How can a company with no revenues still make money? It’s not a trick question. The answer is at the very foundation of the digital economy: advertising.

No matter how dire things get for musicians, writers, movies, websites, smart phone apps, video games, or whole social media platforms, no matter how hard it might be for companies to charge for content, services, or convenience, almost everything we are doing in the digital marketplace can serve as the advertisement for something else. The video game promotes a movie, the movie promotes an app, and the app promotes a video game. Heck, this article indirectly promotes a book.

The trouble is, if everyone is in it for the advertising dollar, who is left to advertise? At no point in history has advertising, marketing, and research ever accounted for as high a percentage of GDP, or total economic activity (and that’s being extremely generous). But right now, it’s pushing at the very top of that range. The reason it can’t go higher is that only so much economic activity can go to promoting the rest of our economic activity. The coming crash in the tech market–and quite possibly beyond–will be triggered by the growing realization that every company in the world can’t be a marketing company.

From the looks of it, this simple math has not trickled up to the leaders in venture capital or the investors in NASDAQ’s tech giants. As they see it, just because a retail business sells products at a loss doesn’t necessarily put it in a worse position–financially speaking–than any other Internet platform, smartphone app, or GPS utility that takes little or no cash from its multitude of users. The goal isn’t to earn revenue, but to sell the whole company to an acquirer or Wall Street . And in the current landscape of pitiful revenue, the fantasy “big data play” has become the holy grail of exit strategies.

So instead of earning revenue, these companies harvest user data. In theory, every user interaction can be sold to market researchers looking for better ways to target audiences, craft messages, categorize psychographics, and predict behaviors. The potential upside of all this data far exceeds the pennies one can collect selling something, or so the logic goes. It makes sense in a world where no one has money to spend online, anyway.

It certainly works, at least in the short run. Musician and producer Jay Z, for instance, figured out early on that he could make more money in data than music. So he gave away an entire album to his fans, only it was in the form of an app, a piece of spyware, that monitored their usage patterns and upsold the associated data to Samsung to do with or sell as they please.

Even on e-commerce sites, in many cases the profitability of retail transactions pales in comparison with that of the big data they leave in their wake. Creating relationships with consumers is really just about engendering enough trust to get them to share their data assets with you.

Producers of everything from culture to products have to be tuned to–if not entirely geared toward–reaching easily identified social audiences. This is not a soft science, like determining the audience for a printed magazine or market for a new style of blue jean. It’s hard data on engagement. Our products–my books, this website, your mutual fund advisement services–will be less valuable as goods and services than as the means through which we accumulate followers on social networks, whom we then sell to brands. So our goods and services better be brand-friendly and our markets pre-selected for their data richness.

Even social media deserves a better role in our lives and businesses than this.

Furthermore, unscrupulous website owners have now learned to use robotic ad-viewing programs to juice their revenue from pay-per-click advertising. Most of these bot programs run secretly on the computers of everyday users in the form of malware, a kind of mini-virus that co-opts a computer’s processing power. Bots now comprise an estimated 25% of all online video ad viewers. That means up to a quarter of all the data out there isn’t even about humans.

In 2015, advertisers lost approximately $6.3 billion in pay-per-click fees to these imaginary viewers. Consider the irony: malware robots watch ads, monitored by automated tracking software that tailors each advertising message to suit the malbots’ automated habits, in a human-free feedback loop of ever-narrowing “personalization.” Nothing of value is created, but billions of dollars are made.

Eventually, social branding has to run out of fodder. As more and more markets lose all revenue potential except what they can make as social media marketing platforms, who is left to buy all this marketing and consumer data? Especially when it’s about robotic ad-clicking algorithms that don’t ever buy anything themselves? Consumer goods like soap and potato chips may have been able to keep mainstream broadcast television alive with advertising, but they cannot support the multibillion-dollar valuations of Silicon Valley and the future of the entire digital economy.

No, instead, companies are going to have to begin to venture into the perilous, experimental, and highly speculative option of charging people for stuff. Think HBO. It’s a radical approach to goods and services in which people pay directly for what you’re offering.

If what you’re doing has any value at all, it deserves to be supported by your customers. If you’re counting on the ads to take up the slack, you may be out of luck.