← Все материалы П49
P49 · Case 03 · Market Formation · Electricity

Electricity: Market Systems Map v2 (full corpus)

The money is not made by whoever owns the technology, but by whoever sells the outcome per unit of consumption and the process re-engineered around the technology — with a lag of 20–40 years. The map of how the electricity market formed, as a working template for the AI-2026 wave.

135 nodes / 133 edges / 77 events 8 deep tracks · ~180 sources coverage GO
Foundation: 8 deep tracks (~17,000 words), consolidated into a connected DAG. Edge/node ratio ≈ 0.99 — a connected graph, not a "scatter of facts."
Corpus coverage · A8

Completeness of the cross-section: 11/11 layers · 5/5 periods · verdict GO

Nodes across the 11 layers

INFRA24 TECH18 THEORY16 PRODUCT16 MARKET16 ACTOR15 BUSINESS9 CAPITAL9 LABOR6 PRACTICE4 EVENT*2
*EVENT nodes are few — the 77 events are broken out in a separate table; not a gap.

Nodes across the 5 periods

P1 <188026 P2 1880–9238 · peak P3 1893–0729 · design P4 1908–2930 · mass P5 1930+12 · reshuffle
VERDICT A8: GO
Low confidence — only 2 of 135 nodes · cross-confirmed hubs (≥2 tracks) — 21 (GE, Insull, ComEd, EBASCO, PUHCA, load_factor). A mild CAUTION: reinforce the hubs with a 3rd source before publication.

4 recorded clarifications / debunked myths

TopicWhat the sources actually show
"Electricity saved the whales"A myth. The whaling fleet collapsed earlier (sperm-oil peak in 1856; fleet 735→39 by 1876) because of kerosene (Drake, 1859), not because of the lamp.
Price parity per lumenNo single dated figure (electric vs. gas) is found in the sources — there is a trend (Nordhaus) and the fact that it was "valued below gas." Flagged unverified.
Engineering talent ×8 (1905→30)Not directly verified — a proxy is given (Michigan: peak of 84 degrees in 1924/25). Approximate.
Gas migrationThe incumbent did not die: gas lost lighting but took heating/cooking; the Welsbach mantle extended gaslight by ~50 years.
Systematized picture

Three phases of market formation: science → standard → mass adoption and reshuffle

PHASE 1 · P1 · 1820–79
Science → reproducible technique

Ørsted 1820 → Faraday 1831 → Maxwell 1865; later, calculation-based engineering (Steinmetz). Physics became technique.

No money was made here.

PHASE 2 · P2–P3 · 1880–1907
Ferment → standard → business model

Three "locks":
Component: the ZBD transformer at 98% efficiency (Ganz, 1885); Shallenberger's meter (1888) made consumption measurable; the Curtis turbine at Fisk St — 5 MW when the next-best was just 1.5 MW.
Standard: AC won on transmission economics. Chicago 1893: Westinghouse $399k vs. GE $554k. 60 Hz US / 50 Hz Europe locked in by 1891.
Business model: Insull — the Wright tariff + load diversity + massing → bills −32%/year, price 20¢→2.5¢/kWh (1892→1909).

PHASE 3 · P4–P5 · 1908–55
Mass adoption and reshuffle

Margin and power settled into the hidden layers. The financial superstructure (holding companies) inflated and burst. Durable margin lives in the patented complement (GE: the lamp as a consumable).

Archaeological cross-section · period × layer

The excavation trench: where the "hole" was that the winner filled

The filling "creeps" from THEORY toward MARKET/CAPITAL/LABOR. The winner's point of entry is the empty business-model layer (BUSINESS in P3), not the technology layer.

Layer ↓ / Period → P1 1831–79P2 1880–92P3 1893–07P4 1908–29P5 1930–35
THEORYFaraday's induction (1831); MaxwellAC calculation (Steinmetz, 1892–93)
TECHdynamo, lampEdison's lamp; Tesla/Westinghouse AC systemStanley's transformer; Curtis's turbineturbines 1.5→208 MW; small household motors
PRACTICEisolated plant vs. central stationload diversity + metered tariffunit-drive factory reorganization; load-building
PRODUCT"light" (a service)metered lightiron→vacuum cleaner→radio→refrigeratorrefrigerator 8%→44%
ACTORFaradayEdison, Westinghouse, Tesla, MorganInsull, GE, UL, AIEE, NELAEBASCO, Hotpoint, MatsushitaFTC, Congress (PUHCA)
MARKETlighting (niche)urban utility marketsappliance market + industrial electrificationmature markets
INFRAPearl Street (1882); isolated plantsNiagara (1895); urban gridsnational generation growth ×20 (1902–30)the grid as a utility
BUSINESSselling the "system"/equipmentregulated-utility metered; the PUC bargain (Wisconsin 1907)vendor financing (EBASCO); installment plans; holding-company pyramidsthe collapse of the holding companies; the appliance business holds
CAPITALMorgan → GE (1892)banking capital into utilitiesEBASCO $1.25B / 33 states (1926); Insull's pyramidthe 1932 crash; PUHCA 1935
LABORthe first engineering programsGE "Test," Westinghouse Night School; engineering degrees ×8150 power programs (1935)
EVENTlamp 1879S.S. Columbia 1880; Morgan's house 1882; war of the currents 1888–93Chicago 1893; Niagara 1895; PUC Wisconsin 1907homes 30%→70%; industrial capacity 5%→75%Insull's collapse 1932; PUHCA 1935
★ Flagship finding for AI-2026

The productivity paradox: value comes not from access to the technology, but from reorganizing work around it — with a lag of 20–40 years

The share of electric drive in industry: <5% (1899) → 50% (1919) → 75–78% (1929). But the productivity jump arrived only in the 1920s — with a lag of ~20–40 years. "Naive electrification" (a motor in place of the steam engine on the old line shaft) yielded ≈0; the gain came only after unit drive + replanning the factory around flow (Devine, 1983, 3 stages).

Stage 1 · group line shaft steam engine → shaft → belts Stage 2 · naive (motor → shaft) M the same old shaft → gain ≈ 0 Stage 3 · unit drive + flow a dedicated motor per machine + a conveyor → the jump

The lag curve: adoption rises at once, productivity only 20–40 years later

1899191919291939 <5% 50% 75% lag ~20–40 years share of electric drive productivity growth
Paul David (1990) built his paper directly as an analogy of the dynamo ↔ the computer on Solow's 1987 paradox — and the IT payoff again arrived with a lag of ~20 years. Today the third term in the series is AI.

Takeaway for the CEO: "an LLM in place of the line shaft" = zero. The wave's margin is captured by whoever sells the reorganization of the process around agents, not access to the model. A rigorous historical argument for the thesis "sell the outcome": the outcome = the re-engineered process.

Hypotheses H1–H5 · verdicts on the data

All five confirmed — with the key evidence and the figure

#HypothesisVerdictKey evidence
H1 Removing a barrier/setting a standard does not lock in the market → a second moat is needed UL + insurance (the admission gate) and the AC standard opened the market to everyone. The hold came from a second moat: Insull — a regulatory monopoly + capital; GE — the lamp patent + vendor financing.
H2 Growth by embedding into confirmed demand + collapsing the price Light entered already-paid-for demand (gas), priced below it. The first customer is not the mass market, but a business with acute pain: S.S. Columbia (1880), before Pearl Street. Isolated plants beat the grid 702:58 (1886).
H3 The "1C army": victory through training an army of implementers GE Test Department (1904): 575 people, 67 colleges, 18 nationalities, on rotation → ready-made engineers into the industry. The winner itself built a talent forge as part of the capture infrastructure.
H4 Winners-by-capital: infrastructure + financial wrap around the channel (with a caveat) EBASCO (1905): the equipment vendor became a bank for the utilities — $1.25B / 33 states by 1926. BUT the superstructure ≠ durability: the holding companies (8 groups ≈73–75% of investor-owned utilities by 1932) inflated and burst → Insull's collapse 1932, ~600k shareholders → FTC 96 volumes → PUHCA 1935.
H5 Margin migrates; the most durable is the patented mass-market complement What survived through all of it was GE's appliance-and-patent business — the lamp at 71.9% of the market, inelastic demand, profit on appliances through the Depression. Infrastructure gets regulated/bankrupted — the patented complement does not.
Ruleset · §4 of the synthesis

6 rules for transfer to the AI wave

1
Don't sell "current" — sell the outcome per unit of consumption + the re-engineered process. Lag arbitrage: whoever first redesigns the work around AI captures the wave.
2
The first customer = a business with acute pain + money + a need for autonomy (your "S.S. Columbia"), not the mass market. Construct the mass market afterward: demos, installment plans/pay-per-result, diversification of the "engine's" load.
3
A second moat is mandatory: a standard/acceptance (a "UL for AI") OR a patented complement OR financing the customer's adoption (EBASCO / vendor financing) OR a regulatory bargain. Removing the barrier does not lock the market in by itself.
4
Build the army of implementers yourself (GE Test = 1C franchisees): cheap tools + training = a distribution moat.
5
Keep the margin in the patented mass-market complement, not in infrastructure/the superstructure. And without excessive leverage — a superstructure bubble kills even a sound model.
6
The incumbent migrates, it does not die (gas → heating): plan not to "kill" it, but for where the displaced party will go — and who will profit from it (copper/coal/appliances = the unexpected winners).