Research · strategy for the AI market
To design the strategy of our products and our company in the AI market, we study how similar markets formed in earlier eras — markets born from a general-purpose technology. The first case we worked through is electricity. This is the story of what we found.
Inside: why · what we did · 9 findings · the pattern across 44 cases · what it means for AI · what's next
Why
Our task is to find the regularities and repeatable rules of entering a market and creating a new one. The AI market is the market of a new general-purpose technology; history has had several — electricity, the automobile, electrical and electronic engineering, computers. All of them travelled a similar path, from discovery to the rise of new industries. If we understand HOW big companies grew on them, those become transferable rules for us today.
We started with electricity and took several cross-sections: how, as a phenomenon, electricity replaced old functions; how it rebuilt the markets needed to produce it; what markets, products and services it created and changed; and, separately, how the very earliest penetration unfolded, when the physical phenomenon turned into the first technologies, products and businesses. The aim is to extract patterns and working strategies — and then study other markets the same way.
What we did
We built a structural schema: electricity → six base functions (light, motion, heat, cold, chemistry, signal) → types of mechanisms → the concrete products people buy. It is the «skeleton» of the whole market, from phenomenon to shelf.
And then the year-by-year picture: how infrastructure was built first, how consumption caught up with a lag, and how displaced markets died in sync.
Especially early on, people didn't buy «electricity» or «a motor» — they bought the effect: light in the home and on a ship, something that moves by itself, finished silver-plated tableware out of an electrochemical bath. The technology was the means; the product was the result.
On a young market you sell the result, not the technology. That is the first thing the data confirmed.
An honest caveat. "Sell the result" is a near-universal truth (it holds for cars and for CRM alike), so by itself it is a weak historical finding. The differentiating question sits one level deeper: why did people buy THIS particular result, and why then — what behaviour and way of life had to change for the result to become wanted. A separate "behavior" layer of the analysis addresses this (in progress).
Electricity closed needs that classical methods couldn't reach: light on ships and in theatres where gas was dangerous; lighting places a gas pipe couldn't get to. And while there was no grid, they brought the whole thing — dynamo, wiring, lamps — deploying a self-contained, turnkey system right on site.
Edison's first commercial lamp order was the steamship SS Columbia (1880): a ship, where gas means fire and there is no grid. Isolated plants dominated for years (891 plants / 230,674 lamps by 1887) — the «obvious» grid model was the minority at the start.
The big companies were built not by the authors of inventions but by those who took a defensible position and built a final business on the technology. And the largest new markets were entirely new services that hadn't existed before: radio, television, the telegraph.
The inventors (Tesla, Sprague) sold their patents and captured almost nothing. The rent went to whoever held the defensible asset — a patent pool, a network, regulated infrastructure, a brand.
As soon as access to electricity was brought into the home and the street, the next product didn't require building infrastructure again — it rode the pipe already laid. On that one base, one after another, light, radio, fridge, washer and vacuum took off — and each new generation penetrated faster.
The first product can be an almost-free «can opener» — it lays the rail to the customer; the money comes from the next products on top of the same base.
The most surprising thing in the economics: electricity and its appliances took only a modest share of a single consumer's budget. The market grew because the NUMBER of consumers grew sharply — households, factories, offices that simply couldn't afford a comparable service before, and now could.
The consumer benefit mattered above all where the alternative was expensive or impossible (cooking vs gas, light vs kerosene). Expensive appliances and access were sold on installment — as cars later were: «cheaper than keeping a horse». And the recipients of the money changed in the process.
The share of money paid for human labour in the cost of a kWh fell, while the share paid for capital rose — specifically for the purchase and the wear (depreciation) of EQUIPMENT: machines, boilers, networks — plus interest to investors. So labour was replaced not by «capital in general» but by investment in equipment and its depreciation. Fuel held steady throughout.
Money turned the market. Huge capital went into infrastructure, to cheapen and lock in access. But —
most of the capital went not into «hardware» but into getting people to START USING it — installment plans, marketing, a low early price — effectively financing future revenue.
The pairing of «capital + those who scale» is what delivered the penetration.
Profit-pool migration →$5M → $495M, absolute Margin flow (Sankey) →who paid for what → where it pooled Interactive graph →298 nodes · centrality × margin
The pattern across 44 cases
Separately, we collected 44 first-entry cases across all six functions (light, motion, heat, cold, chemistry, signal) — who sold first, what exactly, where the power came from, who bought — and looked for the regularities.
Three stable patterns. First: the «power source» at entry was set by the era — early functions took the battery, later ones rode the ready-made grid. Second: B2B, commerce and municipalities paid first — the mass home came in the next wave, on the pipe already laid. Third — the most important for us:
the information function (signal) was sold as a SUBSCRIPTION SERVICE from the very start — back on batteries: telegraph, telephone, alarms. Physical functions (light, motion, heat) entered as EQUIPMENT; selling a physical result «by the meter» was a later invention that needs a grid.
AI is an informational, cognitive function. By this pattern it — like the telegraph and the telephone — is more likely to sell natively as a service / by subscription rather than as a «box», and to enter through the commercial and B2B buyer before the mass home. → full analysis of the 44 cases and case cards
What it means for us in the AI market
These are still hypotheses for the next phase — but electricity hints at where to look.
A map of AI functions → products, but starting from what is ALREADY accepted and bought: where the service needs no proof — offer it and they'll buy.
Turn our position and competence into a final result that can be sold fairly broadly — the way they sold light and the effect, not «electricity».
Where neither the big players nor those doing something similar the old way can reach. Electricity's ships and theatres are our AI edge-cases.
Where the market already pulls — and where we CREATE a consumer who didn't use it before because it was too expensive/impossible. Growth lives in the second.
And the cross-cutting candidate rules: sell the result, not the technology; enter where the old way fails; lay a «pipe» to the customer and sell on top of it; remember that the market grows by the number of new consumers, while margin lives in a protected position — centrality by itself does not guarantee it.
Under the hood
The story walks the main findings, but behind it sits the full analytical apparatus. Here are all the cross-sections and interactives it is built on.
What's next
The next step is to work through the automobile and computers/the internet the same way (a preliminary check already shows the same rules hold, the difference being speed), and then move to an action plan for our company, starting from where we are today.
Assembly report (all materials) →navigation over every cross-section and exhibit Full synthesis →model, 10 laws, cross-case, transfer to AI Quantitative analysis →regressions, correlations, measured facts
This is a narrative of the results of our study of the electricity market. The diagrams are the real exhibits of our analysis (not illustrations); their labels are in Russian, from the source exhibits. The transfers to AI are hypotheses to test, not conclusions. Status of all "laws": candidates — verified on a single case; confirmation requires testing across ≥3 technological revolutions (auto, computer/internet, containerization).