The Learning Curve (Experience Curve) Volume Strategy
Price for the volume you'll win, not the costs you have today.
- Difficulty
- Advanced
- Time to result
- ~ongoing to results
- Steps
- 5
- Confidence
- 85%
The Learning Curve, developed by BCG's Bruce Henderson and Bill Bain and refined by Morris Chang during his time at Texas Instruments, holds that unit costs fall predictably as cumulative production volume rises. A company that forecasts long-run market demand and prices ahead of its current cost structure can capture volume faster than competitors, moving down the cost curve quicker, which further lowers costs and lets it reinvest in R&D and capacity for the next generation. Applied at TSMC, this became the engine behind aggressive capital spending decisions like the 2010 tripling of capex for the 28nm node: rather than reacting to costs, TSMC forecast where the market and its own costs would land, priced and invested accordingly, and used the resulting scale to become the dominant, lowest-cost foundry — a self-reinforcing flywheel.
Origin
Developed at Boston Consulting Group by Bruce Henderson and Bill Bain in the late 1960s/70s; refined in practice by Morris Chang while he ran Texas Instruments' semiconductor business, then applied systematically at TSMC.
Core principles
- 01Unit costs fall predictably as cumulative volume increases (the 'experience curve')
- 02Price and invest based on where costs will be at scale, not where they are today
- 03Winning volume early compounds into a durable cost advantage
- 04Fixed-cost, capital-intensive industries reward the player who reaches scale first
- 05The advantage snowballs: lower costs lead to more volume, which leads to even lower costs
How to run it
- 1
Forecast total lifecycle market demand
Estimate the total unit demand for a technology generation before committing capital (e.g., total 28nm chip demand across all customer segments).
Pro tip Use both a market forecasting group and a dedicated business development team to triangulate demand.
- 2
Price ahead of the curve
Set price based on the unit cost you expect to achieve at full volume, not your current cost, to win share early.
Watch out This can compress margins in the short term and requires financial cushion.
- 3
Commit capital and R&D at fixed, aggressive levels
Lock in R&D (e.g., 8% of revenue) and capex budgets tied to the volume bet rather than renegotiating them reactively each year.
- 4
Win the largest share of customer volume
Aggressively pursue being the primary or sole supplier for major customers so cumulative volume compounds fastest.
- 5
Reinvest scale advantage into the next node
Use the resulting cost and cash advantage to fund R&D and capacity for the next generation, repeating the cycle.
Pro tip Never assume you're 'done' scaling — Chang noted TSMC never planned to stop after two fabs once he internalized the curve.
In the wild
After locking R&D at a fixed 8% of revenue, TSMC's engineers identified 28nm as the technology 'sweet spot.' Chang tripled annual capex from roughly $2.5B to $6B, betting the balance sheet on becoming the volume leader before demand (driven by the smartphone boom) was even visible.
→ TSMC became the dominant leading-edge foundry, a position it never relinquished.
Common mistakes
Stopping expansion too early
TSMC's original plan reportedly considered halting after two fabs; Chang says a real student of the learning curve would never do this, because stopping early cedes the volume advantage to competitors.
Pricing to current costs instead of future costs
Pricing based only on today's cost structure rather than the cost you'll achieve at scale causes you to lose volume to a competitor willing to price ahead of the curve.
Is it for you?
Best for
Capital-intensive, fixed-cost businesses (manufacturing, chips, infrastructure) competing for scale-driven cost leadership.
Not ideal for
Low fixed-cost, low-volume businesses where scale doesn't materially reduce unit costs.
From the transcript
“the simple explanation of learning curve is that uh as you make more of one thing anything uh actually started with refrigerators and the cars…”
“I was a serious student of learning C and I would never I would never stop uh at just two baps”
From the episode
TSMC Founder Morris Chang
Morris Chang