to offer. Year. One thing to keep in mind is that Apple’s supposed intention to use the Arizona fab: bloomberg report. Bloomberg is a highly reputable source, so we assume the information is accurate, but cannot verify it.
apple needs
TSMC’s fab in Arizona will manufacture chips using the company’s N5 (5nm class) process technology family, which includes N5, N5P, N4, N4P and N4X. Now Apple uses his N5, N5P, and N4 for various system-on-hi chips used in smartphones, PCs, and consumer electronics. Meanwhile, Apple is reportedly an alpha customer of TSMC’s cutting-edge process technology and the first company to use TSMC’s N3 (3nm class) node for at least one of its SoCs.
But it’s expensive to design chips for leading-edge process technology (we’re projecting about $600 million for physical implementation and embedded software alone) and manufacture them at leading-edge nodes. Apple uses his N4-based A16 Bionic SoC exclusively for the premium iPhone 14 Pro, which starts at $999. Meanwhile, the mainstream vanilla iPhone 14 (starting at $799) is still using his N5P-based A15 Bionic SoC for 2021 and beyond.
Perhaps a more colorful example is the launch of Nvidia’s 4N (a customized version of N4 with significantly enhanced performance) graphics processors based on the Ada Lovelace architecture. The AD102 and AD103 GPUs are dedicated to the company’s flagship GeForce RTX 4090 and RTX 4080 graphics boards and have price tags of $1,599 and $1,199 respectively (and these are, so to speak, the best available today). graphics card). In contrast, AMD’s over-the-range Radeon RX 7900 XT and RX 7900 XTX (based on Navi 31 GPU) will use TSMC’s N5 technology and will retail for $999 and $899 respectively starting December 13th.
But while the new process is expensive, Apple needs state-of-the-art production nodes not only for its flagship smartphones, but also for PCs. The company’s M1-based Mac has broken several performance records and clearly exceeded expectations. Meanwhile, Apple’s M1 Max, with 57 billion transistors (one of the most complex chips ever designed) implemented on TSMC’s N5P process, is, to put it mildly, the company’s best performance. It shows a high level of ambition for
Features of TSMC
You probably know some vectors of Apple’s SoC development, but you need to analyze them in relation to TSMC’s plans and capabilities. And they’ll probably give you some extra food for thought.
In October, TSMC cut its 2022 CapEx by $4 billion for the second time this year. This is due to the slowdown in medium-term future chip demand and the availability of wafer fab equipment (WFE). Reduced capital expenditures will not have an immediate impact on the company’s business, but reduced spending and longer lead times for fab tools could impact TSMC in the years to come. In fact, WFE availability seems to be a big issue for the semiconductor industry in general.
The move to cut TSMC’s capital expenditures from $40 billion to $36 billion came as a bit of a surprise as the company’s revenue and profits hit record highs in the third quarter of 2022. year-on-year increase, quarter-on-quarter he increased by 11.4%. The company’s Q3 2022 net income increased from $4.849 billion in Q3 2021 to $8.717 billion.
The main driver of TSMC’s recent profitability has been the aggressive growth of advanced manufacturing technologies N7 and N5 (5 nm and 7 nm classes, respectively) by general customers, especially TSMC’s largest client, Apple. It’s a good choice. In Q3, TSMC’s N5 production node family was the largest revenue contributor, accounting for 28% of the company’s revenue (up from 18% in Q3 2021). Profile products of prominent clients such as Apple (M2, A16), AMD (Zen 4 CCD), Nvidia (H100, AD102, AD103, AD104), Qualcomm (Snapdragon 8+ Gen 1). Meanwhile, the N7 manufacturing technology (and Apple still uses it to maleify its A16 Bionic SoC) accounts for 26% of his TSMC revenue (down from 34% in Q3 2021). ). After all, 54% of the #1 foundry revenue in the world came from advanced nodes (up from 52% in Q3 2021).
Nonetheless, TSMC has cut capital expenditures from $40 billion to $36 billion, citing medium-term prospects and the availability of WFEs, but citing US sanctions on China’s supercomputer and semiconductor industry. did not mention.
In fact, demand for PCs and smartphones is on the decline, and some say that it has not hit bottom yet. Meanwhile, high-performance computing products (that is, the vague term TSMC uses for his products for PCs, HPC, and other performance-hungry applications) now account for 39% of TSMC’s revenue. , showing a slow but steady growth. on the other hand, Sales of CPUs for client PCs fell.HPC’s contribution to TSMC’s revenue is also expected to decline, with the foundry’s N7 capacity utilization dropping to about 70% in the fourth quarter of 2022 and the first half of 2023, according to analysts. chinese renaissance.
Capacity expansion delays and impact on Mac
Analysts believe TSMC’s CapEx cuts will slow capacity expansion at TSMC’s N6/N7 production nodes. This makes him one of the company’s most significant revenue contributors today. Meanwhile, the foundry plans to redeploy some of these production lines to make advanced chips at the N4/N5 nodes as it has spare capacity in his N7-ready next quarter.
“TSMC will redeploy some of its N6/N7 idle capacity for N4/N5 and delay expansion of Fab 22 (built for N7/N28),” analyst Szeho Ng told the client. wrote in the memo.
The company said nothing about a slowdown in capacity expansion for the N5 family of manufacturing processes, but expects demand for N6/N7 to recover in the second half of 2023 as the utilization of its N7-ready product line increases. says that
Meanwhile, it is unclear how the CapEx reduction will affect the launch of the company’s N3 (3 nm class) family of manufacturing technologies. And this is where Apple is concerned. I don’t know what the N3 will produce. Apple may use his N3 for processors for PCs. These Macs may not be among the popular ones like the Mac Mini, MBA, or MBP 13 Pro, but there are expensive Mac Studio and MacBook 14/16 machines that can absorb the cost. But the problem is that no one knows what Apple plans to do. All we know is based on accurate comments from the China Renaissance Financial Analysts firm, but not official sources.
“We expect CapEx to be flat in 2023,” the analyst wrote. “The focus of CapEx in 2023 is that Qualcomm/Nvidia plans [Samsung Foundry]The pace of TSMC’s N3 buildout plans has lagged behind the pace of N5/N7 expansions during similar stages of launch given the initial focus of customers and cost concerns.
Overview
We don’t know exactly what Apple has planned for future chips, but it’s certainly possible that the Arizona fab won’t be able to fully meet its ambitious Apple Silicon plans. But would the company need a chip produced there? Absolutely! Is it a marketing gimmick? I can only wonder at this point.
