In 2009 I wrote a post detailing what Netflix's direct costs were for streaming movies and TV shows over the Internet. Since then, Netflix's cost to stream two-hours of video has dropped by about half and I estimate the company will spend about $50M this year with third party CDNs for video delivery.
Two years ago, Netflix paid about five cents to stream a movie and today, pays about two and half cents. While most video contracts with third party CDNs are typically priced on per GB delivered model, Netflix and other large content distributors usually pay the CDNs on a per Mbps sustained model. They pay not for the total number of bits they transfer each month, but rather the total amount of bandwidth they peak at each month, a pricing model also referred to in the industry as 95/5. This means that a customer can burst above their committed rate of Mbps less than 5% of the time with no penalty, but once they go over that, they pay for overages.
Like most videos viewed on the web, Netflix's movies and TV shows are encoded at multiple bitrates to allow for the different broadband connection speeds and performance of today's ISPs. Netflix has given out some details that shows the average user streams Netflix at just over 2Mbps (2000Kbps) even though they encoded video for twice that speed. Most Netflix users watch content in 480p or 720p quality, but Netflix does offer a limited amount of content in 1080p, only available via playback on the PS3.
What's really interesting about the bitrate data that Netflix provides is that it shows just how bad most consumers ISPs really are at delivering video. In 2009, most users were getting about a one and half to two Mbps stream from Netflix. Two years later and Netflix's data shows most users are getting about 2Mpbs. That's almost no video quality improvements in the rate of Kbps being delivered in two years time. That's not to say that Netflix isn't improving their video quality because they are working to make their encoding more efficient as well as working with their partner CDNs to deliver the bits more efficiently. But even with those efforts, Netflix has no control over the last-mile providers, who still treat Netflix traffic as if it's a burden on their network.
While there has been a lot of talk lately about new bandwidth caps by AT&T, that really has no impact on Netflix because as we can see from Netflix's own data, bitrates and video quality aren't exactly growing very quickly, if at all. AT&T said that 2% of their users account for 20% of the traffic and the caps are a way for them to manage performance on their network. I think rather than going after the 98% of their customers with caps they should really go after the 2% causing the problem, but that aside, caps from Comcast or AT&T won't have any material impact on Netflix now or anytime in the near future. Years from now they could, if the ISPs don't raise their cap levels, but right now caps are not a big deal to Netflix.
Some might suggest that the Netflix pricing data shows that bandwidth pricing is falling fast, and while it is for a company like Netflix, most content owners aren't seeing this big of a decline in CDN pricing because they don't grow traffic as fast as Netflix. Overall, I expect the average CDN customer, which Netflix isn't, to see a pricing decline of about 20% this year. Of course the good news for Netflix in all of this is that their pricing to stream video continues to decline and it's cheaper to stream than mail DVDs through the mail. And with video bitrates not growing very fast, Netflix's cost to deliver video remains very stable and overall, is a very small percentage of their cost to run their streaming service.