IC Manufacturing as a Service — Intel Edition

Sam Lin
5 min readOct 24, 2021

--

What is the future held for Intel in 2024? What could be a pivot Intel may still do now?

While Intel is still the No1 Integrated Circuit(IC) vendor with $73B revenue & 16% market share in 2020. The market feels like it’s a falling giant. This week Intel reported a 5% growth to $18B revenue in Q3 2021. But, its stock price dropped more than 11% as many concerns of the heavy capital investment is going to hurt the margin. Let’s take a different look at its future.

Intel Q3'21 Revenue

What the future holds for Intel in 2024

I bet no one can do a better job than new Intel CEO, Pat Gelsinger on reviving Intel by IDM 2.0. But, is that enough? Even it’s too early to tell, I can see two diverse theories. Depending on which thesis you pick, Intel’s future can be very different.

  1. Getting worse before getting better. As Gelsinger suggested, it takes 2–3 years to build a few fab, and even longer for a modern mega-fab. In his ambition, Intel will catch up by 2024 & maybe an unquestioned leader in 2025. Because by then, 3 new factories should solve Intel CPU shortage problems. And, it may either have the surplus to either serve Intel Foundry Service (IFS) business or play pricing games to squeeze competitors.
  2. High margin no more. As System on Chip (SoC) has outperformed Intel’s discrete CPU architecture in terms of performance per watt & unit cost. It’ll be very hard for Intel to reclaim the lost lands anymore. Intel has to compete much harder from now on, and “competition is the loser’s game”. For example, the high margin IC has to be built by the advanced chip fabrication technology (< 10nm), and TSMC leads with more than 80% of the market share in 2020. Even Intel just renamed its process nodes, Intel 7, 4, etc. to align with TSMC & Samsung, Which may make marketing sense, Intel still has many to do & build beyond marketing. It takes much more for the B2B business than the consumer marketing, that Intel used to do well, e.g. Intel inside.

I’m leaning toward the latter because Apple will want to keep designing its chips. Microsoft & others are also trying to “control” their own destinies. Furthermore, Intel also completes with these big customers as an IC vendor. Which makes it hard for them to “trust” Intel to make their ICs. Let me know what you think. But either way, Intel has to learn how to play better with a key frenemy, TSMC. Which is a brand new game for Intel to learn.

Visual Capitalist: Projected Construction of new IC factory(2021–22)

Unbundling manufacturing from design

To buy or to build, is the question a good product owner has to constantly revisit. Previous to the new CEO, there were many talks about Intel CPU outsourcing to TSMC. Which makes the short-term sense to solve the Intel CPU shortage since 2018. But does it for the longer term? I bet Gelsinger does not think so because Intel is building 3 new fabs. On the other hand, Intel has been outsourcing other chips to TSMC, including some of its GPU. So how does that make sense?

Accenture: The inverted pyramid of semiconductors companies to fabrication houses.

The Modularity Theory can help us to think about such questions better. By the framework, we can see:

  1. Intel should make its high-end CPU in-house because it’s highly interdependent on advanced Intel manufacturing technology & capability. Vertically integrated have been giving Intel edges on time to market & performance. Because the high-end CPU is the core business to Intel & built the brand to sell more mid & low-end CPUs, Intel has to lead. And IC fabrication & even packaging technologies are the keys to lead.
  2. Whereas other chips & even low-end CPUs are “modular enough” to outsource because TSMC& etc. have created great “interfaces” to unbundle the manufacturing from the design for a while. For example for GPU, TSMC’s manufacturing technology & capability have been great in serving top GPU vendors: Nvidia & AMD. Also, TSMC & Samsung are competing in the manufacturing service business. Intel’s investment in such an area will make it even more “Performance Surplus”. Therefore, if Intel wants to complete better on GPU, it better focus on design or other core competencies to differentiate.
The Innovator’s Solution: Product Architecture & Integration

Pivoting for a better future

While semiconductor cycles are driven by fluctuating sales growth numbers and it’s likely more difficult to manage as it takes longer to build modern fabs. Therefore, oversupply may be a problem around 2024. Which is a major concern for Intel’s current strategy. Even IDF sounds great, competing with customers is a major concern for Intel. So, what could be a missing piece to increase Intels’ chance of success?

Given time semiconductor sales eventually recover to higher levels than before

2 is better than 1?

What if there are two Intels? One focus is on designing & selling the chips. And the 2nd one focus on IC MaaS. Sure, this is a hard pill for Intel to swallow from the current point of view. But, if Intel is serious about the new strategies, a split is necessary. To be fair, this is a tough maneuver to do for a giant like Intel. But if it’s done right, two Intels may succeed by collaborating in a highly aligned manner, and also innovating quicker for their own goods in a loosely coupled way. In other words, Intel has to create the right organizational architecture to “separate concern” & innovate 2 distinct businesses “in parallel”.

Reduced IC Unit Costs & Increasing Capabilities Drive New Applications

Full Disclosure

The opinions stated here are my own, not those of my company. They are mostly extrapolations from public information. I don’t have insider knowledge of those companies, nor a whatever expert.

--

--

Sam Lin

A Taiwanese lives in Silicon Valley since 2014 with my own random opinions to share. And, they are my own, not those of companies I work for.