STRATEGYWeeks to result

Wright's Law Cost Curve Forecasting

Predict when emerging technologies become economically viable using production learning rates.

Problem it solves

Investors and strategists cannot predict when a technology's cost will drop enough to unlock entirely new use cases and markets.

Best for

Analysts, investors, or strategists evaluating nascent technologies where production scale drives predictable cost declines.

Not ideal for

Mature industries with flat learning curves or markets driven primarily by regulatory or demand-side forces rather than production economics.

Overview

Why this framework exists

Wright's Law states that for every cumulative doubling of a technology's key production metric, unit costs decline by a constant percentage—its learning rate. By identifying the right cumulative metric (e.g., kilograms launched to orbit, kWh of batteries produced), measuring the historical learning rate, and projecting future output doublings, practitioners can forecast cost trajectories and pinpoint the price thresholds at which entirely new applications become economically viable. ARC Invest applies this to rockets: every cumulative doubling of kilograms launched to orbit yields roughly a 17% cost decline, which predicted Falcon 9 reaching ~$1,000/kg and projects Starship reaching sub-$100/kg—the threshold that makes orbital data centers financially feasible.

Core principles

6 total
  1. Costs in technology-driven industries are not random—they follow predictable learning curves tied to cumulative production.
  2. The key variable is the cumulative production metric, not time; growth rate determines how fast the curve is traversed.
  3. Each doubling unlocks new economic thresholds that make previously unviable applications suddenly feasible.
  4. Learning rates are empirically measurable from historical data and tend to remain stable within a technology domain.
  5. Identifying the correct unit metric is the most critical and error-prone step—wrong metric, wrong forecast.
  6. Incumbent advantage compounds: earlier producers accumulate doublings faster, widening their cost lead over time.

Steps

6 steps
  1. Select the correct cumulative production metric
    Identify the single metric that best captures accumulated 'experience' in the technology—for rockets it is total kilograms launched to orbit, for solar it is total GW installed. Choosing the wrong metric produces misleading forecasts.
    Pro tipWhen in doubt, test multiple candidate metrics against historical cost data and pick whichever produces the most stable learning rate over time.
    WarningUsing time as a proxy instead of cumulative output is the most common error; it conflates pace of adoption with the learning curve itself.
  2. Measure the historical learning rate
    Plot log-log cost vs. cumulative output using at least a decade of data. The slope of the regression line gives the learning rate—the percentage cost decline per doubling. For space launch this has measured around 17%.
    Pro tipCross-validate your learning rate against analogous technologies (e.g., compare rocket learning rate to analogous aerospace hardware) to sanity-check outliers.
    WarningShort data series with sparse doublings can produce artificially steep or flat rates; aim for at least three full doublings before trusting the number.
  3. Establish the current baseline cost and cumulative output
    Anchor the forecast by documenting today's verified cost-per-unit and the current cumulative production level. For rockets, this means current $/kg to orbit and total kg launched to date industry-wide or for a specific provider.
    Pro tipUse the most capital-efficient producer's cost, not the industry average, if you are assessing competitive moats or best-case scenarios.
  4. Project future cumulative output doublings
    Estimate how many times cumulative output will double over your forecast horizon using realistic growth-rate assumptions. Model conservative, base, and aggressive scenarios. Each additional doubling applies the learning rate discount to current cost.
    Pro tipTie growth-rate assumptions to concrete demand drivers (e.g., satellite constellation buildouts, government contracts) rather than extrapolating historical growth blindly.
    WarningElon Musk-style timelines—or any visionary founder's stated schedules—tend to embed optimism bias; stress-test your schedule assumptions.
  5. Calculate forecast cost at each doubling milestone
    Apply compound learning-rate discounts to today's cost for each projected doubling. For example, three doublings at a 17% learning rate yields: cost × (1−0.17)³. This gives a cost curve over the forecast period.
    Pro tipBuild a simple spreadsheet with cumulative output on the x-axis and cost on the y-axis; the curve becomes a powerful visual communication tool for investment theses.
  6. Map cost thresholds to use-case viability unlock points
    List target applications ranked by the cost level they require to become economically feasible. Overlay these thresholds on the forecast cost curve to identify when each use case is expected to unlock. For orbital data centers, the unlock is sub-$100/kg to orbit.
    Pro tipTreat the unlock point as an investment entry signal—position before the cost curve crosses the threshold, not after the market has priced it in.
    WarningEconomic viability is necessary but not sufficient; also check for regulatory, infrastructure, and demand-side readiness before concluding a use case will actually scale.

Checklist

Saved in your browser

Examples

3 cases
Space launch: Falcon 9 to Starship cost trajectory

ARC Invest tracked cumulative kilograms launched to orbit across the commercial launch industry. With a measured ~17% learning rate, the Wright's Law model explained Falcon 9's arrival at roughly $1,000/kg—cheap enough to make Starlink's satellite internet economically viable. Projecting further doublings enabled by Starship's full reusability, ARC forecast sub-$100/kg launch costs, the threshold their model identified as necessary for orbital data centers to be economically competitive with terrestrial alternatives.

OutcomeThe framework provided a quantitative basis for ARC's conviction that orbital data centers transition from science fiction to viable business within this decade, contingent on Starship commercialization.
ARC Invest Big Ideas 2026; Daniel Maguire interview, Milk Road AI, May 2026
Solar panel cost forecasting (benchmark analogy)

ARC Invest has applied the same Wright's Law methodology to solar PV, tracking cumulative GW installed globally. Each doubling of installed capacity has produced a consistent ~20% cost reduction for decades. Analysts who used this framework in 2010 correctly projected that solar would reach grid parity in most markets by the early 2020s, enabling massive investment positioning ahead of the inflection—years before consensus recognized the shift.

OutcomeEarly Wright's Law adopters captured outsized returns by investing in solar manufacturers and project developers before grid-parity costs were widely accepted as achievable.
Battery learning curve applied to EV inflection timing

Using Wright's Law on cumulative GWh of lithium-ion batteries produced, analysts tracked the learning rate at roughly 18% per doubling. This allowed projection of the cost level at which EVs would reach upfront purchase-price parity with internal combustion vehicles—approximately $100/kWh pack cost. Investors who anchored their EV thesis to this threshold rather than to arbitrary calendar-year targets positioned more precisely in automakers and battery suppliers.

OutcomeThe framework separated noise from signal in EV investment timing, filtering out premature hype cycles while identifying the genuine inflection point driven by battery economics.

Common mistakes

3 traps
Using time as the independent variable instead of output
Plotting cost against years rather than cumulative production conflates the speed of adoption with the learning mechanism itself. This causes forecasters to miss that a slowdown in production growth will delay cost reductions, regardless of how many years have passed.
Ignoring that economic viability ≠ market adoption
Even when the cost curve crosses a viability threshold, adoption can stall due to regulatory barriers, grid interconnection queues, NIMBYism, or missing complementary infrastructure. Wright's Law forecasts cost; it does not guarantee demand materializes on schedule.
Applying a learning rate derived from one technology to another without validation
Learning rates vary significantly across domains—rockets at 17%, solar at ~20%, nuclear historically near 0% or even negative. Borrowing a rate from an analogous technology without empirical validation for the specific domain produces systematically wrong forecasts.

Origin story

How this framework came to be

Originally formulated by engineer Theodore Wright in 1936 while studying aircraft manufacturing. ARC Invest has extended it across solar, batteries, genomic sequencing, and space launch, using it as a core tool for timing technology investment theses. Application to rockets discussed by Daniel Maguire on Milk Road AI.

Source

Traced to primary
Source · VIDEO
Why SpaceX Wants To Put AI Data Centers In Orbit — Milk Road AI
Milk Road AI · 2026
Open source →

Related frameworks

Browse all Strategy →