Picture a restaurant that charges five dollars for a meal that costs twelve dollars to make. Every customer who walks through the door makes the problem worse. The owner knows this. The investors know this. And yet the line stretches around the block, because the owner is not actually in the restaurant business. They’re in the real estate business, and every loss-making meal is a down payment on owning the block.
This is roughly how modern tech companies think about money, and most coverage of their financials misses it entirely.
The Actual Goal Is Not Revenue. It Is Position.
When Uber was burning through cash in 2016 and 2017, losing billions annually, the instinct was to frame this as a company that hadn’t figured out its unit economics yet. That framing was wrong. Uber understood its unit economics precisely. The losses were a deliberate subsidy to riders and drivers, designed to collapse the existing taxi market and crowd out competitors before either group had time to develop loyalty to an alternative.
The goal was never to make money on individual rides. The goal was to become so embedded in urban transportation infrastructure that regulators, cities, and consumers would have no practical alternative. Revenue was a secondary concern to market position.
This pattern recurs so reliably in tech that it has a name: blitzscaling. Reid Hoffman popularized the term, though the strategy predates the label. The idea is that in winner-take-most markets (and tech produces a lot of those), the cost of losing to a competitor who achieves scale first is larger than the cost of the capital burned getting there. So you spend aggressively, absorb losses, and race.
Why This Works in Tech and Almost Nowhere Else
Try this in manufacturing and you go bankrupt. Build cars below cost and you run out of money before you lock in enough customers to matter. The economics don’t work because physical goods have marginal costs that don’t compress.
Software is different. The marginal cost of serving an additional user on a platform approaches zero. Once you’ve built Spotify’s infrastructure and licensed the music catalog, streaming one more song costs almost nothing. This means that if you can acquire users cheaply enough (or subsidize them aggressively enough), and if you can retain them, the eventual per-user economics look attractive. You just have to survive until you get there.
The key word is “retain.” The model only works if users don’t leave. This is why the loss-making phase almost always coincides with aggressive investment in switching costs, network effects, and lock-in. Amazon spent years as a notoriously unprofitable company, but it was using that period to build Prime, AWS, and a fulfillment network that would be nearly impossible for a competitor to replicate. The losses funded a moat.
The Accounting Tells a Deliberately Incomplete Story
Here’s what gets missed in most earnings coverage: GAAP accounting treats customer acquisition cost as an expense in the period it’s incurred. You spend a hundred dollars acquiring a customer in Q1, and that shows up as a Q1 loss. But if that customer generates twenty dollars of margin per year for the next decade, you’ve made a two-hundred-dollar investment that looks like a hundred-dollar loss on paper.
This isn’t a flaw in the accounting. GAAP is conservative by design, because investors need protection from companies that project rosy futures that don’t materialize. But it means that a company aggressively acquiring high-lifetime-value customers will always look worse on paper than a company that isn’t growing at all.
Savvier investors look at cohort analysis instead: how much did we spend to acquire customers in 2019, and what has that cohort actually generated since? This is the metric that matters, and it’s deliberately not on the income statement. Companies that are genuinely building something valuable will show improving cohort payback periods. Companies that are burning cash without the corresponding lifetime value improvement are just losing money.
The trouble is that most retail investors, and frankly many journalists, can’t tell the difference from the headline numbers.
When the Strategy Is Legitimate and When It Is a Story
Not every loss-making tech company is executing a disciplined land-grab. Some are just poorly run. Distinguishing between the two requires asking a specific question: are unit economics improving as the company scales, or staying flat or deteriorating?
If a company is losing fifty dollars per customer when it has a million users, and thirty dollars per customer when it has ten million users, that’s a legitimate scaling story. Fixed costs are spreading, the model is working, profitability is visible from where you’re standing. If a company is losing fifty dollars per customer at every stage of growth, it has a different problem: its business model doesn’t work, and scale won’t fix it.
WeWork is the cautionary example everyone reaches for, and it’s the right one. WeWork’s losses grew in proportion to its revenue, with no evidence that operating leverage was improving. The company was not building a platform with compressing marginal costs. It was signing long-term commercial leases and subletting them short-term, a business with real and persistent fixed costs on both sides of the transaction. No amount of Silicon Valley framing changes that arithmetic. The startups that skipped venture capital early often avoided exactly this trap, because without unlimited outside capital, you’re forced to prove unit economics before you can scale.
The Investor Role Is More Deliberate Than It Looks
Venture capital and growth equity investors aren’t passive bystanders confused by the losses. They’re the ones financing the strategy, and they understand the bet precisely.
The structure is: find a market where winner-take-most dynamics are plausible, back a company to subsidize its way to dominance, accept years of reported losses, then exit either through an IPO or acquisition at a valuation that prices in the eventual monopoly rents. The losses aren’t a bug in this model. They’re the mechanism.
This is why the same investors who would never tolerate a losing manufacturing company will cheerfully fund a social platform burning cash. The terminal value calculation is different. A dominant social platform with strong network effects can price however it wants once the competition is gone. A manufacturing company in a competitive market cannot.
The risk, which materialized for a lot of companies in 2021 and 2022, is that the winner-take-most outcome never arrives. Interest rates rise, the cheap capital dries up, and you’re left with a company that has been burning cash for a decade without ever achieving the dominance that justified the burn. A lot of “tech” companies turned out to be conventional service businesses that had been dressed up in software framing to access software-company valuations.
What Amazon Actually Proved
Amazon is the loss-making-company story that everyone cites, and it’s worth getting the specifics right. Amazon was not uniformly unprofitable for twenty years. It was consistently reinvesting any profit it generated into new businesses, choosing to show near-zero earnings rather than pocket cash flows it could deploy productively.
This is subtly different from a company that genuinely cannot generate a profit. Amazon’s retail business became profitable on a unit basis fairly early. The company chose to build AWS, Prime, fulfillment infrastructure, and a dozen other things instead of returning that money to shareholders. Jeff Bezos was explicit about this in shareholder letters going back to the late nineties: free cash flow per share was the metric that mattered, not reported earnings.
The lesson isn’t that losses are fine indefinitely. The lesson is that mature, functioning unit economics, combined with aggressive reinvestment in defensible new businesses, can look identical to structural losses from the outside. Most companies that invoke Amazon as justification for their burn rate do not have Amazon’s underlying unit economics.
What This Means
If you’re trying to evaluate whether a loss-making tech company is building something real or running an elaborate story, the framework is simpler than the financials suggest.
First: are the per-unit economics improving as scale increases? If not, growth makes the problem worse, not better.
Second: is the company building something with genuine switching costs, or does a competitor’s discount threaten to unwind everything they’ve built? Subsidized growth that produces no loyalty is just expensive customer rental.
Third: is there a credible path to a market structure where the company can eventually charge prices that reflect the value it delivers? Monopoly economics require either a network effect, a switching cost, or a proprietary asset. Companies without one of those things can’t justify the patience the strategy demands.
The years of losses are a bet that one of those three things will crystallize before the money runs out. When the bet is right, it looks like genius. When it isn’t, it looks like WeWork.
