Why it matters: The cost of progress is getting steeper with each new manufacturing process TSMC develops for Apple's A-series chips, which power the iPhone and iPad. A recent analysis sheds light on the increasing wafer prices and the diminishing transistor density gains Apple faces.
Let's rewind to 2013 and the A7, Apple's first 64-bit chip built on TSMC's 28nm process. At the time, those 28nm wafers cost Apple $5,000 each, according to Creative Strategies CEO Ben Bajarin's supply chain sources. Those wafers packed a billion transistors into the A7's dual-core CPU and quad-cluster GPU.
Fast forward to today, and the wafers for Apple's latest A18 Pro chips now cost $18,000 apiece – over 3.5 times the price of the A7's wafers. This translates to an eye-watering increase in cost per square millimeter, rising from $0.07 on 28nm to $0.25 on 3nm.
To be fair, the A18 Pro is fabricated on TSMC's cutting-edge 3nm process and crams an astonishing 20 billion transistors into its significantly more powerful CPU, GPU, and neural cores.
Got a detailed price/die/density analysis of Apple A-silicon over time at TSMC. Some nuggets.
– Ben Bajarin (@BenBajarin) December 18, 2024
From A7 to A18:
- progression from 28nm to 3nm
- Most dramatic shrinks occurred early (28nm → 20nm → 16nm/14nm)
- Steady increase in transistor count from 1 billion (A7) to 20…
Despite these impressive advancements, the rising costs are becoming harder to justify as performance gains diminish. Transistor density improvements have slowed in recent years as well.
Bajarin notes that the biggest density gains came during the transitions to 20nm and 16nm, and the "glory days" of 10nm and 7nm with the A11 and A12. Those two chips saw transistor density hikes of 86% and 69% respectively compared to previous generations.
However, in the past few years, these gains have slowed to a crawl. Between the A16 and A18 Pro, transistor density increases have dropped to single-digit percentages, largely due to diminishing returns in SRAM scaling. Despite this, Apple has to pay the piper much steeper prices for each new process node.
Also read: Aiming for Atoms – The Art of Making Chips Smaller
As for what's Apple's play here, it's all about maximizing the key performance-per-watt metric. As Bajarin explained to Tom's Hardware, IPC throughput gains are getting harder each generation, so the company has had to switch gears to focus more on power efficiency and keeping operating costs in check. Bajarin added that Apple has successfully maintained relatively stable die sizes while dramatically increasing transistor density.
It should also be noted that as a top customer, Apple has a few advantages over other TSMC clients. Rumor has it that Apple is the only client paying TSMC per "good die" rather than per full wafer, giving it a competitive edge over its rivals.
Apple's 3nm TSMC wafer costs soar to $18,000, more than tripling since the 28nm A7 chip
