Felix Salmon had a question on Twitter:

While Salmon’s question was driven by the news that AT&T would spin out WarnerMedia and merge it with Discovery, under Discovery’s management, it was also prescient given a report in The Information a few hours later that Amazon is considering buying MGM.

It remains to be seen if Amazon will actually follow through with a purchase, and even longer to know whether or not it was the right play; there are reasonable arguments both for and against. What has always been clear, though, is that AT&T’s foray into media was a bad idea that made zero strategic sense.

Being open to Amazon’s purchase of a studio, and opposed to AT&T’s acquisition of a media conglomerate (which I was from the beginning) isn’t based on some sort of bias against old-line firms versus tech firms; rather, it stems from an understanding that the Internet really is different from the analog world.

AT&T and the Economics of Distribution

Distribution is one of the most frequently used words in both tech and media, and in the vast majority of cases it is used wrongly; explaining why the marriage of AT&T and Time Warner was destined for divorce is not one of them. Take it from a legend of the industry, in the Wall Street Journal:

Cable mogul John Malone, a major Discovery shareholder, said that although he believes Time Warner is doing fine, merging content and distribution usually doesn’t make sense. “I think that the technology of connectivity and digital technologies are one focus, and creating content that people get addicted to is another focus,” he said. “And you seldom would find both of those in the same management team.”

Look no further than Jeff Bezos and Amazon to see why Hollywood can be a distraction, but in my estimation the problem of merging content and distribution are much more fundamental.

Start with the latter: distribution, properly understood, entails the build-out of physical infrastructure. Comcast needs to lay cable and fiber; DirecTV, another AT&T misadventure, needs to launch satellites and install dishes; AT&T’s core mobile business requires buying spectrum and installing cell phone towers and base stations. These are not businesses for the faint of heart, given the massive capital costs.

The payoff for this investment, though, is a competitive moat. Comcast, for example, may only compete with satellite for TV service, and the local phone company for Internet access; no other cable company is going to lay competitive cable lines, because Comcast could simply lower prices and drive them out of the market. This isn’t even predatory pricing, because the actual utilization of its infrastructure is effectively free on a marginal basis. Laying cable is expensive, but using it is cheap.

That said, to the extent competition exists, it is quite brutal because it is zero sum; an AT&T customer is not Verizon customer is not a T-Mobile customer. The combination of scarcity in customers and zero marginal cost for incremental usage can lead to price wars, which T-Mobile used over the last decade to take a good amount of share from Verizon and AT&T; perhaps that is why both tried their hand at content.

Differentiation and Commoditization in Content

Verizon’s approach was to get into the commoditized content space, acquiring and eventually merging AOL and Yahoo into a company bizarrely named “Oath”. The problem for Verizon is that Oath had the opposite problem of its core business: digital content requires effectively zero fixed costs to get started, but never-ending marginal costs to produce, which means it had effectively infinite competition for both eyeballs and advertising dollars, and no differentiation in either market.

AT&T, on the other hand, acquired highly differentiated content with its acquisition of Time Warner (which it renamed WarnerMedia). The problem for AT&T is that differentiated content has a business model that is orthogonal to AT&T’s core business. Whereas AT&T competes for customers in a zero sum game, content is best leveraged by reaching as many customers across as many distributors as possible. That means that what would have been best for AT&T’s core business — being the exclusive way to get access to WarnerMedia content, thus giving a reason for customers from Verizon or T-Mobile to switch carriers — would have been value destructive to WarnerMedia, because the cost of producing its differentiated content would have been amortized across fewer customers.

AT&T instead went in the opposite direction: by creating HBO Max, WarnerMedia stopped selling content to the highest bidder and instead started bidding for content itself, which was the worst of all possible worlds, at least in the short run. HBO Max content was limited in reach — it was only available to HBO Max subscribers — which meant that its content was leveraged against an even smaller base than, say, AT&T’s customers. The potential payoff, of course, was a service like Netflix, which reaches everyone everywhere, no matter their carrier or cable provider (which, of course, raises the question as to what strategic benefit would have accrued to AT&T’s core business even in the best case scenario).

This is where it is important to be precise about the meaning of “distribution.” For Netflix, distribution is the Internet, which is to say it is completely commoditized. You can watch Netflix on your phone, on your TV, on your console, on your computer, on your set-top box, basically through any device that has an Internet connection. This distribution isn’t technically free — Netflix has huge bandwidth bills, and if you watch enough of it you may hit bandwidth caps — but from a strategic perspective it might as well be.

In fact, everything that is distributed via the Internet is effectively free. Netflix, Oath, HBO Max, Google, Facebook, Wikipedia, Spotify, YouTube, Stratechery — all are effectively free to access for anyone from anywhere. This is why I get so confused when companies or regulators complain about Google or Facebook controlling distribution; neither company controls the cables or routers or switches that deliver content. They have zero control of distribution. Rather, what those companies control is demand.

Distribution Versus Demand

What are the ways one might read this Article? My subscribers are reading an email, or perhaps an RSS feed, or listening to a podcast. New readers may have had this Article forwarded to them from a friend, followed a link on social media, or searched for an article distinguishing between “Distribution and Demand.”

The sheer scale of the Internet is such that the last two channels have many orders of magnitude more potential readers than the ones I directly control. After all, people use social media to connect to basically everyone on the planet, and post basically anything; a link to my site is one link among an infinite number of other links. It’s the same thing with search: everyone uses search to find information about anything you might think of; the scale of the Internet — thanks in part to the never-ending content creation that characterizes businesses like Oath — is such that there are almost always scads of results for your query.

What makes Google and Facebook so successful is that they are the linch pin upon which these massive markets pivot, in large part because both services increase in functionality with scale: more sites and more links mean better results in Google, which drives increased usage, which provides a feedback function allowing Google to refine its results; similarly, more people and more content mean stronger networks and more engagement on Facebook, which drives increased usage, which provides a feedback function improving the recommendation algorithm.

Both services built monetization engines that perfectly align with and benefit from their core feedback function: Google gives you search ads that are more relevant with more advertisers, and which improve in relevance with more customer clicks; Facebook shows you ads that you are interested in because they are similar to the content you already enjoy, and like Google, more ads mean better ads for the marginal user, which improve in relevance with more customer clicks.

The problem for a company like Oath is that its content was simply grist for Google and Facebook, and to the extent it earned page views, the accompanying ad inventory wasn’t nearly as differentiated as ads sold by Google or Facebook. Oath had distribution — anyone anywhere could access its sites and apps for free — the problem is that it didn’t have demand, from either users or advertisers.

HBO Max, meanwhile, did have demand for its content, but because that content required a subscription to HBO Max, it placed an artificial cap on its distribution to whoever was willing to subscribe. This can work — again, look at Netflix — but it takes billions of dollars, both in actual spending and, just as importantly, in foregone content sales, to pull it off. That may make sense for a startup, but its nigh-on suicidal for a (real) distribution company like AT&T.

Amazon Prime

There is another important difference between a company like Oath and a company like Google: while the former has minimal fixed costs and ongoing marginal costs, the latter has massive fixed costs and minimal marginal costs. Google spends a massive amount of money on both R&D and in capital costs to dot its servers all over the world, and link them together with an expansive private network; Google leverages these costs across the virtuous cycle of users, content suppliers, and advertisers.

Another company that expends massive amount of money on R&D and capital costs is Amazon;1 the company spends billions of dollars on everything from distribution centers to delivery drivers, paying for it with an ever-expanding virtuous cycle of suppliers, customers, and, increasingly, advertisers. Because these costs are fixed Amazon wants to both expand its customer base and also increase its share of wallet with its existing customers; one of the most effective ways to do this is convince customers to subscribe to Prime, which, among other benefits, offers two-day (increasingly one-day) shipping. This encourages shoppers to start their searches for products on Amazon, where they may click on sponsored search results and ultimately buy on Amazon, without even considering alternatives.

This is not, to be clear, because Amazon controls distribution. Other e-commerce sites are, as they say, “only a click away” — Google in particular is eager to direct shoppers to alternatives, which are easily accessed from your phone or computer. Rather, Amazon’s goal is to control demand: its best customers go to Amazon because Amazon has what they need and gets it to them quickly, and because they already paid for shipping with their Prime subscription.

Notice how much different this kind of competition is from that engaged in between true distribution companies: AT&T and Verizon spend a lot of up-front, but the payoff is physical lock-in; Google and Amazon, meanwhile, spend massive amounts of money on fixed costs, but they don’t have any lock-in at all: they win not by limiting customer choice, but by being the top choice in a world where alternatives are easily accessible. At the same time, while a dependency on physical infrastructure limits AT&T and Verizon’s scale, companies like Google, Amazon, and Facebook have free distribution — remember, they’re on the Internet! — which means they can serve anybody.

This is why Amazon’s long-running investment in Prime Video, and potential acquisition of MGM, makes far more strategic sense than AT&T’s gambit ever did. Amazon has to work to win and retain customers on a continual basis, ideally to its Prime subscription service, and bundling differentiated content is a great way to do that. Moreover, Amazon isn’t trying to build a subscription service from scratch, thus drastically limiting its leverage on that content; Prime has 147 million members in the United States, and 200 million worldwide, and none of those customers are making a zero-sum choice between Prime and, say, Netflix. After all, distribution is free.

Aggregators and Platforms

You can probably guess my vote in Salmon’s poll:

I do get the argument that Prime Video is a waste of money for Amazon; Brad Stone notes in his new book Amazon Unbound that “there was little evidence of a connection between viewing and purchasing behavior” and that “any correlation was also obfuscated by the fact that Prime was growing rapidly on its own.” I have no beef with people who voted “No” on both — what is important is that we all agree that AT&T’s approach didn’t make any sense at all.

This is why getting definitions right is so important: if you conflate controlling distribution with controlling demand, you are liable to waste billions of dollars on acquisitions that make no sense, or, in the case of regulators, spend years pursuing court actions against companies like Google that result in zero change in the relevant markets. I wrote in a 2019 article:

There is a Sisyphean aspect to regulating power predicated on consumer choice: look no further than the European Union, where regulators are frustrated that remedies for the Google shopping case aren’t working, even though those same regulators were happy with the remedies in theory; the problem was trying to regulate consumer choice in the first place.

The premise of that Article is that effective regulation meant distinguishing between Aggregators and Platforms; drawing the distinction between delivery and demand is making the same point in a different way. Aggregators win by consolidating demand; platforms exert dominance by controlling distribution. Search results and social networks are the former; App Stores and access-control are the latter. Not understanding the difference is, as both AT&T and Verizon have learned, exceptionally costly.

1. I am, for purposes of this article, going to focus on Amazon.com, not AWS, although all of the same principles apply.