Blog Post

Online Video is Still Too Expensive

Stay on Top of Enterprise Technology Trends

Get updates impacting your industry from our GigaOm Research Community
Join the Community!

There are a number of ways to deliver digital video content, and delivering that content online is the most expensive of them all. How expensive? Dan Rayburn recently took a look at CDN pricing, and while it is on the downswing, it’s still pretty expensive. For example, a customer who buys 100 terabytes at $0.15 a gigabyte would have to pay around five dollars in bandwidth to serve 1000 views of a three minute video. That means they’d have to get at least five dollars in CPM (cost per thousand) advertising just to break even, and that doesn’t include storage costs.


My math isn’t what it used to be, and companies are understandably closed-lipped about their expenses, so please feel free to correct my numbers. Obviously corporations like Google (GOOG) and Apple (APPL) aren’t paying that much for bandwidth, but the expense is not insignificant, and they aren’t the only people paying — The Register pegs the cost to UK ISPs of delivering BBC iPlayer content at $1.30 an hour. And when compared to delivering video over broadcast, cable or satellite, it’s definitely more expensive per viewer hour.


The good news is that it’s getting cheaper — Rayburn’s numbers show a decrease since the last quarter, and data from TeleGeography shows long-haul 2.5 and 10 Gbps wavelength rental prices declining steadily over the last few quarters and are now nearly half what they cost only a year ago. Google, which deals in massive amounts of bandwidth, has purchased backbone fiber-optic lines to connect their server farms to each other and to peer nodes with ISPs in major metros.

Another problem with costs is the penalty for success — if everyone suddenly tuned in to watch Heroes online instead of on their televisions, NBC would wake up with one huge bandwidth bill hangover in the morning. There are ways to mitigate this popularity tax, such as 95th percentile billing. And of course, using peer-to-peer technology would make things cheaper for distributors, if not ISPs. Cable and satellite connections could care less, cost-wise, if everyone or no one is watching at any one time.

It’s also true that cable companies are facing a bandwidth crunch of their own — there’s only so much content you can pipe over the existing back end hardware and coaxial cable. IPTV promises to provide something of a solution, as not every channel would have to go downstream to every user at any given time. But that would require an infrastructure investment and therefore impact profit margins for near foreseeable future. The cable network has already been built, and therefore beyond maintenance costs there’s no overhead for cable companies.

The easy answer is simply to provide much lower quality video than you would get on cable. As I’ve argued in the past, quality is all about bitrate, and the higher the bitrate, the bigger the bill. To broadcast high definition content over the Internet would be incredibly expensive. Even standard definition bitrates typical of digital cable would probably make YouTube turn from profit to loss (or more loss) even at the $20 CPM rate the new video advertising is said to be making.

So while it’s certainly free and easy to publish video online, actually delivering that video to millions of users in high quality is inevitably more expensive than traditional broadcast. If the telecom companies manage to charge companies tolls for priority over the last mile connections, it could get even more expensive.

14 Responses to “Online Video is Still Too Expensive”

  1. I think that few people realize the true cost of advertising online. I really don’t even feel that advertising online makes much of a difference. People may assume differently, however, considering that more people watch videos online nowadays than they watch TV, but mostly people don’t pay attention to online advertising. At least that’s been my experience.

  2. Well after days of brainstorming, i have come to the conclusion that internet video is not really the key to taking over the internet and the universe, infact what is happening is that, there is now more videos than ever, more than you could watch, soon there will be 1000’s of videos a day for every viewer, so will 1 person watch 1000 videos a day? of course not, people only watch a few good ones, which means.. the video content will become useless, keeping content that is useless will cost money, the solution has come up with is a brilliant one, which wont be revealed in this post because… well there is no because..

  3. I rent a dedicated from AIT for $40 a month and data limit 2,000 GB per month. Now supposing I did hit 2,000 GB, that works out to $0.02 per GB. Much better than other costs mentioned here.

    Of course, in reality I can’t get more than about 800 GB per month. After that the server gets slow. Still that is only $0.05 per GB.

    Also wonder why no one is talking about Real Media. I experimented with different formats, and found that most bandwidth efficient.

  4. Morgan Warstler


    Talk to Olympusat they might get you your signal uplinked cheaper. Remember tho, there’s a reason murdoch gave up on distirbution.


    I think the broadcaster world is comfortable doing On2 VP7 at <700kbps and that’d keep eveybody psyched for a while. Seems right for Fox no?

    Either way Mathew,

    Doesn’t it seem more likely that the broadband video pricing structure in the US is really more likely to be Peak Hour Maximum payment and everything after midnight EST is free? Per gig based on office workers watching 3 minute YouTube clips doesn’t make sense – people watch TV when the offices close.

    Maybe not, but seems reasonable.

    Real issue is packetshaping in 24 mos. no?


  5. Ok, but the ITMS charges $2 for a 43-minute episode of Lost, and their bandwidth costs are (according to your model, and again disclaiming my math) about 3 cents. That doesn’t seem so bad!

    Just for the purposes of comparing ITMS bandwidth costs to that of ad-supported products (and this is flawed for a thousand reasons but still useful for orders of magnitude) let’s think of that 43-minute ITMS episode as being roughly equivalent to 17 three-minute video clips. At Apple’s retail prices, they are charging (retail) $150 for every thousand three-minute views, and paying (surely far less) than $5 per thousand in bandwidth costs.

    Moving back to the advertising-supported world, the full episode player uses roughly the same bitrate as ITMS and includes 3 ads in that same episode of Lost. I believe that this is the highest bitrate in the industry. At the bandwidth cost in your example, it works out to about $23 CPM in bandwidth costs. Cripes that’s a lot, but let’s hope they’re paying way less than that for bandwidth, and since it’s for insanely premium ad inventory, I think they can charge way more.

    Other ad-supported video players are using much lower bitrates successfully. A 500kb/sec file, according to your costs, comes at a CPM bandwidth cost of about $1.75. No big whoop, right?

    No doubt it’s more expensive than broadcast. But it’s not just about cost — it’s about value. Advertisers are willing to pay a higher CPM for interactive inventory versus broadcast because those ads perform better for them.

    Internet delivery costs more than broadcast delivery. But it should, because it’s better for both the viewer, and the advertiser, right?

    –Matthew Rechs

  6. Barry, it’s a good question. And as you see from the first chart, $30,000 won’t buy you as much bandwidth unless you’re moving data in serious bulk with a CDN. Of course, you can always distribute for free using YouTube, Revver or, but they’ll take a cut of any ads for providing the service.

    Matthew, your numbers are exactly what I used. I chose the 1.5Mbps bitrate because that’s about standard for video from the iTunes Music Store. Many online video streaming or progressive download sites encode around 300Kbps to 750Kbps, but the quality is obviously not good enough for a full-screen experience.

  7. Jackson: I’m checking your math here, and what’s missing is the bitrate your example video is encoded at. That matters a lot.

    I tried to reverse-engineer your math. Keeping in mind that my math skills are surely no better than yours, it looks like you’re assuming a average bitrate of 1.5 Megabits/sec (187 kbytes/sec, or ~ 33Mbytes for a 180sec video file). That seems high.

    Email me if you want to trade Excel spreadsheets.

    –Matthew Rechs

  8. the problem monetizing video is that users who watch videos are the ADD demographic of the internet.

    Even regular users who watch videos watch and then leave. The only way to effectively tag videos is to have the users do it and the popular videos have almost no correlation to a product being sold.

    When they figure out a way to compress videos better than .flv, which is leaps and bounds better than previous formats, they might start seeing some ROI if they ad overlays, pre-rolls, etc. Bascially everything that will ruin the User Experience and cause users to go elsewhere….

  9. Barry Bowen

    I want to start a new movie channel. It could cost me $30,000/month for satellite air time to distribute a signal for cable operators to pick up. Or I could spend that $30,000 on bandwidth and storage for an IPTV channel.

    Now how much bandwidth would $30,000 buy?