Electricity: from Edison's "system" to Insull's "current by the meter"
The commercial prize went not to the inventor of the technology (S0), but to the player who sold the outcome as a metered service (S4) and locked in the standard and scale (S6). Pioneer ≠ the one who captured the money.
Period 1879–1935 · Case R1 · deep-research · 2026-05-31
Anchor case P024 sourceslayers L1+L2
⚠ Honest disclosure. Layers L1 (dossier, ≥3 sources on key players) and L2 (track: 4 cross-sections + S0–S6 + 7 questions) have been completed. L3 (cross-case Q-runs) is not performed here — it is populated during the R11 synthesis. Roughly ~80% of the key facts (dates, figures, business model) are source-confirmed; the analytical linkages and the pass through the framework are operator inference. The "immaturity window 1882–1893" is analytical annotation, not a citation. The thesis "Insull is the true winner" is a judgment call: in capital terms GE/Morgan won as well. A precise dated "parity point" for price per lumen (electricity vs gas) is not given by the sources as a single number — it is flagged [approximate]. This is the entry into gate G0; G0 itself still requires R2.
Cross-section 1
Technology maturity curve
"It worked" (demo 1879 / commercial 1882) ≠ "it sells like hotcakes." In the immaturity window ≈1882–1893, two incompatible designs competed (Edison's DC vs Westinghouse/Tesla's AC), and the technology was being pushed for sale head-on — which ran into the physical ceiling of DC (transmission of ~1–2 km). The turning point was the victory of AC (Chicago 1893, Niagara 1895) and Insull's model.
Caption: AI-2026 ≈ "inside the immaturity window" — the dominant design for agentic pipelines has not settled; the mass-adoption curve still lies ahead. The window annotation is operator inference; the anchor dates are verified.
Cross-section 2
Substitution matrix
What they TRIED to sell (dead end)
What ACTUALLY SOLD (caught on)
What did NOT sell
What was SUBSTITUTED (incumbent)
"Electricity as a technology"; dynamos/generators to factories; "the Edison system"; DC over long distances (physically unable to deliver)
First — electric light to wealthy districts, theaters, ships (isolated plants). Then — metered current to the masses (Insull). Then — cheap household appliances (Matsushita: lamp, iron, adapter)
Arc lamps for the home (too bright/dangerous); a pure DC city grid; "selling the technology" in the immature phase
Gas lighting (the principal casualty), candles (tallow/wax), whale/spermaceti oil, kerosene lamps; the work of lamplighters
Key: Edison deliberately engineered the system to a price competitive with gas — copying the business model of gas companies (central supply + meter) in order to enter an already-existing job-to-be-done, "give me light in the evening." He aimed not at a "new category" but at substituting a specific incumbent on the incumbent's own economics.
Cross-section 3
Turning-point pattern: the parity line
The turning point was not "the technology got better," but the superposition of three forces:
1 · Transmission parity in technology (the AC standard)
After Niagara (~1893–95), current could be transmitted over tens/hundreds of km — an entire region from one large station. This made scale possible.
2 · Price parity (price per lumen)
Price of 1000 lumens: $785 (1800) → $0.23 (1992) — down 99.97%. Filament efficiency in 1883 — 2.6 lm/W versus gas at ~0.25 and a candle at ~0.1. Pearl St already in 1882 — "at a price competitive with gas." [parity point — approximate]
3 · The Insull model (the true turning point for mass adoption)
demand-metering + load diversity + massing of production + regulated monopoly. Bills −32% in a year → current profitable at a low price → mass availability. The turning point is carried by S4 (the model), not S0 (the technology).
In parallel, Matsushita carried appliances across parity (a lamp lasting 40 h vs 3 h, an adapter 30% cheaper) — to the threshold of "like running tap water."
Value chain
Stages S0–S6: where the profit settled
Stage
What it was
Who held it
Where profit settled / got stuck
S0 Base technology
Generation and transmission (dynamo, lamp)
Edison (DC), Tesla/Westinghouse (AC)
By itself it carried no money; got stuck on "selling the technology head-on"
S1 Enabling practice (moat)
The AC standard + transmission grids (after Niagara)
Westinghouse/Tesla → the industry
Moat: the standard decided whose ecosystem would survive — without it, no scale
S2 The offer that sold
Light → metered current → cheap appliances
Edison → Insull → Matsushita
Profit appears here — what sells is the outcome, not the technology
The core of mass adoption. The model made the low price profitable
S5 Investment chain
Capital for stations and mergers
Morgan → GE (1892); banks → utilities
Financial capital consolidated production; later — the holding-company bubble
S6 Institutionalization (moat)
Patents, the AC standard, scale, regulated monopoly, the GE brand
GE, utilities, the regulator
The business became transferable without its founder (Edison was removed — GE lived on)
Triple refusal to sell the technology: (1) Edison sold not "the bulb" but the system/light-as-outcome (S2). (2) Insull climbed higher — he sold metered current as a service (the S2 outcome via the S4 model). (3) Matsushita sold a cheap end-appliance to the masses (S2). The main prize went to those who climbed to S4 (Insull) and locked in S6 (GE: standard + scale).
★ Headline
Map of actors by layer: surface vs hidden locks
On the surface, the "war of currents" was raging (whose lamp/dynamo). The real levers lay in three invisible locks — and the profit settled not with the inventors.
L0 SURFACEEdison (DC) · Tesla/Westinghouse (AC)
Mechanism: the public "war of currents" — inventors and integrators. Visible heroes, but not the holders of the margin. Edison was pushed out of GE, DC lost.
L1 CAPITALJ. P. Morgan (the formation of GE) · Insull's holding pyramid
Mechanism of power: whoever supplies the capital for consolidation appoints whose technology "wins." Morgan stitched together GE (1892), pushing out Edison; the first president was Coffin. Insull: control without ownership — ~$1M controlled $2.5B (1:2500); 95+ holding companies, 255 operating companies; collapse on 8 April 1932, ~600,000 shareholders → PUHCA-1935.
Mechanism of power: regulation and certification = barrier to entry and adoption gate.
NELA (Insull president from 1898): the largest propaganda campaign against public power; FTC investigation 1928–34.
UL + insurers (Merrill, lab 1894): the insurer would not cover a fire with uncertified wiring → a UL listing = a mandatory pass for an appliance to reach the market.
PUC-bargain (Insull's speech 1898, Wisconsin 1907): voluntary state regulation of tariffs in exchange for exclusive territory = a legal barrier to entry + cheap capital. "He designed his own cage."
AIEE (1884): a common technical language → interoperability of vendors' equipment → a market becomes possible at all.
Mechanism of power: the owner of a key node + patents = rent from the entire industry. The Stanley transformer (demo 20 Mar 1886, Great Barrington) — without it AC cannot be transmitted → it determined the outcome of the war of currents. Without the meter there is no Insull model; without the turbine there is no "massing" (Harrison St 6.4 MW 1894 → 16.2 MW 1903). The patent as a weapon: Westinghouse legally barred Sangamo from making induction meters until the end of 1910.
L4 COMPLEMENTORMatsushita (appliances) · GE Appliances
Mechanism of power: an appliance creates load → load feeds the utility → the utility promotes the appliance. Every watt is monetized twice (the sale of the appliance + the sale of the current to run it). Matsushita: the adapter socket 1920 (−30%), the bicycle lamp 1923 (40 h).
Mechanism of power: demand does not arrive on its own — it is constructed, because an underloaded station = a loss. Demonstration kitchens, volume-discount tariffs, appliances on installment. Electrification of homes: 1/7 (1910) → 35% (1920) → 70% (1930).
L6 KNOWLEDGECharles Proteus Steinmetz (GE)
Mechanism of power: he turned AC from an "art" into reproducible engineering → replicable by an army of engineers, and GE locked this in as a corporate moat. The law of hysteresis (AIEE 1892), the symbolic j-method (1893), the 1897 book, ~200+ patents.
Where power and margin actually settled: not with the inventors (L0), but in layers L1 (control of capital), L2 (rules / certification / the monopoly bargain) and L3 (the key component + patent). Three invisible locks: adoption (UL + insurer), monopoly (Insull's PUC-bargain — competition was killed not in the market but in the state legislature), replicability (Steinmetz + the patents on the meter/transformer).
Cross-section 4
Value-redistribution map: who lost, who won
Lost (casualties)
Gas lighting companies (displaced by the early 20th c.) — the principal casualty
Electric utilities (Insull) — a mass metered service
The coal industry (fuel for the stations)
Electricians / wiring, and later — all of household electrical equipment
Pioneer Edison ≠ the ones who made the money. Edison invented and built the first station, but he was pushed out of GE, and DC lost. The money was made by: (1) Morgan / GE — on capital and production; (2) Insull — on the mass metered utility ("the Henry Ford of electricity"); (3) Matsushita — on the cheap mass appliance. The irony: Insull himself went bankrupt (the pyramid's collapse in 1932) — but because of financial leverage, not a poor operating model; the utility model outlived him.
Transfer
Three rules for AI-2026
1
Sell "current by the meter," not "the generator." The prize went neither to Edison (the technology) nor to the seller of the "system/equipment," but to Insull, who sold the outcome as a metered (consumption-based) service. For AI: don't sell the LLM/agents/"the technology" — package an alienable outcome with a clear unit of consumption (a task completed, a ticket closed, an artifact delivered) and a metric the buyer is already accustomed to paying for. The turning point is carried by S4 (the model), not S0.
2
Wait for / create your own "AC standard" and play on load diversity. Mass adoption arrived once the design settled (AC after Niagara) and Insull learned to smooth load by mixing dissimilar consumers. For AI: while the standard for agentic pipelines has not settled — sell the outcome to narrow early adopters, but engineer the economics so that one "engine" serves heterogeneous clients (load diversification = low unit cost = mass adoption). Cheapness is a consequence of scale, not of a discount.
3
Aim to substitute the incumbent on its own economics — and don't kill yourself with leverage. Edison won the entry by engineering the system to a price competitive with gas. Matsushita crossed the threshold by making the appliance "like running tap water." BUT even Insull's correct model collapsed because of an over-engineered financial superstructure. For AI: (a) find a job the client already pays for, and cross the parity line on its metric (10–100× lower in time/price); (b) keep a healthy unit economics without excessive leverage — a superstructure bubble kills even the right model.
Which hidden layers to occupy (from the map of locks):
L2"UL for AI." Become the "insurable standard" — eval/safety/acceptance, without which an agent cannot be put into production (compliance, audit, insurance of AI errors). Don't build the model — become the gate through which other people's models reach production. The margin belongs to whoever issues the pass.
L2The regulatory bargain. Insull won by being the first to come to the regulator. Participate in shaping the framework (eval protocols, agent safety standards) = a barrier to entry built by the hands of the state.
L3The patented shovel. The meter/transformer are boring, but without them there is neither the Insull model nor AC. The analog: the "engine"/SDLC tooling + protection (contracts/IP/data moat). Sell not the gold (the model) but the shovel.
L5Load-building. Demand for current was constructed (demonstration kitchens, installments, volume tariffs). For AI: an underloaded engine = a loss → construct demand (demonstration outcomes, consumption credit, per-volume task tariffs) and maintain load diversification.