Working paper · Wright's Law · Cost curves

GPU compute is the fastest cost decline ever measured. 89.1%.

Wright's Law applied to 150 technologies, including the first systematic experience curves for carbon removal and AI compute. The headline numbers, where they came from, and what they mean for climate and capital.

Nick Gogerty · February 2025 · SSRN Working Paper · ~9 min read
150 Technologies analyzed
20.9% Median learning rate
89.1% GPU compute (highest ever)
5–16% Carbon removal range

The 90-second version

Costs for most things fall by about 20% every time we double how much we have made. That is Wright's Law, and it has been true for a hundred years across 150 technologies. This paper updates the dataset to 2024 and adds the two domains everyone is currently arguing about — carbon removal and AI compute.

The headline number is GPU compute: 89.1% cost reduction per doubling, the highest learning rate on record. Nuclear, by contrast, has gotten more expensive with every reactor built — an anti-learning rate of −88.1%. Carbon removal is still early, but the better-positioned technologies are learning at 5–16%, comparable to where solar PV sat in the 1980s.

The practical takeaway: before you forecast cost decline for any new technology, ask which side of the modular-versus-bespoke line it sits on. That matters more than the technology itself.

Key findings

Five things this paper actually shows

1finding-1 · learning rate distribution

Wright's Law holds. Median learning rate: 20.9%

Across 97 technologies with sufficient production data, the median learning rate is 20.9%, mean 20.1%. Two thirds of the curves (66 of 97) fit Wright's Law with R² above 0.80.

This is broadly consistent with Nagy et al. (2013), the most-cited prior validation, which found a median in the 15–20% range across 62 technologies. The slight upward shift comes from updating fast-learning digital technologies and adding AI compute.

What it tells you: when someone forecasts the cost of a new technology, the prior should be a learning rate near 20%. The interesting question is always which direction the technology deviates from that prior — and why.

Use this in
2finding-2 · gpu compute

GPU compute is the fastest cost decline ever measured

Cost per GFLOPS fell 89.1% per doubling of cumulative production (R² = 0.995, 12 observations, 2003–2024). A 667-fold reduction in 21 years.

Three things compounded to produce this: transistor scaling (the classic Moore's Law mechanism), architectural parallelism (GPU core counts grew from ~100 to over 16,000), and yield learning at fabs. None of these would have produced 89.1% on its own. Stacked, they did.

The implication for AI economics: LLM inference costs are now declining at roughly 69% per year, about 3× faster than Moore's Law alone. Frontier inference reaches cents per million tokens by 2028 if current rates hold. The marginal cost of intelligence approaches zero well before any physical limit kicks in.

Use this in
3finding-3 · carbon removal

Carbon removal is learning at 5–16%, comparable to early solar

First systematic experience curves for CDR. Mean learning rate 7.5% across eight removal technologies. DAC liquid solvent leads at 16.4%; DAC solid sorbent shows no learning yet.

The DAC split is the interesting result. Liquid solvent (Carbon Engineering-style) is approaching solar-PV-grade cost decline. Solid sorbent (Climeworks-style) costs initially rose as plants scaled from 50 tCO&sub2;/yr pilots to the 36,000 tCO&sub2;/yr Mammoth facility — first-of-a-kind plants typically need a manufacturing revolution before learning takes hold.

If DAC liquid solvent's 16.4% rate persists for three to four doublings, costs reach the $50–70/tCO&sub2; policy-relevant zone before 2035 with aggressive deployment. That is a conditional projection — CDR data is thin (3–8 points per technology) and forecast error grows about 2.5% per year.

Use this in
4finding-4 · nuclear

Nuclear is getting more expensive with every reactor built

Nuclear power exhibits anti-learning: a learning rate of −88.1%. Costs rise with cumulative capacity. The opposite of Wright's Law.

Why this happens: nuclear is bespoke and site-specific. Each plant is essentially a custom build, with escalating regulatory requirements, shifting safety standards, and institutional knowledge that has to be relearned every cycle because the gaps between projects are too long. The factory-style learning loop never closes.

The lesson generalises beyond nuclear. Modularity and manufacturability are prerequisites for strong learning. A technology can be physically promising and still fail to follow Wright's Law if it cannot be standardised, repeated, and improved on a short cadence. This is the test we should apply to every "promising technology" claim — including in CDR.

Use this in
5finding-5 · symbolic efficacy

The AI symbolic layer has grown by a factor of 100 million in twelve years

Φ-s (Symbolic Efficacy Index) compounds compute efficiency, energy efficiency, algorithmic efficiency, and deployment reach. From 1 in 2012 to ~95 million in 2024. Doubles every ~6 months (R² = 0.996).

Individual experience curves measure cost decline along a single dimension. Φ-s captures what happens when four independent improvement vectors compound. Compute cost down 13×, energy efficiency up 21×, algorithmic efficiency up 342×, deployment reach up 1,000× — over the same twelve years.

The total improvement exceeds Moore's Law over the same period by more than six orders of magnitude. Solar PV gains by more than seven. This is not a single technology getting cheaper. It is a phase transition in the complexity of the symbolic layer.

Use this in
Visualised
GPU compute is the fastest cost decline ever measured.
Wright's Law learning rate, % cost reduction per doubling of cumulative production. Selected technologies from a 150-technology dataset. Negative values indicate anti-learning — costs rising with cumulative production.
Learning rates across 9 selected technologies Horizontal bar chart. GPU compute leads at 89.1%. Hard Disk 56.4%, DRAM 42.8%, NAND 38.1%, Li-ion battery 36.1%, Solar PV 20.1%, DAC Liquid Solvent 16.4%, Wind Denmark 7.9%. Nuclear shown in red at minus 88.1%, the opposite direction. −80% −40% 0 +40% +80% GPU compute 89.1% Hard disk 56.4% DRAM 42.8% NAND flash 38.1% Li-ion battery 36.1% Solar PV 20.1% DAC liquid solvent 16.4% Wind (Denmark) 7.9% Nuclear −88.1% Highest ever measured ↓ First systematic CDR estimate ↓ Anti-learning ↓ Information & storage tech Energy Carbon removal Anti-learning
Source: Gogerty (2025), Experience Curves Extended, Table 2 & Table 3. Dataset: Santa Fe Institute Performance Curve Database, IRENA, Epoch AI, CDR.fyi. Selected 9 of 97 fitted technologies; full dataset CC BY 4.0.
Audience

Who this paper is for

Climate VCs & technology investors

If you are forecasting cost decline for a deep-tech investment, the prior is a 20% learning rate. The interesting question is what makes your bet faster or slower than that.

Policy & climate strategists

If a technology is bespoke and site-specific (nuclear, large hydro), don't assume it will follow Wright's Law. If it's modular and factory-built (solar, batteries, DAC liquid solvent), expect cost decline conditional on deployment.

CDR buyers & portfolio managers

Use the 7.5% mean CDR learning rate as a baseline. Allocate more to technologies that are already showing 12%+, less to those flatlining. Treat the early estimates as preliminary indicators, not stable parameters.

AI & compute strategists

If frontier inference costs are halving annually, your three-year capacity plan is wrong. Model on a one-year halving, not three. The marginal cost of intelligence is going to zero before physical limits.

Cite this paper

Citation

Gogerty, N. (2025). Experience Curves Extended: Wright's Law Across 150 Technologies, Including Carbon Removal and Artificial Intelligence. Carbon Finance Labs SSRN Working Paper. https://carbonfinancelab.com/wrights-law-extended/
@techreport{gogerty2025experience, author = {Gogerty, Nick}, title = {Experience Curves Extended: Wright's Law Across 150 Technologies, Including Carbon Removal and Artificial Intelligence}, institution = {Carbon Finance Labs}, type = {SSRN Working Paper}, year = {2025}, month = {February}, url = {https://carbonfinancelab.com/wrights-law-extended/} }
Gogerty, Nick. "Experience Curves Extended: Wright's Law Across 150 Technologies, Including Carbon Removal and Artificial Intelligence." Carbon Finance Labs SSRN Working Paper, February 2025. https://carbonfinancelab.com/wrights-law-extended/.
Related

Continue reading