Tech’s false idol: Aggregation Theory and the Netflix foil
The myth of zero marginal costs and its systemic implications
Despite Netflix owning and operating its very own content delivery network (CDN), Open Connect, the streaming giant’s content delivery costs are certainly not zero — and they’re not trivial either. This is an important assertion, because “zero marginal costs” are the foundation of latter-day strategy among businesses both small and large, from startups to big tech. Woe the idolatry!
Part of the confusion comes from a lack of attribution: It’s hard to precisely quantify Netflix’s streaming delivery costs, because they’re lumped-in with “Other Costs of Revenue” on Netflix’s SEC filings. This rather benign obfuscation of marginal costs — as is common among Netflix and its peers — has cleared-the-way for the “zero marginal cost” false narrative. However, I’m going to try and estimate those costs herein, in an effort to determine whether or not marginal costs can even be dubbed “effectively zero”, as classical Aggregation Theory and its acolytes assert at their own risk…
Qualifying marginal costs: They do exist!
To begin, let’s take a look at the following footnote from Netflix’s most recent 10-Q:
For the Domestic and International streaming segments, amortization of the streaming content assets makes up the majority of cost of revenues… streaming delivery costs and other operations costs make up the remainder of cost of revenues. We have built our own global content delivery network (“Open Connect”) to help us efficiently stream a high volume of content to our members over the internet. Streaming delivery expenses, therefore, include equipment costs related to Open Connect, payroll and related personnel expenses and all third party costs, such as cloud computing costs, associated with delivering streaming content over the internet. Other operations costs include customer service and payment processing fees, including those we pay to our integrated payment partners, as well as other costs incurred in making our content available to members.
From that disclosure alone, we can infer a couple of relevant insights. First, the majority of Netflix’s cost of revenues are associated with content acquisition — specifically the amortization of content assets. Second, streaming delivery costs are definitely a non-zero line item — no matter how defrayed they may be by Netflix’s proprietary CDN. I’ll discuss the former in a moment, but, for now, I want to hone-in on the latter.
As that excerpt can attest, marginal costs — the cost for Netflix to serve its next customer’s activity or even its next customer — are objectively not “zero”. They do exist! I’ll discuss the marginal costs of adding an incremental subscriber a bit later, but, for now, this footnote from Netflix’s financials has at least manifest a real marginal cost associated with serving active subscribers: When a user presses “play” to watch a show or movie via Netflix, there’s a cost for Netflix to deliver that stream to the user’s device.
That said, we’ll have to quantify that cost were we to establish whether or not it is even “effectively zero”…
Quantifying marginal costs
Per its aforementioned 2019Q3 calendar year filings, Netflix’s cost of revenue took a 50% margin bite out of the company’s top line domestically and 70% internationally. Again, given a lack of next-level detail, we cannot suss-out how much of that is CDN-related.
That said, Netflix does break-out “non-cancelable contractual obligations… primarily related to streaming delivery and cloud computing costs”, which total $635M in the next 12-months (ntm). So, if we take that number as a percent of Netflix’s aggregate ntm obligations — a total of $10.5B legally-binding obligations including almost $9B for content acquisition — then we arrive at a 6.1% pro rata share attributable to streaming delivery. That is as good of a proxy for streaming’s operational “cost of revenues” as we can get.
Applying that 6.1% share to Netflix’s annual run-rate for streaming-only cost of revenues (mrq*4 = $12.3B) equates to a $750M annual spend, which represents a 3.6% margin bite out of total segment revenues from streaming delivery alone — not including DVDs.
In sum, a variable cost that represents 3.6% of top line revenue is neither trivial, zero, nor effectively zero. It’s also not a trivial, zero, nor effectively zero marginal cost.
To ratify the estimates I’ve made herein, we can refer to a 2015 analysis of Netflix’s cost structure, in which ARK Invest estimated a 1.3% to 8.3% contribution margin for streaming delivery in the US:
So, go ahead and pick your poison, however my 3.6% estimate for Netflix’s marginal cost of revenues attributable to streaming delivery is not only an evidence-based derivative bootstrapped from financial filings, but also captured by ARK’s independent range (1.3% to 8.3%).
Other marginal costs: The fixed cost ratchet and reflexivity
In addition, this all ignores the step-like function of fixed costs, which is almost impossible to derive from these standard filings. This also ignores the R&D and capital investments it took to even drop these streaming costs down to ~3.6% in the first place — investments subjected to continual depreciation, amortization, and the need for ever more expensive innovation. Finally, this ignores the step-like function of content acquisition costs, which most financial models disregard (perhaps) due to the long-tail monetization opportunity of evergreen content.
If you were to zoom-in on a particular point in time, it’s unlikely that you’d observe a perfect correlation lock-step between Netflix’s content acquisition costs and net subscriber adds. However, were you to zoom-out and observe the relationship over a longer time horizon, the two do track one-another pretty closely. Take licensed content as an example…
- T=0:
Per “The Prisoner’s Dilemma Wedge”, Netflix started by licensing Starz’s library for far greater fixed cost outflow than its rake from variable subscriber inflow; - T=1:
Eventually, net sub inflows exceeded that Starz content acquisition outflow to provide Netflix with positive cash flow in the narrowest of unit economics interpretations; but then - T=2:
To acquire new users and retain old ones, Netflix ultimately had to license additional content, which cost more than all of the positive cash flow it had accrued throughout the prior timeframe; and then - T=3:
Additionally, so as not to lose content via attrition, Netflix had to renew its Starz license at an higher fixed cost than before, which added to its marginal cost ratchet
In other words, it’s perhaps more apt to characterize these static fixed costs as more dynamic and ratcheting — in which case they look more like marginal costs! These costs outrun subscriber revenues for a period; then the subscriber base rallies from behind to catch-up to the content library at equilibrium; then subs exceed content for a period; then new and increasing content costs jump ahead of subs again; lather, rinse, repeat.
In such a way, Netflix’s real influence on the entertainment industry has been to actually increase the cost of all content — from rights to production to talent — rather than reduce it. As a consequence, not only has the marginal cost of content acquisition increased Netflix’s cost of licensing 3rd party rights, but it’s also increased its own cost of producing proprietary Originals. An exercise in reflexivity!
We can also look at this from the revenue side of the equation — instead of the cost side discussed up to this point. In common size, a lot of these tech companies’ unit costs are not infinitesimally small relative to their unit revenues. For instance, using YouTube as an even more apt example of an aggregator: Were ads fetching $7 CPMs, that would translate to $0.007 of gross revenue per ad served, which would represent, at most, $0.00315 net revenue to YouTube (at its 45% max cut). So, even were YouTube’s marginal cost $0.0001 to serve a video plus an ad, then that’s still a 3%+ slice out of unit margins. And, realistically, I think those numbers are all generous/conservative in YouTube’s favor.
But, I didn’t even need to make these detailed arguments — qualifying and quantifying the marginal costs — because these marginal costs are the foundation of Proof of Work (PoW) blockchains like Bitcoin and a number of other cryptocurrencies, for which miners secure the blockchain and verify transactions by incurring compute and storage costs. ‘nuff ced.
A critique of strategy, not sustainability
Considering the context of these real costs, it’s worth noting the impressive strides by the likes of those such as YouTube and Netflix to exponentially decay their costs with huge CDN innovations. To be clear, however, the effect of these cost-saving inventions on these business' bottom lines are realized via mere classical economies of scale and operating leverage, not zero marginal costs, per “A Critique of Aggregation Theory”:
[T]hese observations apply just as well to yesteryear’s industrial leaders — whether Carnegie’s US Steel, Rockefeller’s Standard Oil, or Bell’s AT&T — who all increased the value of the end-user experience via vertical and horizontal expansions. These aren’t novel concepts. Even “effectively zero marginal costs” are not markedly different than those featured by early TV and radio, which spread across geographies with low marginal distribution costs and effectively zero marginal cost to end-users.
In sum, Aggregation Theory appears to be more of an observation about economies of scale, increasing customer selection, and decreasing consumer price than anything else… Were Ben just talking about decreasing marginal costs, then he’d be describing mere economies of scale; but he specifically underscores “zero” as “what makes aggregators unique”.
In other words, nothing new under the sun.
Stepping back for a moment, this analysis really isn’t an indictment of Netflix’s business or business model — nor the sustainability thereof. On the contrary, Netflix has enjoyed positive, accelerating operating income and EBITDA for which it still has a long runway — particularly in its international segment.
Strategically, Netflix certainly is facing headwinds amidst the launch of Disney+ and other horsemen of the streaming wars apocalypse. While these newfangled competitors are not perfect substitutes for Netflix’s service, they are both complementary unto Netflix’s core userbase and rivalrous unto its marginal customers. In the competition for the latter, non-core subscribers, Netflix is at a slight competitive disadvantage, net-net, by virtue of being a one-dimensional business, in that streaming is the only cash cow it can milk, whereas Amazon and Disney can monetize supplemental business lines like Prime and theme parks, respectively. This is a broader topic for me to cover more in-depth at another time, but for now let’s just say that Netflix still has many levers-to-pull and many offsetting competitive advantages here. Plus, a lot of households are going to get multiple complementary services for free: Disney+ for Verizon Wireless Unlimited customers; HBO Max for AT&T customers; Apple TV+ for iPhone owners; Prime Video for Amazon Prime subscribers; and NBC’s ad-supported Peacock for everyone. All of these dynamics secure Netflix’s current and projected walletshare for the foreseeable future — at least a multiyear window.
Returning to the original thesis at the top, this serves more as an important revelation about the mis-diagnosis that undergirds Aggregation Theory and its resonance downstream. One such example is the lack of both operating leverage and economies of scale embedded in a lot of VC-funded business models, as discussed in “The New News Bundle”: There’s a widely-held belief that our marginal costs will go away and we’ll be left to simply amortize fixed costs ‘once we win this market’ or ‘once we grow big enough’ or ‘once we raise enough capital’ or ‘once we accumulate enough loss leaders’. Such misplaced faith leads to terminally wrongheaded projections for future capital-intensity and discounted cash flow. (The closing paragraphs of “Three’s Company” also do a good job of commemorating the broader strategic, academic, and regulatory consequences of worshipping Aggregation Theory’s false idol.)
But, I digress. To avoid repeating myself ad nauseum, I’ll simply conclude as follows…
I’m not going to debate either the useful life of Netflix’s IP from an accounting standpoint or the value of its back-catalog from a strategic lens, but I would suggest we at least re-evaluate these content acquisition costs for the purposes of Aggregation Theory and general strategic analysis: Whether Netflix acquires preexisting IP or produces its own in-house content (i.e. Netflix Originals), it’s clear that incremental video hours are required to not only acquire marginal subscribers, but also retain preexisting ones.
For this reason, I really do think subscription technology businesses, like Netflix, whose narratives — and therefore valuations — explicitly cite anything akin to ‘effectively zero marginal cost leverage’ should start disclosing marginal cost items, customer acquisition costs (CACs), and customer lifetime values (cLTVs or CLVs) in investor filings. I also really do think that not only FASB/GAAP accounting but also investors need to re-consider how we traditionally handle intangibles, including things like the capitalization of research and development (R&D) — or lack thereof.
Real operating leverage
There’s an abundance of content out there, and more gets produced every day. I don’t want to waste my time or attention, but information overload means that signal gets buried under all of this noise.
Luckily, there is now a network that gives me a summary for everything I want to read and an annotator for anything I want to save or share. That network is Annotote, an antidote to our information overload:
