Another Great Quarter for TSMC: Apple's Silicon Partner and Roadmap Pacesetter

A few days ago, TSMC, arguably the world's most important and advanced pure-play chip company (certainly largest thanks to its ~$500B US market cap), announced unsurprisingly strong Q1 2022 earnings results. TSMC revenue grew 36% YOY and net income grew 45% YOY on a NTD (Taiwanese currency) basis. That translates to $17.6B in USD-basis sales and $7.3B net income, respectively.

TSMC management, via its earnings conference call (transcript and audio archive taken by SeekingAlpha, which didn't seem paywalled at time of writing, but might be now) were bullish on full-year 2022 as well as what it termed a multi-year "megatrend" in 5G smartphones and HPC (high-performance computing). Smartphone and HPC platform semiconductors, incidentally, each accounted for about 40% of TSMC's Q1 2022 revenue. (4:00 mark of archived conference call)

Volume Production for Chip Wafers vs. Volume Production for the Entire Chip-Enabled Widget

A bit later in the call, longtime TSMC CEO C. C. Wei had some remarks about the highly-anticipated, leading-edge, next-generation 3-nanometer lithography process which TSMC calls "N3". Specifically, he stated: "Our N3 schedule is unchanged, and we’re on track for volume production in [the] second half of 2022 with good yield." (15:20 mark of archived conference call)

Somewhere along the way amidst the pandemic, I got the impression that N3 was facing potential delays. Spoiler alert: It's not. 2H 2022 volume production has been planned since at least April 2020.

At first, I was excited by the possibility of this technology shipping "at the usual time" in late 2022. But my first impression was most likely mistaken. Apparently, "volume production" is just the start of a long supply chain process from the foundry (high-enough yielding chip wafers processed at scale, chips produced) to foundry customer product integration. To say nothing of the numerous planning steps and interdependencies I've left out.

From what little I can tell, it takes months for this unbelievably oversimplified sequence of events to happen:

(1) Wafers are painstakingly, um, lithographed and produced in sufficient quantity to yield the amount of chips required

(2) An OEM like Apple, Qualcomm, etc. takes those chips, volume-installs them on logic boards, and ramps up production of their own finished products (mostly iPhones for Apple, mostly Android-destined Snapdragon chipsets for rumored TSMC customer Qualcomm).

How many months does this all take, in Apple's case? From my casual-observer standpoint, it seems that a Fall iPhone release requires TSMC starting "volume production" of a given lithography platform by the second calendar quarter of that year. So in a "typical" scenario, volume production sometime between April through June will allow for a September-October iPhone launch later that same year.

This is where C. C. Wei's "second half of 2022" volume production timeframe greatly alters the reasonable projection for when N3 will be ready in volume for (preferred) TSMC customers. July through December doubles the usual three-month time window, and implies an additional lead time of between 3-6 months, when comparing April to July and June to December. While it's certainly possible that TSMC could ramp up chip wafers more quickly and with better yield curves than previous lithographic nodes with the appropriate (multi-multi-billion) investments in fabs, production equipment and high-skilled technicians, we also have to keep in mind:

>> the smaller lithography itself won't necessarily yield more chips per wafer, since Apple will inevitably look to boost transistor counts on their A- and M-series SoCs year after year

>> the smaller the lithographic process, the more incredibly complicated it is to "print" all of those ridiculously small transistors - likely adding process steps and therefore "cycle time" (which is longest at the outset, but tends to quicken over time with process maturity)

If we assume cycle time assumptions are built into Wei's volume production timeframe, we can then attempt our best guess of when these shiny new N3-process chips can be incorporated into what's almost certainly going to be iPhones (especially since TSMC is optimistic on yields, and sees N3 revenues from I-wonder-who's-buying-those-chips in Q1 2023).

A 3-6 month delay from the typical cadence might place "integrated product" availability between December 2022 and March 2023.

A December iPhone release wouldn't be too much later than the previous late-in-the-year recordholder, iPhone 12 Pro Max (launch date Nov. 13, 2020). But a March iPhone release would be quite a different story, considering that every iPhone 13 model launched on Sept. 24, 2021.

No matter. For at least some of the upcoming iPhones, I have high armchair analyst conviction that Apple will have "no choice" but to wait until the technology is ready.

The Intersection of Technology Cadence and Technology Leadership

Moving out the launch of any iPhone 14 model potentially up to six months beyond the "typical" September 2022 timeframe might seem outlandish. Apple has consumers to keep happy, and shareholders to answer to. There's also something called the iPhone Upgrade Program, which more or less promises a new iPhone every 12 months ("Get a new iPhone every year"). An 18-month cadence could cause a bunch of consumer outrage, and Apple's lack of disclaimers (one possible example: "we're not guaranteeing a new iPhone model will launch every 12 months") would hastily be fixed, after Apple unwittingly painted itself into a product cadence corner.

Apple management is definitely hoping that N3 wafers are ramping up rapidly closer to July 2022 than December 2022. But if there aren't enough chips for new iPhones until March 2023, it must wait, for one fundamental reason. Technology leadership - using and benefitting from the efficiencies and higher performance potential of 3nm Apple silicon (mostly in future iPhones) from the moment it's ready.

There is zero benefit to taking delivery of all those untold millions of next-generation chips and sitting on them until September 2023 or even June 2023. Apple won't be TSMC's only N3 customer, of course, and Qualcomm (to name just one competitor) would love nothing more than to upstage its chip-and-someday-5G-modem frenemy with a Snapdragon 900-series-or-something chipset on a non-Samsung 3nm process while Apple's still stuck on 5nm.

It's even conceivable that Intel could play the part of Apple-humiliating "we shipped 3nm first" spoiler - on TSMC's own N3 process, no less.

Given the competitive landscape, I think Apple chose wisely, and planned ahead for the hopefully-not-too-long wait for 3nm - preparing for an ultra-fast-paced 3nm-tech iPhone rollout the moment it has enough stock for a meaningful product launch.

Now comes the fun part - what does Apple do with the "iPhone 14" in the face of an extended, N3-availability-driven product cadence?

Gaming Out the "iPhone 14" Product Launch

Scenario 1: All-or-Nothing

The "all-or-nothing" approach has Apple not launching any of the "expected" iPhones (14 Regular, 14 Pro/Max, or possibly that rumored "14 non-Pro-but-still-Max") until 3nm is ready. This, of course, guarantees that the A16 Bionic would be 100% next-generation 3nm tech.

The upsides are all consumer-friendly (more performance, likely lower power consumption, and enabling new camera/feature possibilities), aside from the possibility of price hikes due to higher costs, inflation, etc.

The downsides are just as obvious - a perceived delay of as many as six months, roughly half a smartphone sales cycle in present-day terms. It probably wouldn't materially harm Apple's iPhone installed base in the long run, but it might sow some serious perceptional doubt amongst investors and analysts who've come to expect Apple always having a new iPhone ready just in time for the last calendar quarter.

Going all of 2022 without a new-SoC iPhone might, for the first time, herald a change in industry-wide smartphone release cycles from more-or-less-annually to every 18-24 months. Life will go on just fine without new smartphones every 12 months, but that's one heck of an industry shift that most in the industry (especially carriers) may not be fully prepared for just yet. It's far from clear if Apple itself is prepared at this point, even though the trend of increasingly hard-won performance gains points to slowing chip advances becoming an inevitability.

Scenario 2: Hedging Their Bets - Hybrid Launch of Third-Run 5nm A-Chips in Late 2022, Followed By True-Next-Gen 3nm in 2023

The next, perhaps(?) more likely approach would be Apple releasing some new iPhones in the expected Fall 2022 timeframe, and N3-based product for the top-end iPhones...whenever it's ready, whether it's year-end 2022 or March 2023.

The advantage of this approach is that Apple may not need to wait on TSMC to roll any other technological advancements into annual iPhone updates (for instance, a better camera system with more elements in the lens). And if Intel could continue to evolve its 14nm-platform chips for around five years, Apple can certainly find ways to achieve legitimate performance gains and/or efficiencies over three years on the 5nm platform. Quick example: For the Late 2022-bound mainstream-flagship iPhones, Apple could simply work from the top-binned iPhone 13 Pro 5-core GPU to immediately boost graphics performance. Maybe Apple brings back the "s" line of iPhones, and uses "iPhone 13s" with enhanced 5nm chips while saving "iPhone 14 (Pro/Max)" nomenclature for N3-based SoCs!

There are significant "downsides" to this cautious/"fallback" approach, though. Until now, Apple has always used every new flagship iPhone - Pro or non-Pro - to showcase its newest SoC. By decisively splitting the silicon innovation pipeline this development cycle, Apple would create a permanent stratification between Pro and non-Pro iPhones. After all, what's the operational point of having the mainstream iPhone flagship "fall behind" an iPhone Pro some years, only to "catch up" to the latest A-chip in others?

But wait, it gets sillier. What happens when the mainstream flagship iPhones get 3nm SoCs, perhaps by late 2023? Do you then call the "iPhone 13s" successor an "iPhone 14"-class device, confusing consumers even further? (I'd personally find no shortage of irony if Apple suddenly went in the direction of "Galaxy S" and "Galaxy A".)

True, anyone paying attention to the feature divide from iPhone 11/11 Pro and even farther back can make a very legitimate case that iPhone flagship tier differentiation is exactly what Apple's been working towards for years.

- Want three or more non-selfie cameras, and the best ones, plus LiDAR? Go Pro.

- How about the fastest GPU, the smoothest/brightest display, the most storage, or actual ProRes accelerators? It's Pro or nothin'.

- Prefer a rumored dash-dot TrueDepth sensor array instead of a "tired" not-even-that-old second-gen smaller Face ID notch? You know the starting price of admission.

Putting mainstream-flagship and premium-flagship iPhone lines on two separate technology tracks increases the predictability of new iPhone releases, but it also introduces one new chip line for Apple to design and ramp just as it's looking to leave 5nm behind (because there's no way Apple does "new phone, same old chip"). Additionally, with all signs pointing to N3 being a just-fine-yielding process, there's no reason for Apple to not produce as many 3nm-SoC iPhones as possible, since they're the highest-margin, highest-volume devices Apple makes into the foreseeable future.

You'd also expect lower-priced flagship iPhones to generally sell in higher quantities than, say, the newest Pro Max (which is still very popular, to be clear). So if Apple reserves 3nm SoCs exclusively for the Pro iPhone lineup, it's incredibly doubtful it'll use anything close to all of the chips TSMC can supply them.

Leaving any "new iPhone" one chip generation behind to (1) save on costs, (2) bring a new iPhone line to market "on schedule", or (3) add product differentiation just doesn't sound very Apple. It might even open a feature umbrella for competitors to out-spec them or further compete on price. Since even iPhone 13 and 13 mini are advertised as being faster than any of the Android competition, anything to undercut the performance narrative of mainstream flagship iPhones could militate against "third-run 5nm" or something similar for any new-generation iPhone.

Apple can't entirely control the flow or timing of new technology, yes. But until something drastic happens, it will always be first in line to acquire and volume-ship the latest and greatest chip tech that foundry partners and the laws of physics allow. I'd argue it's incredibly foolish to squander such a powerful advantage simply to avoid the perception of "an iPhone delay".

Anyway, Apple must have committed to the tech feature set of "whatever comes after the iPhone 13-series" long before I wrote this. Did they go with a contingency plan, or did they "roll the dice" on the 3nm availability timeframe? Around 5-6 months from now, we'll have a much better idea of which hard choices were made.

3nm Chips are Also Very Important for Mac

Try as I might, I'm unable to think up a good reason for Apple to delay putting Macs on the 3nm Apple Silicon roadmap as soon as reasonably possible. Unlike the tablet market, where Apple handily leads in any meaningful revenue/mindshare metric, PCs are as hotly competitive a market as ever (wattage joke aside). Adopting (read: scaling up) 3nm tech for Apple Silicon as soon as it's available also creates the widest possible performance-per-watt gap with the PC competition, and the best opportunity to gain share.

It's important to point out that M1 Pro and Max only became available on MacBook Pro starting late October 2021, and desktop users only gained access to M1 Max and Ultra a month ago in the case of Mac Studio, so "Pro-grade" Apple Silicon is still very, very new.

Unlike iPhone users, Mac users are used to waiting well over a year for updated Mac products, and you need look no further than the approaching-18-months-without-an-update first-wave M1 Mac mini, MacBook Air and Touch Bar MacBook Pro 13". Some Macs barely meaningfully update at all. Since patience is an expectation for Mac users, Apple would be leaving sky-high stacks of first-generation 3nm wafers on the table by not planning prosumer-and-up 3nm-SoC Macs to launch by at least mid/late 2023.

Sure, waiting until 3nm for M1-based Mac mini, MacBook Air, and iMac might be too long of a wait. But Apple doesn't necessarily have to turn to a third-run 5nm process to create a worthy M1 successor. For a consumer-level M2, the A15 SoC building blocks, which already feature a faster GPU and Neural Engine, would be a more-than-worthy "stopgap" to keep the "baseline" Apple Silicon momentum going.

It's a deliciously complex question of whether Apple feels compelled to fully scale a new M-chip generation from baseline all the way through workstation-level, near-ritualistically building up its chip learning and expertise while it continues down its long-term Mac SoC roadmap. If we follow the iPad silicon model, a relatively lower-volume product compared to iPhone, maybe not - iPad-specific silicon included the A5X, A6X, A8X, A9X, A10X and A12X/Z, skipping X-series silicon for the A7, A11 and A13 chip generations.

Timing is Everything, and It's Somehow A Bit Exciting

Life's been kind of a bummer for the world since the end of 2019, to say the least. Some parts of the globe are still under COVID lockdowns, despite being more than two years out from the start of the pandemic. Inflation is...just bonkers, and hardly confined to the US. The chip crunch is as intense as ever, and global supply chains might well be under greater strain than before. Ukrainians are fighting for their very existence, and the once-unthinkable invasion is having horrific downstream impacts even on the world food supply, which threatens multiple regions.

So maybe it's not such a bad thing - if we're fortunate enough to have the time and space for reflection - to look on the bright side every now and then. And believe it or not, 3nm chips are definitely something to look forward to. Unlike 5G modems relying on a Rube Goldberg-like series of things going just right to work as hyped, these chips only require a power source to work their magic, and 3nm tech (yes, including Samsung's eventual 3nm chips) will surely enable some very interesting technology use cases in the future.

For iPhone users and long-suffering Mac users alike (and iPad users, eventually), 3nm is all but certain to provide serious performance, efficiency and feature dividends to A-chip devices and maybe even Macs beginning 2023 and for some years afterward. With TSMC firmly on schedule and Apple having every incentive to deploy 3nm tech as quickly as possible, Apple's near-term iPhone and prosumer Mac release timeframes may be surprisingly clear - and quite promising. That might take some of the edge off a possibly underwhelming tech calendar and a gloomy-so-far 2022.