Economics of Online Video 2: Unit Cost Structure


After performing a detailed analysis of the economics of streaming video, we continue to believe it is a very tough business–with high capital costs and low profit margins.  In the first instalment of this series, we explored the ramifications of this.  In the second, we take a closer look at unit costs.

Streaming video has a similar cost structure to many text- and graphics-based Internet businesses, with three significant additional costs: bandwidth, storage, and transcoding.  Each of these items increases the costs associated with video streaming relative to that of static content–without a reliable offset in terms of additional revenue.  In contrast to most text-based Internet content, moreover, these video-related costs are variable, meaning that they rise in direct proportion to video usage (not revenue–usage).  This means that the industry will not suddenly become wildly profitable as revenue increases.  On the contrary: It will likely continue to struggle to eke out a profit for many years.

In addition to the three major video-serving costs, which we detail below, there are two other critical inputs into most streaming video business models:

  1. The percentage of videos that are monetizable. Low for video dumps like YouTube; high for network sites and professional “show” hosters like  Because video streamers incur the same costs for monetizable videos as non-monetizable ones, this assumption is critical.
  2. Content production/licensing costs. Relatively low for TV networks, which can repurpose content, and high for the more plush online show production* (costs of shooting, editing, studio, etc.) and YouTube (royalties).

Most companies that store and serve video lie somewhere along a continuum of, say, 20%-100% on these two expense items, with the specific inputs having a huge impact on the potential profitability of each model.  To illustrate the importance of these costs, we have modelled:

  1. A base “streaming video” model that lays out the basic cost structure
  2. A “TV Network” version, which adjusts for low royalties and a high percentage of monetizability.
  3. A “YouTube” version, with huge scale, high royalties, and a low % of monetizable content.
  4. A “niche network” version (e.g.,, with medium royalties, high targeting, and a high percentage of monetizable content.

We are grateful to Mike Hudack of, Dwight Merriman of ShopWiki (an SAI investor), and others for help with this preliminary cost analysis.  Please weigh in in the comments or via email ([email protected]), and we’ll refine as we get more info.  Details after the jump. 

*We’re not suggesting that online video production costs a lot relative to TV, movies, etc.  Relative to those, they’re dirt cheap.  The expensive ones still cost a lot relative to the revenue they can produce, however.)

Here are the key considerations for the potential profitability (and value) of streaming video:

Revenue. Online video monetization should continue to improve, and, ultimately, online video should be as accepted and important an ad medium as, say, paid search.  Recent data suggests that “run of site” video CPMs range from about $5-$20, with targeted sponsorships ranging from $15 to, on occasion, $300-$500.  The high end sponsorships appear to be a bizarre outlier, and as with most other forms of online advertising, we expect that CPMs will drop as the thrill and novelty wears off.  So we are not expecting soaring CPMs to bail out the industry’s high cost structure.  In our modelling, we’ve used a CPM range of $5-$25, with a “base case” of $15.

Percentage of Content That is Monetizable.  We have no doubt that, contrary to popular perception, dancing cat videos will eventually generate some revenue.  Blow-job videos and pirated TV content, however, probably won’t–at least not on sites like YouTube.  Regardless, the percentage of videos that are monetizable at, say, YouTube, is far below that at, say, (a niche network featuring professional “shows”) and TV networks.  This assumption is critical, because if the streamer only monetizes, say, half of its videos, the “effective CPM” will be cut in half.  Our base assumptions are as follows:

YouTube:       30% monetizable.            80% monetizable.
TV network:   100% monetizable.

Content production / royalty costs.  Assuming you’re playing by the rules, you either have to pay to develop video content yourself or pay someone else for theirs.   In the text-based world, Google pays about 80% of revenue out in royalties (“rev share”).  If the video royalties are anywhere near this level, YouTube’s profitability is going to be minimal (if that).  We expect Google will adapt to the high-cost-structure reality by vastly reducing the revenue share it pays to video producers (which won’t sit well with them).  In the meantime, however, we’ve modelled a high cost here:

YouTube:        70% payout             50% payout
TV network:    20% payout

Bandwidth.  Video streaming eats bandwidth.  Bandwidth costs are declining rapidly, of course–which is the great business-model hope of many video streamers–but, importantly, these cost declines are often offset by increases in average video file size, as resolution increases.  For the purposes of this analysis, we have optimistically assumed that the costs of bandwidth, storage, and transcoding (see below) will continue to decline rapidly and that increases in average video resolution will not eat all these benefits.  Specifically, we use a range of $0.05 to $0.15 per gigabyte and a 20mg average file size, which produces a $1.00-$3.00 current per stream CPM.  We have assumed that in the “future,” bandwidth, storage, and transcoding costs will decline by 75% versus today.   If file sizes increase rapidly, this could easily prove too optimistic.  A low-cost P2P solution, meanwhile, is likely years away.

Storage and Transcoding.   To estimate storage and transcoding costs, we have estimated capital equipment costs and then converted them into per-stream costs.  These costs should decline rapidly, too, but not if video file sizes continue to increase. 

Please see the model links above for more details.