ExpertsFMTSMC Founder Morris Chang
Acquired · January 27, 2025
5 Selected Lessons:
heuristic · 1 case · 1 framework · 1 priority · 1 reasoning_pattern · 1
Never compete with your own customers — the pure-play foundry wedge
Source
Acquired — TSMC Founder Morris Chang (January 27, 2025) · at 1:35:04 ▸ · at 1:41:34 ▸ · at 1:49:11 ▸ · at 2:20:02 ▸
When this applies
when deciding what parts of a value chain to participate in, where customers also operate in adjacent parts of that chain
The takeaway
When you can choose your position in a value chain, prioritize never structurally competing with the customers you depend on, even over short-term advantages integration might offer.
Why it works
By never designing chips or owning end products, TSMC removes any structural conflict of interest with customers, which builds a trust advantage no integrated competitor can match (even a competitor's own customers prefer the non-competing supplier, as with Samsung and Apple).
One level up
Structural trust, not just technical capability, becomes the dominant long-run differentiator in capital-intensive, relationship-dependent businesses.
Where it stops holding
This stops working when there's no viable independent architecture/design ecosystem (e.g., an x86-dominated world) to supply a pure manufacturing-only player with customers, so the specialization has nothing to serve.
Detail
hierarchy:
- 1. Never compete with your customers in their core business.
- 2. Build manufacturing excellence and customer trust as the differentiators instead of design/architecture.
- 3. Let customers (even fierce rivals like Apple and Samsung) both rely on you precisely because you have no stake in their competition.
rationale: Chang: 'I was explaining to you... customer trust, you know. I mean, they were always very superior, you know, Intel... None of them either trusted Intel or liked Intel.' Hosts: 'They are truly the only foundry at the leading edge that does not in any way compete with their customers.'
In their words
They are truly the only foundry at the leading edge that does not in any way compete with their customers. They don't have their own end product division. They don't design their own chips.
The learning curve as a pricing and capital strategy
Source
Acquired — TSMC Founder Morris Chang (January 27, 2025) · at 2:00:55 ▸ · at 2:04:13 ▸ · at 2:07:26 ▸ · at 2:36:28 ▸
When this applies
when competing in a fixed-cost, scale-economics business where unit costs fall predictably with cumulative volume
The takeaway
In a scale-economics business, price toward your projected long-run cost curve early to win volume, because volume itself is what gets you to that cost curve.
Why it works
Because unit cost falls with cumulative experience/volume, a producer who prices aggressively early (even below current cost) can capture disproportionate volume, which accelerates its own cost decline faster than competitors', creating a compounding, self-reinforcing lead.
One level up
In markets governed by experience/scale effects, the fastest path to a sustainable cost advantage is to act now as if you already have the scale you're trying to win.
Where it stops holding
This stops working when market-size forecasts are wrong by even modest margins (the hosts note 5-10% errors can tank profitability for a node generation), or when the business isn't actually fixed-cost/scale-driven.
Detail
procedure:
- Forecast total end-market demand at maturity for a given node/product.
- Price aggressively (even below current unit cost) early in the node's life to maximize share of cumulative volume.
- Let volume drive the actual unit cost down the learning/experience curve faster than competitors.
- Reinvest the resulting cost advantage into further price leadership and R&D for the next node.
rationale: Chang: 'I had a role in refining it to the point where a semiconductor company can use it effectively... I found it highly fruitful just as a thinking tool.'
In their words
The insight is that like the goal that you are playing for is to be the largest volume player kind of at the end of the game
Rising capex thresholds create natural-monopoly dynamics
Source
Acquired — TSMC Founder Morris Chang (January 27, 2025) · at 2:36:28 ▸ · at 2:40:27 ▸
When this applies
predicting industry consolidation in any capital-intensive business where the entry cost to stay competitive keeps escalating
The takeaway
In an industry where leading-edge capex requirements rise over time, expect consolidation toward a small number of players (potentially one dominant player) as an eventual structural outcome, not an accident.
Why it works
As the capital required to build next-generation capacity rises faster than most competitors' balance sheets can keep up, fewer and fewer players can afford to stay at the frontier, compounding the leader's advantage via the learning-curve/volume flywheel on top of the capex barrier.
One level up
Escalating fixed-cost barriers to competing at the frontier of an industry are themselves a powerful, self-reinforcing selection mechanism for monopoly or oligopoly.
Where it stops holding
This stops working when a technology discontinuity resets the capex requirement to a lower baseline, or when government intervention/subsidy artificially sustains multiple competitors.
Detail
approach: Track the absolute capex required to remain competitive at the frontier over time; if it is rising super-linearly, infer the market has natural-monopoly characteristics.
signals:
- current leading-edge fab costs ~$20B and rising toward $40-100B per the hosts' extrapolation
- capex growth tracking net income growth, reinforcing the leader's ability to outspend competitors
rationale: Hosts: 'How many players are really going to be left standing with the ability to deploy $100 billion to build a building with some machines in it? This market has natural monopoly characteristics.'
In their words
Right now, it costs, I don't know, on the order of $20 billion to build a new fab. Eventually, it will cost $40 billion, $80 billion, 100 billion. How many players are really going to be left standing with the ability to deploy $100 billion to build a building with some machines in it?
Don't lay off by subjective performance rating if you'll need the people back within a year
Source
When this applies
when a cyclical downturn tempts leadership toward a knee-jerk headcount cut, especially one dressed up as a performance action
The takeaway
Before cutting headcount in a downturn, calculate the full round-trip cost (severance + retraining + lost trust) against the likely duration of the downturn, and don't cut if you expect to need the people back inside that payback window.
Why it works
Performance ratings are assigned by hundreds of individual supervisors with no common bar, so a layoff built on them is arbitrary and destroys trust; separation costs plus ~6 months retraining time make rehiring within a year a net loss.
One level up
Decisions made from fear during a crisis should be tested against a simple cost-benefit math, not executed as reflex.
Where it stops holding
This stops working when the downturn is structural rather than cyclical and the roles genuinely will not be needed again, making the retraining-cost argument moot.
Detail
rule: If you need the laid-off people back within a year, you shouldn't lay off at all.
threshold: Separation expense ~ half a year of pay; retraining a replacement takes at least half a year.
rationale: Performance-based layoffs rest on hundreds of individual supervisors' subjective ratings and are not something employees will respect.
In their words
If in a year you have to hire people back, you have to hire the laid-off people back, then you shouldn't lay off because the layoff, the separation expense is usually half a year, about half a year. And it takes at least half a year to train a person. So if you need the people back within a year, you shouldn't lay off.
Chasing adjacent new businesses (solar, LED) instead of doubling down on a huge core market
Source
Acquired — TSMC Founder Morris Chang (January 27, 2025) · at 25:15 ▸
When this applies
when a company diversifies into adjacent markets while its existing core market is actually still underappreciated and about to become enormous
The takeaway
Before diversifying into an adjacent 'new business,' size the realistic ceiling of that market against the ceiling of the core business you may be underestimating.
Why it works
Solar was commoditized by state-subsidized Chinese competition and LED was blocked by a small patent cartel, while the core semiconductor market was uncapped by Moore's Law demand growth — so the 'safe diversification' instinct missed the far larger opportunity already in hand.
One level up
The biggest opportunity is often hiding in plain sight inside what looks like 'just the core business.'
Where it stops holding
This stops working when the core business genuinely is reaching a ceiling, in which case adjacent diversification is the correct hedge.
Detail
situation: In 2009 TSMC stood up 'new businesses' in solar cells and LED, placing a trusted CEO-candidate in charge of them, believing they were promising alongside the core IC business.
decision: The company invested resources and leadership time in these adjacent businesses.
outcome: Solar was ruined by heavily subsidized Chinese competition; LED's market was too small and patent-locked by a few incumbents. Both failed, and the executive leading them eventually left to become MediaTek's CEO, while the core IC business became a $600B+ market.
lesson: Chang: 'it's always interesting to me when companies think, oh, we should look at other new businesses, when in reality, semiconductors became a $600 billion a year market... You were already in the best market.'
In their words
It's always interesting to me when companies think, oh, we should look at other new businesses, when in reality, semiconductors became a $600 billion a year market. And, you know, solar is a small fraction of that. LEDs are a small fraction of that. You were already in the best market.