VC
About
Media
Startups & Experiments
Writing
Topics
Resources
Contact
Posts
VisionaryJuly 8, 2026 / 11 min

How Business Models Evolved: From Owning the Factory to Owning the Loop

A century of business model evolution in one essay: how companies went from making money on owned assets and headcount to software margins, network effects, and AI-era compounding loops — and what founders should do about it.

VCVik ChadhaFounder • Operator • Investor
How Business Models Evolved: From Owning the Factory to Owning the Loop
Part of the SaaS Scaling series
Explore topic
First-hand Experience
25+ years building companies across three business model eras(Co-founded Backupify (acquired by Datto) and UnifyCX (6,000+ employees), now building AI-era companies through Scalable Ventures)
Strip any company down far enough and you find the same three questions: how does it create value, how does it deliver that value, and how does it capture some of it as money? Everything else — the org chart, the tech stack, the fundraising strategy — is downstream of those answers. That trio is the business model, and it is the single most underexamined choice most founders make. They agonize over product and hiring while inheriting their business model unconsciously from whatever era they grew up in. That inheritance is now dangerous, because the answers have changed twice in a century — and the second change is still in motion. I have had the odd fortune of building companies across all three eras: a labor-based outsourcing business that scaled to more than 6,000 people, a SaaS company that rode software economics to an acquisition, and a portfolio of AI-era companies at Scalable Ventures that make money in ways neither of the first two could have. This essay is the pattern I see when I line those experiences up: what the 20th-century model actually was, what software changed, what exponential models change again, and what founders should take from the shift.

How Did 20th-Century Business Models Make Money?

The dominant business model of the 20th century was linear: value was created in owned assets, delivered through owned distribution, and captured as a margin on each unit sold. To make more money, you made more units, which meant more factories, more machines, and more people. Revenue and cost grew on roughly the same line. Ford's River Rouge complex is the archetype. Raw materials went in one end and finished cars came out the other, with Ford owning as much of the chain in between as it could manage. That was not a quirk; it was the winning strategy of the era. When coordination between firms was slow and expensive, owning everything was how you got quality, speed, and cost under control. The great 20th-century companies — the automakers, the oil majors, the consumer packaged goods giants — were all variations on this theme: asset-heavy value creation, physical distribution, per-unit capture. Three properties of that model are worth naming, because everything that follows is a reaction to them:
  1. Growth required proportional inputs. Doubling output meant something close to doubling plant, equipment, and headcount. There were economies of scale, but they flattened. No factory got meaningfully cheaper to run because the previous million units had been produced.
  2. The constraint was capital. Because growth meant assets, the scarce resource was money to buy them. This is why the 20th century belonged to whoever could finance the most capacity — and why banks and public markets sat at the center of the economy.
  3. The moat was physical. Competitors had to replicate your factories, your distribution network, your shelf space. That took years and fortunes, which made incumbency durable.
None of this was a mistake. It was the optimal design for a world of physical goods and expensive coordination. The trouble started when the goods stopped being physical.

What Did Software Change About Making Money?

Software broke the link between units sold and cost incurred. The first copy of a program might cost millions to create; the ten-millionth copy costs approximately nothing. Microsoft was the early proof: it captured a per-unit price — the license — while paying almost no per-unit cost. That combination, industrial-era pricing on top of near-zero marginal cost, produced gross margins the linear world had never seen and funded three decades of software eating everything. But the deeper shift was not the margin. It was that the best software businesses stopped selling units at all and started selling relationships. SaaS replaced the license transaction with a subscription: value delivered continuously, revenue captured continuously, and — the part that changes everything — the customer relationship compounding instead of resetting to zero after every sale. I lived this one directly. Backupify started as a dorm-room idea and grew into a company protecting 160 petabytes of data for 40,000 customers before Datto acquired it. Note what the asset was: not servers, not offices, not headcount. The asset was the recurring relationship and the accumulated trust of being the system a customer never has to think about. Each new customer made the economics better, not merely bigger — infrastructure costs amortized across more accounts, and the product hardened against edge cases every new environment surfaced. A 20th-century company gets bigger as it grows. A well-built software company gets better as it grows. The internet era added a second invention on top of software margins: the network model. Google's search results improve because people use Google, which attracts more people, which improves the results. Airbnb built one of the world's largest lodging businesses while owning no rooms; its asset is the two-sided network itself, and it captures value as a take rate on transactions it neither manufactures nor fulfills. In these models, the users are not just customers — they are unpaid contributors to the product. Value creation was crowdsourced; value capture stayed with whoever owned the point of aggregation. By the 2010s, then, the frontier model looked like this: near-zero marginal cost of delivery, recurring or transactional capture, and a moat made of accumulated data and network position rather than steel and shelf space. The constraint had moved too. Capital was abundant; the scarce resources were distribution and talent. Which is exactly why the next shift matters so much — because it attacks the talent constraint directly.

What Makes an Exponential Business Model Different?

Software collapsed the marginal cost of distribution. AI is collapsing the marginal cost of cognition — the analysis, judgment-adjacent work, and production that until now could only be bought as human hours. When both distribution and a growing share of the work itself approach zero marginal cost, you get what I call an exponential business model: one where the core loop compounds instead of merely scaling. The distinction matters, so let me be precise. A linear business grows by adding inputs. A scalable business (classic SaaS) grows without adding proportional inputs. An exponential business grows and improves through the same motion, because usage feeds a loop that makes the product better, cheaper, or smarter for the next user. Three loops show up again and again:
  1. The data loop. Usage generates proprietary data; the data improves the model; the improved product wins more usage. This is the AI-era version of a network effect, and it is buildable in markets far too small or unglamorous for classic network effects to form. HiveDesk, one of our portfolio companies, runs on exactly this: custom models trained on years of workforce scheduling data — shift patterns, preferences, compliance constraints — that no generic tool can replicate. The scheduling engine is not a feature bolted onto the product. It is the business model: customers pay for decisions, and every customer makes the decisions better.
  2. The leverage loop. AI absorbs work that previously required hiring, so growth stops pulling headcount along with it. Revenue scales against compute, which gets cheaper on a curve, instead of against salaries, which get more expensive on one. I have written a whole essay on what this does to org design in Blueprints for the 10-Person Unicorn, so I will not repeat it here — the point for this essay is that the tiny team is a consequence, not the cause. The cause is a business model in which the cost side simply no longer moves with the revenue side.
  3. The outcome loop. When AI does the work, you can price the result instead of the tool. The industrial era priced units. The SaaS era priced seats — access to a tool, with the customer still supplying the labor. The exponential era increasingly prices outcomes: resolved tickets, qualified meetings, completed workflows. This is the most underrated shift of the three, because it changes who captures the productivity gain. A seat-priced product hands most of the AI dividend to the customer. An outcome-priced product splits it.
A caution on the word "AI" here: sprinkling a model into your product does not make your business exponential, any more than adding a website in 1999 made a company an internet business. The test is whether one of these loops is actually turning — whether usage measurably compounds into advantage. Deciding where AI genuinely earns its keep inside an existing company is a different question, one I address in AI Strategy for CEOs; this essay is about the model-level question of what business you are actually in.

How Do You Tell a Linear Business in Exponential Clothing?

Most companies that talk like the third era still make money like the first. The history books' favorite cautionary tale is Kodak, which famously developed early digital camera technology and still went bankrupt — not because it missed the technology, but because its business model captured value through film, and no internal invention was allowed to threaten the profit pool. The lesson founders usually draw — "don't ignore new technology" — is the wrong one. Kodak didn't ignore it. The right lesson is that a business model can defeat a technology roadmap, and it usually does. I say this as someone whose own company sat on the wrong side of one of these transitions. UnifyCX, the outsourcing business I co-founded and helped scale past 6,000 employees, was built on the classic 20th-century capture mechanism: labor arbitrage, selling human hours at a margin. AI is ending that era, and we had to rewire the model from the inside — AI agents absorbing the routine tier of work, humans moving up to the ambiguous and high-stakes interactions, pricing shifting from seat counts toward outcomes. I wrote about that transition in detail in What Scaling a 6,000-Person Outsourcing Company Taught Me. The relevant point here: the hard part was never the technology. It was accepting that the thing we billed for had to change. So here is the diagnostic I use, whether I am looking at a portfolio company or a pitch. Five questions:
  1. Does the marginal customer make the product better, or just the revenue bigger? If nothing compounds — no data loop, no network, no accumulating advantage — you have a linear business with good margins, which is fine, but price and plan accordingly.
  2. What does growth pull along with it? If every new revenue milestone has a hiring plan attached, the cost side is still industrial no matter what the product is made of.
  3. What are you actually billing for? Hours and seats are proxies for value. Every era's transition has been a repricing from proxy to something closer to the value itself. If AI cuts the customer's effort by 80% and you charge by the seat, your revenue just volunteered for the cut.
  4. Where does the moat accumulate? Assets depreciate. Data, distribution, and trust can appreciate. Ask what gets more valuable each month the company operates — if the answer is "nothing, we just have more customers," the moat is a queue, not a wall.
  5. Would your current profit pool veto your best idea? This is the Kodak question, and it is uncomfortable on purpose. If the honest answer is yes, the risk to your company is not a competitor. It is your own P&L.
This diagnostic is also, not coincidentally, how operating experience changes investing. When you have personally owned a P&L through one of these transitions, you read a pitch deck differently — a lens I unpack in The Operator-Investor Advantage.

What Should Founders Take from the Shift?

The compressed history — assets to software to loops — is interesting. What you do with it on Monday matters more. Five takeaways:
  1. Choose your business model as deliberately as your product. Write down, in one sentence each, how you create value, how you deliver it, and how you capture it — then ask which decade each sentence belongs to. Most founders discover their product is from 2026 and their capture mechanism is from 2006.
  2. Design at least one compounding loop from day one. Data, network, or leverage — pick the one your market allows and instrument it. A loop bolted on at scale almost never turns; the data schema, the pricing, and the customer promise all have to be shaped around it early.
  3. Price the outcome, or at least start the migration. You do not have to leap from seats to outcomes overnight; hybrid structures (platform fee plus usage, base plus success fee) let you learn what customers will accept. But if your pricing model assumes the customer supplies the labor, its clock is running.
  4. Let the model set the org, not the reverse. Headcount plans, fundraising size, even geography are consequences of the business model. The reason we can build companies on 40 to 60 percent less capital inside the venture studio is not frugality as a virtue — it is that exponential models simply need less input per unit of progress, and raising like a linear company forces you to spend like one.
  5. Assume the next transition is already underway. The gap between eras is shrinking — roughly eighty years of industrial dominance, thirty of packaged software, fifteen of SaaS. Whatever loop you build, hold it loosely. The founders who survived each previous shift were not the ones with the best version of the old model. They were the ones willing to cannibalize it first.
The factory owners of 1926 were not fools, and neither are the seat-based SaaS operators of 2026. Each built the optimal machine for their era's constraints. The only durable mistake, across a century of business model evolution, has been assuming your era is the last one.

Related Reading

Rethinking How Your Company Makes Money?

If you're a founder auditing your own business model against this shift:
Scalable Ventures logo
Flagship VentureFounder & CEO of Scalable Ventures — The AI-focused studio where I build, fund, and scale capital-efficient B2B SaaS.
Visit Scalable Ventures

Related Articles

Explore more insights on entrepreneurship, AI, and leadership:

Explore More

Dive deeper into related topics and resources:
On this page