Apple Earnings, An Interview with Jay Goldberg About Chips and Intel

Author: Ben Thompson (Stratechery), interviewing Jay Goldberg URL: https://stratechery.com/2021/apple-earnings-an-interview-with-jay-goldberg-about-chips-and-intel/ One-line: A chips-industry conversation on Apple’s 5G iPhone strength in China, the economics of the fabless model, Apple’s push to re-integrate vertically down to the chip, and the cultural problems facing Intel.

Key claims

• 5G drove a strong iPhone cycle in China, where the 5G network is well established and the overwhelming majority of phones sold are already 5G — Apple benefited from pent-up anticipation for a 5G iPhone.
• 5G matters more (1) where it is widely available and (2) where most computing — including heavy streaming video — happens on the phone. If Chinese consumer behavior runs ~3–5 years ahead of the US, the explosion in mobile video is just getting started.
• The fabless model took off because a leading-edge fab now costs ~$15B; abstracting design from manufacturing let a “big flowering” of specialized design and chip companies emerge.
• Qualcomm has a “lowest common denominator” problem — it must segment one chip across many customers (high-end/low-end, every feature) — a liability versus Apple, which optimizes its A-series for exactly its own needs. (Vertical Integration)
• This fits the Clayton M. Christensen model: new entrants integrate top-to-bottom because it’s hard and integration makes it work better; as volume grows it makes more sense to modularize at different layers of the stack.
• Apple appears to be going backwards — re-integrating not just handset + OS but all the way down to the chip — possibly because the mobile market is large enough to capture economies of scale internally, getting both the cost advantage of modularity and the advantage of integration.
• Intel’s structural risk: a $15B fab has to stay full; with PCs flat (and possibly declining if Apple’s M1 wins), an underused fab is a massive problem.
• Intel’s deeper problem is cultural — immense engineering talent, but the company cannot do customer service and has decades of telling others “how it’s going to be,” not accepting it is no longer in control. The new CEO’s task is to knock heads together; finance people shouldn’t be running a company that should be run by engineers.

Notable quotes

Keep in mind that 5G in China is—the network is well established. And the overwhelming majority of phones being sold are 5G phones. And so I think there was some level of anticipation for us delivering an iPhone with 5G. And so, iPhone did extremely well.

Because a leading edge fab today costs $15 billion to build, and very few companies can afford that and the economics just work out for fabless — you don’t want to have to be building your own fabs and designing your own chips. The ability to abstract those two and to specialize in one side of it or the other, has allowed this big flowering of design companies to chip companies to take off.

Qualcomm has a little bit of a lowest common denominator problem in the sense that even if they have a really advanced chip, they have to segment it across multiple customers. They have to have a high end one and a low end one and it has to have all kinds of different features and so I think that’s a little bit of a liability versus Apple, who can do their A series and just optimize it for exactly what they need.

Vertical Integration — they’d be top to bottom and that fits the classical Clayton M. Christensen model where when you’re new to a market, you have to do everything, because it’s really hard and makes it work better. But over time as volume comes along, it makes more sense to modularize at different levels of the stack, and that’s what we’ve seen happen.

My read on Apple is they are going backwards, they are actually starting to completely vertically integrate not just the handset and the OS, but now all the way down to the chip level. Is this is a function of just how large the mobile market is where they can get sufficient economies of scale, even though they’re doing it themselves, that they get both the cost advantage that would usually go with modularity, but also the advantage of integration.

I think the issue that I see that I worry about for Intel is you mentioned at the top of our recording that a new fab cost $15 billion. And that means that, that fab has to be full. And if that fab is not full, then you’re going to have a massive problem. And the worry for Intel, as PCs have flat-lined, Apple might cause it to decrease if the M1 ends up being as much of a winner as they think it is.

Ultimately, the problem Intel has is a cultural problem. I have a lot of friends at Intel, I think the world of their talent, they’re immensely smart people there, but there’s something about the culture that’s not working. And you and I both know it they cannot do customer service, they’ve tried for years, and they just cannot get it right.

This is the hope for the new Intel CEO, someone’s got to come in, and knock heads together and I’m just so skeptical that happening because we have decades of this company telling others how it’s going to be and not accepting that they’re not in control anymore.

Intel had a finance person running it. I’m a finance person, finance people shouldn’t be running that company. It should be run by engineers. There’ve been a series of bad decisions, it’s not all on the operations team. And so maybe they can turn around, they’ve done it before, memory failed and they went to processors, they made it big.

Notes

• 5G matters more if (1) it is widely available and (2) if all of your computing, including huge amounts of streaming video, are done on your phone.
  • If China’s consumer behaviors are ~3-5 years ahead of the US then the explosion in video is just getting started — relevant to Twilio daily-sentiment tracking.
• fabless model & consolidation: what impact, if any, has the switch to a fabless model had on consolidation? Has it been a driver of it, or a beneficiary of it?
  • When this all started, the big revolution was the shift from doing your design vertically integrated with hardware (not just manufacturing) to a merchant silicon model.
• Apple re-integration question: when does mobile saturation happen, if at all?

How it connects

• Jay Goldberg — the interviewee; semiconductor analyst on chips, fabs, and the industry’s modularize-vs-integrate cycle.
• Intel — the central worry of the interview: full-fab economics, cultural/customer-service failure, and the new-CEO turnaround hope.
• Apple — the counter-example: re-integrating down to the chip (A-series, M1) against the industry’s modular grain.
• Qualcomm — the “lowest common denominator” merchant-silicon foil to Apple’s bespoke optimization.
• fabless — the $15B-fab economics that made design/manufacturing specialization the default.
• Vertical Integration / Clayton M. Christensen — the integrate-then-modularize framework the whole conversation runs on.
• 5G — the demand driver behind the strong China iPhone cycle.
• Twilio — daily-sentiment connection on the coming explosion in mobile video.