VC Investment Thesis: Riding External Cost Curves
The Core Concept
A technology’s viability and market disruption is primarily driven by its ability to ride an existing, rapid decrease in costs — powered by external macro-forces (Moore’s Law, battery density improvements) rather than internal R&D. When a startup builds to ride such a curve, it shifts from being a “toy” that undershoots market needs to a disruptive force as the external cost decline makes it exponentially more accessible than incumbents.
The pattern has a name: Wright’s Law. Every doubling of cumulative production drops unit cost by a predictable percentage, independent of any one company’s R&D spend. Disruption follows when a technology crosses a cost threshold, not when a company invents its way across it.
The key question is not “what technology is getting cheaper” — it’s “who captures the value when the cost curve moves, and why them specifically.”
The Investment Framing: Cost Curves as Infrastructure
Treat the declining cost curve like a new highway being built. The highway builder rarely makes the money — the gas station, the motel, the distribution warehouse at the exit captures the value. Find the business that becomes newly viable (or newly dominant) at a specific cost threshold, and fund it slightly before that threshold hits.
Most investors chase the cost curve itself — fund the solar panel manufacturer, the battery chemistry startup, the sequencing hardware company. That’s usually wrong, because the curve is exogenous. You’re funding a commodity producer racing toward zero margin.
Three Structural Positions Worth Funding
1. The aggregator / distributor who owns demand. When input costs fall, whoever controls customer relationships and distribution wins. Rooftop solar installers like Sunrun didn’t invent solar panels — they owned the homeowner relationship and installation channel. As panel costs fell, their margins expanded.
2. The platform that becomes newly viable at a cost threshold. Some business models are simply impossible above a certain cost level. Whole-genome sequencing at 200, you get consumer genetics, newborn screening, oncology monitoring, and pathogen surveillance — four entirely different markets.
3. The picks-and-shovels player one layer above the commodity. Not the manufacturer of the cheap thing, but the company that makes the cheap thing useful. As GPU costs fell, value went to cloud providers (aggregated capacity), then to model developers (built on top) — not to GPU makers.
The Three Investment Buckets
Bucket 1: Curve Quality — Is this real and predictable enough to build on?
- Learning rate (% cost decline per doubling of cumulative production)
- Remaining halvings before saturation
- Exogeneity — is the decline driven by macro forces or one company’s R&D? (Latter is fragile)
- Reversibility risk — regulatory, geopolitical, supply chain shocks that could stall or invert
- Historical consistency across multiple cycles and manufacturers
Bucket 2: Threshold Dynamics — At what cost point does the real market unlock?
- Distance to the critical threshold where mass adoption becomes economically rational (not just technically possible)
- Number of discrete thresholds, and what new use cases each unlocks
- Whether adoption is automatic at threshold or requires additional behavior change / infrastructure
- Time to threshold vs. company runway — can the business survive the wait?
Bucket 3: Value Capture Position — Does this company get structurally better as the curve arrives?
- Layer in the stack: commodity manufacturer vs. aggregator vs. application vs. picks-and-shovels
- Whether competitive moat strengthens or weakens as input cost falls
- Demand ownership — does the company control the customer relationship independent of input cost?
- Market structure at maturity — consolidation window or immediate commoditization?
Sequence matters: Bucket 1 is table stakes. Bucket 2 determines timing and survival risk. Bucket 3 is where the actual returns live.
Types of Cost Curves
1. Manufacturing Learning Curves (Wright’s Law) — Solar PV (20% decline per doubling for 4 decades), lithium-ion batteries (86% decline since 2013), semiconductors, consumer electronics. Endogenous to production; no single company owns it.
2. Observation-Law Curves (Moore’s Law and derivatives) — Time-based, self-fulfilling because capital allocation organizes around the expectation. Compute, storage (Kryder’s Law), bandwidth. When the expectation breaks, the curve slows.
3. Commodity & Resource Extraction Curves — Lumpier and more reversible. Lithium, cobalt, polysilicon, shale. Geopolitical shocks and cartel behavior can invert them.
4. Infrastructure & Network Density Curves — Cost per unit of service falls as physical/digital infrastructure reaches density. Telecom, cloud computing, ride-hailing, last-mile logistics. Tends to produce winner-take-most dynamics faster than manufacturing curves.
5. Regulatory & Policy-Driven Curves — Most fragile. Offshore wind (2015–2022 cost declines heavily policy-driven, not just hardware learning), drug approval pathways, fintech licensing. Can reverse overnight with a policy shift.
6. Data & Algorithmic Curves — Newest, least well-understood. AI inference cost, computer vision, drug discovery. Can see step-function discontinuities from architectural breakthroughs (transformers, AlphaFold). Higher-variance than Wright’s Law.
Curve Stacking
The most powerful disruption setups occur when two or more curve types compound simultaneously. Solar + storage works because solar manufacturing and battery manufacturing curves decline in parallel. AI inference works because GPU manufacturing, data accumulation, and algorithmic efficiency curves run concurrently. When multiple curves stack, threshold dynamics accelerate non-linearly and the window for building a dominant position is shorter than most investors expect.
Key Risks
- Curve stall: Build the investment case assuming the curve moves at half its historical pace.
- Window compression: The curve moves fast enough that the window for building a durable business is shorter than the fundraising cycle.
- Commodity trap: Funding the curve driver rather than the value captor.
- Threshold miss: Adoption requires more behavior change or infrastructure than anticipated even after cost parity.
Industries Riding External Cost Curves
Solar PV, EVs / lithium-ion batteries, genomics / DNA sequencing, semiconductors / compute broadly, drone hardware, cloud storage / bandwidth, onshore wind.
Counter-examples (curves that don’t work): Nuclear power (bespoke on-site construction, ~60% of LCOE is capex that can’t be standardized); large-scale hydropower; regulation-intensive projects generally.
See also: minilateralism-climate-pathway | 10-technology