In a market economy there are only three foundational ways to earn money:

1. Rate. Charge for your time and/or effort.
2. Rent. Own something, and charge others for access to it.
3. Risk. Get paid to absorb a risk someone else doesn’t want.

That’s the whole list. Everything else is a blend, a special case, or a disguise.

This isn’t an original taxonomy so much as a compressed one. Classical economics names four factors of production: land, labor, capital, and entrepreneurship. The first three map directly. Land and capital are both “own a thing and charge for it,” so they share the rent bucket. Labor is rate. Entrepreneurship is the odd one out. I haven’t given it a bucket because it isn’t itself a source of income. It’s an activity that changes which bucket you’re earning from, and how much. I’ll come back to it.

Risk gets a bucket of its own because absorbing uncertainty for a fee is a real and distinct business, and it’s neither labor nor ownership.

One scoping note. I’m talking about durable, market-driven revenue. Gifts, tips, and patronage are excluded, not because they don’t exist but because they aren’t market exchanges in the relevant sense. And when they do become durable (a subscription to a creator, a standing tip for a regular service) they reveal themselves as one of the canonical revenue sources: payment for continued output, or payment for continued access.

First, some definitions

Rate doesn’t mean strictly “billed by the hour.” A salaried employee is paid a rate. So is a freelancer who charges by the project, or a consultant whose name commands a premium. The test is not whether effort is involved at all. It’s whether another dollar of revenue normally requires another corresponding unit of your effort. If you go on holiday and the money goes on holiday with you, you’re earning a rate, whatever the invoice says. Reputation doesn’t change the category; it just raises the effective hourly fee.

Rent is any owned asset that yields a return without requiring a new unit of your effort for each unit of revenue. Buildings and bonds are the obvious cases, but the category has to include intangibles or it’s useless: patents, copyrights, brands, audiences, datasets, network positions. All of these are owned, all of them can be licensed or sold access to, and all of them keep paying while you sleep.

The objection I hear most is that intangible assets were created by effort, so surely they belong under rate. But so was the building. A landlord who poured the foundation himself still earns rent when the tenants pay, not wages. The construction was labor; the edifice is capital; the cash flow is a return on the capital. A patent, a song, and a codebase work exactly the same way.

Risk is the narrowest of the three categories, and I’ll spend the least time on it. It is income earned specifically because you accept variability, uncertainty, or downside that another party wants transferred or financed. Insurers, underwriters, warranty providers, and market-makers are paid because someone else prefers a known cost to an unknown one. Investors fit less neatly: they buy a risky residual claim, and their expected return compensates them in part for bearing uncertainty. It’s tempting to file entrepreneurship here too, since founders are famously exposed to failure. I don’t, and I’ll explain why in a minute.

A few quick tests:

• A patent royalty is rent, plain and simple.
• A personal brand is rent when it’s licensed and a rate premium when it’s attached to the person’s own labor.
• A consultant’s project fee is still rate, even when it’s quoted as a lump sum.

Rate, rent, and risk are the pillars the rest of the essay rests on. They’re important, and dull, so let’s move on.

Entrepreneurship is a step function, not a category

You’ll notice entrepreneurship didn’t get its own bucket. That’s deliberate. Founders don’t invent a fourth way to make money.

Nor, despite the mythology, are they primarily in the risk business. The vast majority of venture-backable entrepreneurship has nothing to do with absorbing risk for a fee. Investors finance the uncertainty by buying equity; founders use that financing to do something else entirely. That comes in two parts: finding a step function, and re-centering where the margin sits.

The founder bets that some technological or process change produces a discontinuity—a step increase in the value a unit of effort can produce, or a step decrease in what it costs to produce a unit of value.

Both step directions matter. The value step is the one we often celebrate—an engineer-hour that used to build one invoice now builds a tool that processes millions. The cost step is quieter but just as powerful—a task that used to take a team now takes a person, or a script, or nobody.

The step is where the first half of the founder’s margin comes from. Everything downstream (the pricing, the moat, the eventual conversion of labor into assets) is about leveraging and defending a step that already exists. This makes the contrast case quite instructive. Being 10% better than the incumbent is not a robust startup strategy. It’s effectively setting up a rate business with worse economics than the incumbent’s, because the incumbent already has the customers and the scale. Only a step creates durable space for a new entrant to stand.

The Startup Loop: How disruptors build margins

Here’s a pattern successful companies often follow, whether or not anyone inside would describe it this way.

It starts with effort, and someone has to pay for it. A new company has no asset yet—nobody rents a platform that doesn’t exist. What it has is labor, and there are two ways to fund it. The first is to sell the labor directly to customers (consulting, custom work, implementations) and build the product on the side, or out of the residue of the client work. Revenue is rate from day one. The second is to sell equity to investors, who finance the build by taking a risky claim on its future value, and to charge rent-like fees from the first customer onward. Plenty of companies take the second route and never send a services invoice in their lives. The two approaches look different on the income statement, but they’re the same move: convert labor into an asset before the money runs out. The difference is who finances the uncertainty in the meantime.

Look at the marginal effort behind a young company’s revenue and the answer is much the same whichever route it took. Stop the effort and the consulting clients leave; early subscribers churn to whoever is still shipping. Early-stage revenue is often rate-like in substance, whatever the invoice calls it. The company’s goal is to make that revenue progressively less dependent on each new unit of effort.

Success is the conversion of that labor into a rentable asset. A codebase. A platform. A brand. A network. A dataset. A regulatory position. Something that keeps paying after the effort that built it has stopped.

Two textbook cases. Software is the one everyone knows, though I’ll complicate it in a moment. The purer example is franchising. An operator runs a business that earns rate (flip the burgers, serve the tables) and bears the operating risk of doing so. Franchising codifies that operation into an asset (a brand, a playbook, a supply chain, a site-selection system) and licenses it. The franchisor keeps the rent. Labor and operating risk go to the franchisee. Franchising is not a business model that migrates from rate to rent. It is the packaging up of the migration itself.

Installed, perpetually licensed software was always rent-like. The vendor spent capital to build an asset and charged per-seat or per-organization for access to it. The income was just lumpy—a large license fee up front, maintenance on top, and an optional upgrade every few years that the customer could decline. SaaS didn’t convert rate into rent, as the folk history has it. It converted lumpy, low rent-like income into smooth, higher rent-like income. And it did it by taking on an obligation the perpetual-license vendor never had: development continues for the entire span of every contract, because there is no upgrade to sell when the upgrade is the subscription. By the marginal-effort test, SaaS looks at first like a partial reversion toward rate: stop shipping and the customers leave.

SaaS companies therefore spend their maturity pushing that revenue back toward rent-like economics. Vendor lock-in through integrations and data gravity. A slowing development cadence once the product is largely “done.” Brand authority. A value-added reseller network that sells and implements on the vendor’s behalf, and pays for the privilege. The subscription invoice looks identical in year one and year ten. What changes is whether the tap would turn off.

Mature incumbents stack the categories. Look at the invoice a large customer receives from an established vendor. There’s a subscription or a license, priced per seat or per organization (rent). There’s onboarding, implementation, and “professional services” (rate). Often there’s a partner network doing customizations and integrations (rate) and paying the vendor for certification and referrals (rent again). A mature franchisor charges royalties on the brand, leases the buildings to its own franchisees, takes margin on required supplies, and bills for training. From the customer’s side it reads as one product. From the vendor’s side it’s extraction from two or three categories at once, and the excess margin lives in whichever layer the customer can’t easily separate from the rest.

The challenger’s entry move is often to unbundle. A new entrant takes the stacked product and re-offers it in a single category, usually the one the customer would prefer to be paying in. Red Hat is the canonical case: the incumbents charged rent for the operating system, so Red Hat gave the operating system away and charged a rate for support. The transition can run the other way. Tax software (rent) did the opposite to accountants (rate), and website builders (rent) did it to design agencies (rate). The incumbent sold rate, the challenger productized it into rent, and customers preferred the predictable fee to the hourly one. The second half of the entrepreneur’s arbitrage is between categories. You attack whichever layer carries the incumbent’s margin by offering the product in the other one.

These are different moves. The step function changes the economics of the activity itself. Unbundling changes who pays for what, and in which category. The strongest disruptors pull both levers at once, since the step is what lets the unbundled offer come in at a price the incumbent can’t match. But a founder with a step and no unbundling still has a business. A founder with unbundling and no step is running a price war against an incumbent with layers of natural advantages.

Once the challenger is entrenched, it starts rebuilding the revenue stream stack it originally stripped out. The support company adds a proprietary console. The subscription company adds pricing tiers, implementation packages, and a certified-partner program. Margin re-builds. And eventually someone unbundles them, and the cycle starts over. This is why incumbents fight unbundling so ferociously. They’re not defending a product. They’re defending the right to extract margin in two categories simultaneously.

The loop assumes the challenger has time—time to entrench, rebuild the stack, and enjoy the margin before the next challenger arrives.

That conversion isn’t equally available to everyone. Turning effort into durable rent usually requires capital to build the asset, or a durable enough market position that switching away from you is painful, or enough influence to shape the rules so that your activity becomes structurally rentable (patents, licensing regimes, professional accreditation, zoning). Those are exactly the things incumbents have and challengers don’t. Incumbency compounds, in part, because incumbents can convert their labor into asset income and then pull the ladder up behind them.

Technology re-commoditizes rent

Rent depends on one thing: a durable barrier to replication. If anyone can build a copy of your asset cheaply, nobody will pay you a premium for access to yours. So rent is only as durable as the cost of replication is high.

And technology keeps knocking that cost down. Every infrastructure wave collapses the replication cost of some existing class of asset. Cloud infrastructure eroded the moat of “we own the servers and can afford the sysadmins.” Open-source frameworks and libraries eroded the moat of “we wrote the hard plumbing.” AI code generation is now eroding the moat of “we wrote the application logic.” Each wave takes something that used to require real capital or scarce skill to replicate and makes it radically easier to copy.

When replication cost falls faster than creation cost, an odd thing happens to the incumbent’s pricing. The model still looks like rent. There’s a subscription, a per-seat license, an annual contract. But the economics slide back toward a rate market, because a competitor—or increasingly, a customer—can reproduce much of the visible product for the cost of a few engineer-hours. The subscription becomes a convenience fee. The income has quietly become rate-like: fresh labor can now substitute for access to your asset.

This slippage isn’t uniform. Re-commoditization hits hardest where the code was the asset: horizontal tools, generic dashboards, thin wrappers around someone else’s API. It’s far weaker against moats that were never about code. Network effects (a marketplace’s two-sided liquidity), proprietary data (a dataset competitors can’t assemble even with identical software), workflow lock-in (the years of muscle memory embedded in a customer’s daily habits), and regulatory capture all survive codegen more or less intact. Anyone can clone a chat application’s interface in an afternoon. Nobody can clone its message history, its integrations, or the fact that ten thousand companies have built their workday around it.

Franchising is the instructive contrast to software rent erosion. Its asset is brand and physical footprint, not code, and it has largely shrugged off every one of these waves. Nobody has yet written a framework that generates a thousand well-located restaurants.

AI codegen is itself a step function of the kind described above—a discontinuous drop in the cost of producing software. It is also an unbundling engine, because it lets customers convert a vendor’s rent into their own internal rate by building the thing themselves. It pulls both levers at once, from the customer’s side. That’s why it’s so much more destabilizing than cloud or frameworks were.

Double compression breaks the loop

Now remove the assumption that a challenger has time to entrench before the next one arrives.

The re-commoditization wave does two distinct things to a software company. First, it shrinks the window for converting effort into durable rent. The old sequence was: build the asset, accumulate customers, entrench the switching costs, slow the development cadence, and let the subscription harden into rent. That took years, and the later years were the point because they produced the highest margins. Now a competitor can approximate the visible surface of your product before you’ve finished entrenching the parts that were never really code. The time required to produce a credible lookalike is increasingly measured in weeks, often before you’ve entrenched the parts that can’t be copied so quickly.

Re-commoditization also drops the floor. The same tools that let competitors clone your asset also let more people produce equivalent coding and design output faster and cheaper. The fallback that software companies have always relied on—retreating from product margins to services margins when the moat fails—is now worth far less than it used to be, because the rate you could charge for the underlying labor has fallen too.

Put those together and pure SaaS loses much of the soft landing it once had. Rents don’t fall to zero overnight. That’s ordinary competition and companies survive it. What’s new is that the services floor they’d fall to has also dropped, and it dropped for the same reason and at the same time.

For some companies, the startup maturity loop stops turning. The loop required time for a company to entrench before being challenged in turn. When time-to-disruption falls below time-to-entrenchment, the rebundling phase may never arrive. A company that entered the market by unbundling an incumbent can get stuck in that unbundled, single-category, rate-like state. The partial reversion toward rate-based fees that every SaaS company starts with, and spends its maturity reversing, becomes the end state instead of the starting point. It looks like a software business. It bills like a software business. But in economic terms it’s a services firm with a subscription form on the website.

This produces a cohort effect that I think is under-appreciated. Companies that finished the conversion to durable rent before the window closed (the pre-codegen incumbents with genuine data, network, and workflow moats) get to keep their rebundled stack. Everyone who was mid-conversion when the window slammed shut does not. So a wave that looks like it’s disrupting incumbents is, for one specific vintage of incumbent, doing the opposite: entrenching them while denying the same path to their would-be successors. The ladder hasn’t been pulled up. The building it leaned against has been demolished.

If you want an empirical hook, look at the repricing of public software companies in 2026, which the trade press has taken to calling “SaaS-mageddon.” Read it as a regime change rather than a cyclical dip. The market is moving an entire asset class from the rent column to the rate column. But let’s not tie this thread too tightly to one year’s headlines. The mechanism is the point, and the mechanism will still be there when the headlines have moved on.

Rent doesn’t die, it moves

Two implications follow.

The first is genuinely good news. Re-commoditization is a recurring, structural, anti-incumbency force. Each wave drags some quantity of rent back toward rate-like economics, strips incumbents of margin they were extracting for having been early rather than for being good, and hands the savings to customers. It does this without requiring regulators to act or executives to be virtuous. It’s one of the few reliable counterweights to the compounding advantage of incumbency.

The second is to update where you expect future margins to come from. The rent doesn’t necessarily evaporate. Much of it flows downhill. When application-layer rents compress, some of the margin reappears at whatever layer still has a high replication cost. Right now that’s compute, foundation models, and distribution. The loop from earlier predicts what the owners of those layers will do: stack categories, charge in two or three columns at once, and defend the stack ferociously. And the cohort effect suggests their window may be long, because nobody has yet found the wave that re-commoditizes them.

Put the two together and the pattern is clearer than either half. A re-commoditization wave often doesn’t eliminate rent so much as move it, and you can watch it move by following which categories the revenue is flowing between.

At the layer the wave hits, revenue flows from rent toward rate-like economics. The incumbents’ subscriptions decay into convenience fees, and the margin they were collecting for ownership goes back to whoever is willing to do the work. That’s anti-incumbent, and it’s real. But the same wave can move revenue from rate toward rent one layer down. The tools that made the application cheap to replicate are themselves owned, metered, and hard to replicate, so some of each engineer-hour saved at the top reappears as a compute-hour, an API call, or a distribution fee paid to the substrate. That’s pro-incumbent, just for a different incumbent. Same wave, opposite effects, depending on which side of the flow you’re standing on.

That gives the framework its actual use. It can’t tell you where rent will be in ten years, but it can tell you how to find it now. Follow the revenue as it changes categories, and look for the layer where rate is turning into rent rather than the other way around. That’s where the window of opportunity for margins is longest, that’s where the stacking is happening, and that’s where the next unbundling will have to come from.