Complements and Constraints
Startups Investing in Startups, Flywheels, Stripe, Netflix
The Information recently published a pair of articles ($) covering how Stripe and Coinbase have built fairly prolific venture investing operations, despite being venture-backed companies themselves. They note that this is an interesting phenomenon, but not entirely a new one; Google and Salesforce started venture investing at a relatively early stage as well.
This is particularly interesting because these four companies are four of the most valuable companies founded in the last two decades, all worth well in excess of $50 billion. Indeed, if you made a “market cap created per year since founding” table, these four might all be in the top ten.
To the best of my knowledge, it is not particularly common for companies to get into venture investing until they are mature and generating more cash than they know what to do with. These four companies, however, all started venture investing at a time when they were still raising money themselves.
This is surprising because selling equity to venture capitalists is very expensive. The Information reports that Stripe started investing in 2017; at the time, Stripe was reportedly valued at only $9 billion, while today it is valued at $95 billion. Coinbase started investing in 2018, when its most recent round was reportedly at a $1.5 billion valuation; today it is publicly traded and worth $50 billion.
It seems strange for a growing startup to raise dilutive capital only to turn it around and invest in other startups. You would think that high growth companies would focus on their core business, save the dilution, and leave venture investing to the venture capitalists. Yet, this seems to be a common practice at some of the most successful startups! It seems like a pattern worthy of further investigation.
It is worth looking at a story of a successful venture investment to look for plausible explanations. Here is a famous one:
Back in 2007, Netflix was a small public company that rented DVDs by mail on a subscription basis. They had only 7 million DVD-by-mail subscribers at the time, a far cry from the 207 million streaming subscribers they have today.
Netflix’s vision at the time was that video content would eventually all be streamed over the internet as broadband became more prevalent, and it was in 2007 that they launched a streaming service that was free to their existing DVD subscribers.
Video streaming had a number of obstacles at the time, one of which was the difficulty of getting the content to your TV. At the time, you had to watch on your PC, buy a bulky and expensive internet-connected set-top box with a hard drive, or (most commonly) use an internet-connected gaming system such as an XBox. TVs at the time were not connected to the internet.
So it was in 2007 that Netflix launched Project Griffin, to develop a small, inexpensive, easy to use device to stream content from the internet to your TV.
The program progressed successfully, but as the launch neared, Netflix CEO Reed Hastings got cold feet. He realized that getting into the hardware business would put them into competition with other hardware makers whose distribution they would need:
“Imagine you drive a Chevy, and the gas station is branded Ford,” says the top-level source. “Would you feel comfortable filling up your tank there? I think that’s why Reed didn’t think this could be done inside Netflix.”
Hastings decided that they could not take the risk of antagonizing other hardware makers and possibly getting Netflix blocked from popular devices. Netflix instead spun this project into a different company, Roku, invested $6 million into it, and The Netflix Player by Roku was released in May 2008.
The rest is history. Roku was a huge hit, and so was Netflix. Roku maintained their market lead in streaming devices, even as Apple, Amazon and Google later entered the market. Netflix would soon expand their content library and start charging separately for streaming subscriptions, and they would grow rapidly as well.
Today, Netflix has over 200 million subscribers and is worth $220 billion, and the stock is up over 200-fold from 2007. Roku is still the leader in streaming devices and software and has over 50 million active users, and the company is now publicly traded and worth over $40 billion.
The first observation from this story is that Netflix seemed to be unconcerned about maximizing the direct value of their investment in Roku; one might even say they almost gave it away. They invested $6 million in Roku in 2008 after spinning it off, but their SEC filings suggest they sold their stake for $7.5 million in 2009. (Extrapolating from Roku’s S-1, that stake today if they had held it would be worth $2 billion, or as Netflix calls it, a rounding error.) Menlo Ventures ended up funding many of the early rounds for Roku and would walk away with big investment returns.
This story also demonstrates that Netflix had a clear concept of the total user experience, and a clear understanding of the importance of bottlenecks. Netflix understood from the beginning that they were selling convenience, not just content.
Video delivered over the internet on demand was always eventually going to be more convenient because it eliminated having to make a trip to Blockbuster, or request DVDs by mail, or tuning in to a particular channel at a particular time, but that did not mean there were some serious obstacles to overcome to make the home streaming experience a smooth one.
The consumer only cares about the overall user experience, which is no better than the weakest link of the value chain. A consumer will only subscribe to Netflix if the overall experience is worth paying for, and does not care if the weak link is something under Netflix’s control, such as the quality of content or good UI, or if the weak link is something that is not, such as slow internet or difficult hardware setup.
In this scenario, goods like broadband internet and set-top boxes are often described as economic complements to Netflix’s streaming service, as they are typically consumed together. Classic examples of economic complements are left shoes and right shoes, or cars and gasoline.
Netflix invested in the development of an inexpensive internet-connected set-top box because they believed it would increase the demand for their new streaming service, not necessarily because they believed that Roku would necessarily become a valuable company in its own right. They believed that there was a lucrative future in selling right shoes, but they had to make sure left shoes were readily available too.
This case study also illustrates the importance of focusing on bottlenecks and working backwards from them. There was a popular book in the 1980s, The Goal, which advocates a philosophy known as the Theory of Constraints. The general idea is that any system with dependencies is by definition being bound by one bottleneck at any given time, and available resources should be directed only at alleviating that bottleneck.
As a simple example, imagine you have ten pounds of flour, ten pounds of sugar, ten pounds of chocolate, and ten pounds of butter. If you obtain a few dozen eggs, you will be able to make a lot of brownies; therefore all of your resources should be directed into obtaining eggs, and ignoring the other ingredients until you have enough eggs and the bottleneck moves elsewhere.
This is easy to see in a toy example like this, but in real life, it is common for companies to try to, for example, obtain cheap sugar under the theory that sugar is an ingredient in brownies and cheaper sugar will therefore somehow lower the average cost of brownie production.
In the Netflix example, they identified that the lack of availability of good, cheap, easy to use streaming devices closed off the market for their product in households with no gaming system. As the first major entrant in the market for streaming video, it made sense for them to fund the development of such a device as it would enable them to sell more subscriptions, even if they didn’t make any money on the device itself.
In 2010, Apple released a similar competing $99 device, and by the end of 2010, both Roku and Apple had sold a million units. If we assume that all of these devices resulted in new Netflix subscriptions - probably safe to say since the only other service of note at the time was an early version of Hulu - that’s 2 million incremental subscribers of Netflix at $100 per year, or $200 million of incremental cash flow per year!
This illustrates the power of focusing on complements in a business where you have high incremental market share and high margins on incremental units. Goods such as software and video delivered over the internet are known as “information goods”, which cost almost nothing to produce copies of and can be consumed by infinitely many people at the same time. Such goods by definition have nearly 100% gross margin on incremental sales.
Recall here that Netflix was licensing content at flat rates from the studios and did not yet have much competition, so additional subscribers from the availability of complements truly resulted in more cash to Netflix. The math would not work as well in the current more crowded streaming market, where the benefit of spending to improve a complement would be split between Disney, HBO, Amazon and others. Nor would it work as well in an industry with higher marginal costs, where less cash would drop to the bottom line.
What really makes the logic work is the positive feedback loop that is a common feature of industries involving information goods. In the case of Netflix, more subscribers means more cash that can be invested in new content that will in turn attract more new subscribers. Netflix would soon spend $100 million to produce two seasons of House of Cards, its first original show, which debuted in 2013. This brought in more subscribers and more revenue which was spent on more shows and today, in 2021, Netflix’s content budget for the year is $17 billion.
Netflix has other virtuous cycles. More subscribers also means more word of mouth, more cash to spend on improving its product, and more data to improve its content selection, all of which lead to more subscribers and more cash and yet more subscribers. All of these elements combined to compound Netflix’s lead over other streaming providers very quickly. HBO went from denigrating Netflix as the Albanian army to complaining about Netflix’s market power in just a couple of years. Today, other media companies are fighting to catch up, but it is difficult to compete with a business that is now bringing in nearly $30 billion of subscription revenue each year.
The value of Netflix will be determined by the sustainability of its competitive advantage over its rivals. Its size today gives it considerable scale advantages; it can now spread the fixed costs of content production and product development over a much larger customer base than its peers, which should allow it to be very profitable in the future while still offering a better value proposition than its competitors. That it has been able to grow to this level from nothing in only fifteen years is a testament to the economics of information goods and the power of the internet.
The tendency of feedback loops and economies of scale to magnify and solidify early advantages would seem to justify Netflix’s early focus on nurturing the availability of complements to improve the overall user experience and to attract additional early users. Roku was only ever a small segment of Netflix’s early user base, but its availability early on undoubtedly resulted in faster growth and better user satisfaction than there would have been otherwise, which allowed Netflix to gain an edge before incumbents like Disney and HBO were able to get their act together.
Netflix is just one example that Marc Andreesen cited in 2011 when he said that “software is eating the world”. It is frequently much more efficient and flexible to accomplish a task with software and distribute it over the internet than to use a legacy method like mailing a DVD.
We would therefore expect to see this pattern a lot in new software categories. The current market leader should be interested in any investment that improves the overall value proposition to consumers, even if it does not directly produce attractive returns (although it must be economically sustainable). It will capture most of the benefit from new adopters, and positive feedback loops means this value will compound down the road.
Often this will mean that the market leader will build certain complements themselves, but it is equally viable for them to fund other companies that build complements too. In fact, as we saw in this example, sometimes building a complement internally creates a conflict that can best be resolved by investing in other companies that build the complement, instead.
You may have heard of “commoditizing the complement”, a concept popularized in a 2002 essay by software entrepreneur Joel Spolsky, where he observes that technology companies try to get the price of complements as low as possible to stimulate demand for their product. He cites examples of companies funding the development of open source software, and Microsoft commoditizing PC manufacturers. Gwern expands on this with more examples and cites Hal Varian’s Information Rules, which covered the topic in more depth when it was published in 1999. (The book is highly recommended; Varian demonstrated a deep understanding of the economics of the internet by joining Google as Chief Economist way back in 2002.)
Spolsky’s observation is frequently true, but perhaps incomplete. Price is only one dimension; when you have a system with a lot of dependencies, it is much more important that your complement is available at all. Certainly we have seen one demonstration of this as we come out of the pandemic, where the auto supply chain is snarled because of a shortage of inexpensive chips. Prices signal to other parties what is worth investing in; low prices are not always better.
The Roku example is a good demonstration of this. The market clearly expects Roku to be a very profitable company in the future, and to date Netflix has done nothing additional to try to commoditize set top makers. Sometimes the best strategy is to let your partners make some money so they continue to invest in the ecosystem.
Turning back to the examples cited in the beginning of the essay, Stripe is now one of the leading payment providers for businesses that transact over the internet. We do not have an S-1 from them yet, but in general, payment facilitators such as Stripe are able to capture a take rate on the total volume of payments transacted on their platform.
Beyond payment facilitation, Stripe provides a host of other services that make it easy to start and operate a business online, from Stripe Atlas (business incorporation) to Stripe Treasury (banking-as-a-service). It is likely that these other services are less about generating revenue directly and more about making it easy to create and grow new online businesses, which will in turn generate payment volume that they can capture revenue from.
If that is the case, then they should be relatively indifferent between building these services internally and funding startups that provide these services, if there is no benefit from integration. In a recent interview with Stratechery’s Ben Thompson, Stripe President John Collison suggests as much (link):
I think there’s a set of questions that we obsess about around what’s required to grow the internet economy, plus what are the impediments holding back businesses...sometimes that might inform us in the sense of we’re going to go build Stripe Capital to solve this need that businesses have, and sometimes it might take the form of investing in a Pulley or an Accord or something like that.
Coinbase is a clear example of a startup that benefits directly from funding startups that develop complements. Crypto is a new space, and according to the Coinbase S-1, they are growing their share of it, with 11.1% of all crypto assets by value being held on their platform at the end of 2020, up from 4.5% at the end of 2018.
The more people that use crypto, the more people will hold and trade crypto on their platform, and the more revenue they will capture. The Information quoted a Coinbase exec as saying “The primary goal is to let 1,000 flowers bloom”. In their S-1, they reported having already made over 100 investments by the end of 2020.
It makes sense for them to encourage the development of anything that will get people to use crypto even a marginal amount, given their lucrative transaction fees. Coinbase reported income before tax of nearly $1 billion in the first quarter of 2021; by comparison, they reported that their venture investments had a combined carrying value of only $35 million.
I would conclude that successful startups end up getting into venture investing because their early success and the economics of their industry creates a unique set of incentives whereby it just so happens they will sometimes achieve the best outcome by investing in companies that build complements. It cannot be overlooked also that early success allows a startup to raise a lot of cash, which makes it easier to get into a venture strategy.
It certainly does not imply that most startups would be more successful if they started a venture capital arm. Indeed, this strategy mostly compounds the advantages of startups that are very successful in the first place, but we would not expect it to turn a struggling startup into a market leader.