CDN Market Pricing Down, but Overall Growth Continues
Last week I had the pleasure of attending Streaming Media East, including the annual Content Delivery Summit that takes place the day before. As many of you will know, this is the sharp focus of my own personal interest in the sector, so as always, I enjoyed many of the seminars in full.
While I have the honor of chairing Content Delivery World annually in London, there are some key distinctions between the two events that make them both complimentary and valuable.
The first of these distinctions is that, in true keeping with Streaming Media's broader editorial policy, the Content Delivery Summit takes a broad focus on case studies from customers and consumers of CDNs, where Content Delivery World provides a platform for CDNs to present themselves to the market directly. This is not an absolutely black-and-white distinction, and both events will have sessions that buck that trend. However ,Content Delivery Summit is a much harder event to program, because motivating clients of CDNs to speak about their experiences "altruistically" while informing their competitors of their strategies is much more complex than motivating the sales and marketing teams of the CDNs themselves to lay out their wares. This is largely enabled by Dan Rayburn—who programs the Content Delivery Summit—since he has such longstanding and high esteem in the sector, and a vast personal network to draw on. Knect are positioned differently and have a deep heritage in helping telcos and networks market themselves.
Both fulfill their purposes well, but there is one annual Content Delivery Summit session that stands head and shoulders above all others, and that is Rayburn's own "CDN Market Update: Web Performance and CDN Pricing Trends." Even on those years that I haven't been in New York myself, I have tried to tune into a webcast as soon as it is available. The session is an empirical barometer of the sector. It is also relatively fixed in its format, which in itself allows year-by-year comparison as one reviews past editions. This in itself provides a consistent reference on several key data points, and yet is interspersed with broad commentary and clearly identified opinion from Rayburn's own experience and insight as he has collated the data.
So while the presentations were very interesting throughout the event—particularly the opening keynote by Leonidas Kontothanassis, engineering director, Google on their cloud strategy—let's focus in on the key points in Rayburn's update this year since they reflect a relatively neutral perspective.
Pricing is Down as Much as 45% (Average 22%)
For the largest commercial committed contracts, pricing per GB is down to $0.0025. This is down from $0.003 a year ago.
Rayburn derives his pricing through a comprehensive survey of nearly 900 customer contracts that he painstakingly retrieves from the many direct contacts he has. This is by far the largest such analysis of its type, and is absolutely a bellwether for the state of commercial pricing in this space. There are variations in the drop as one scales the contracts down to smaller volumes of traffic. When one accounts for general telco pricing drops (and other market forces), this may not reflect on a decrease in margin for the CDN operators. Indeed, given increased efficiency in the various architectures, and also given continued consolidation in the sector, Rayburn highlighted that this pricing drop is actually only a small part of the wider picture. Value-added-services and broader ecosystem offerings, coupled with a divergence into other models outside of simple media traffic delivery such as (in particular) security and software delivery, mean that the CDNs are broadly stable, if not growing as a whole.
In fact he noted that while video takes up the largest percentage of traffic on a CDN, it typically contributes the least amount of profitable revenue.
Most CDN customers have, or are well underway with a multi-CDN or hybrid CDN strategy now. This is in part a financial/business risk-spreading exercise, and in part a simple QoE optimization strategy allowing for variance in service from all CDNs over time.
Live is Emerging, Not Leading the Way
A central feature of the sector this past year has been the rapid emergence of new and reasonably exciting live and live-linear consumer propositions. Several of the larger publishers have finally managed to facilitate rights negotiations to both rebroadcast existing live events and channels online, and also secure specific online broadcasting exclusive opportunities—with particular mention of both Hulu and sporting events. However Rayburn was also cautious to note that volumes were still quite low when compared to the numbers that surround traditional live broadcast models, with even the largest subscriptions ranging from around 2 million down to (more typically) 150,000. In fact he estimated that a grand total of all U.S.-based live subscription models was still in the region of 3 million, representing less than 3% of the wider Pay TV market.
For my own part I have to note that those numbers are actually quite impressive, and if they have emerged in just a year or so then this trend should make Pay TV operators sit up and take immediate notice, even if by volume they are not yet to be considered a direct threat.
He also, critically, noted that those subscriptions are almost always spread across multi-CDN contracts, and so while the sector is showing a clearly emerging trend, this growth is not concentrated in any one CDN's revenue. Indeed, just to add salience to this, he threw up the note that with an approximate 80 hours average viewing per month, at $0.004 per GB the Average Revenue Per User (ARPU) was still a very small amount of any CDN's gross income.
Mobile Growth in User Volume, not Necessarily Traffic
Mobile, mobile, mobile: We all hear astounding figures about the explosion of mobile. But Rayburn had a glass of cool water for the mobile sector, and one that I feel was a point well-made. Typically mobile video bitrates settle at 750Kbps. This compares to a direct-to-smart-tv or direct-to-laptop delivery on fixed line which may settle at an average of around 3.7Mbps. While this may sound like comparing apples and oranges, Dan made the excellent point that still the majority of mobile video is actually delivered to the device on Wi-Fi. This means that mobile usage actually erodes fixed line traffic volumes as users migrate from their laptop to a fixed-line/Wi-Fi connected tablet or smart phone. He gave the example of Netflix, which may serve bitrates as high as 17Mbps (for their few 4K consumers) to fixed line consumers, who may switch to a mobile or tablet device in the home and drop to a data consumption of less than 5% of that peak rate. So mobile uptake in its broad sense may actually reduce traffic volumes as it replaces some fixed-line delivery models.
It was a point I had not considered in this way before, and only now as I review my notes do I regret not finding to question him more about it . However, I am sure that there are (or soon will be) some forthcoming posts from him on his blog with more of a break down of this. It is a topic that deserves some attention.
4K is Still Very Much a Pipe Dream
Much has been said about 4K, but in practice 4K still faces many challenges. As mentioned above, Netflix only delivers a very small amount of 4K content. Bandwidth requirements are extremely high, and so far the increased costs are proving hard to justify, let alone show a profitable return. 4K is well established in the production space, and it does doubtlessly bring benefits in a number of ways when shooting content. Yet at the consumer end 4K presents a number of deployment challenges, and it is often debatable if the end user can really benefit. Many of the largest live events and live streaming services, for example, are still optimizing their peak bitrate at 750Kbps. At this speed not only does it guarantee a lower cost of operation, but it also generally ensures a more consistent subscriber experience across the multiple devices and connectivity scenarios.
VR is Even Further Off
"VR is so far away from a reality that CDN's shouldn't even be allowed to talk about it," said Rayburn, and I think I clapped spontaneously at this point. (I do like a strong disruptive opinion!). Remember the Content Delivery Summit is about the business of CDNs, which are themselves, large commoditized service providers. VR is undeniably an exciting and potentially transformational technology. It requires large traffic volumes, and so is of course of interest to CDNs. But so far there is little more than some experimentation available in the market, adopted by an extremely small community, and so while CDNs are right to be abreast of this new technology, in practice VR is still very pre-market.
In-House CDNs Rarely Worth It
Even though some services have subscriber bases of ~15 million, Rayburn argued that there are few if any large publishers who can justify a return on investment of building their own CDN. CDN is a complex business and requires a specific set of skills—often very telecoms- and network engineering-based. These are not skills that publishers have in house, and given commoditization in the sector and an abundance of good operators available to the publishers, the cost of building a CDN in house, coupled with the complexity of supporting such a model only makes sense to the Apple/Amazon /Google "hyperscale" operators at this time. These comments were focused on U.S. businesses in particular.
Dan was asked about his thoughts given changing trends in operator infrastructure, and if he thought operator CDNs would disrupt the market in the immediate future. He acknowledged the changing landscape but said that the build-outs were slow, and at the moment the larger CDNs were proving effective in carrying the OTT traffic into and across most operator networks. So while there is some change occurring, it was too early to be able to see any data on that.
From my own point of view I would agree with that, although I personally think that is likely to change rapidly when one looks further ahead than just one year. However my view on that is based purely on my own first-hand (engineering focused) work in the space with various telcos, and not on the broad accountable market view of today that Dan (rightly) is putting forward.
Major CDN Vendor Roundup
As a final series of comments I want to highlight some specific notes he made about particular large vendors in the space. These notes are purposefully brief, and if you want to know more detail do obviously contact Rayburn for more specific information, or you can indeed ping me if you want some more engineering-focused commentary.
Typical media services have seen price cuts of as much as 30%. Akamai continue to be the 800 lb. gorilla in the OTT CDN space. While this position has typically meant that the very largest publishers have been key customers, Akamai suffers from attrition as these are the few customers who do in fact turn to the advantages of a DIY/in-house CDN. Central to this has been Apple's estimated drop from ~$150 million annual spend with Akamai to ~$10 million over the past three years. And while Amazon has stated that they use third-party CDNs in part today, by 2018 they aim to have brought all that traffic in-house.
Akamai have said "We've seen some of this in past: as traffic grows, people who have been trying to do it themselves at some point may find that it wasn't as effective as they thought it would be, and then they come back to us." I guess only time will tell if that is wishful thinking from Akamai or not.
On the subject of Amazon, Dan was relatively effusive. He pointed out that they face different challenges than most other organizations. They have little or no capital restraint. In fact the most significant challenge Amazon face is in recruitment of the right skills. "There are literally thousands of jobs" at AWS at the moment, Rayburn said, so simply onboarding the right people is a massive hold up for them. He conjectured that this might have some bearing on why the massive acquisition of Elemental two years ago may not be emerging as a full integration, noting that AWS Elemental was really coming to market as a part of Amazon Instant Video, rather than being maintained as it was as an independent vendor, or (more as he would have expected) as a tightly integrated part of the wider AWS offering to AWS' many customers.
He noted that AWS and Google Cloud were strongly positioned as platform providers, and in some ways Google's acquisition of Anvato is a strong platform play, where AWS could probably eventually do the same with Elemental. He also noted that both differ greatly from Azure whose offering was much more "siloed."
One of two big surprises was hearing Rayburn effusing strongly about Limelight's "transformation" in the past year. It has been rare to hear him talking positively about Limelight for a good while, but clearly their recent turnaround has struck a strong chord. They have recently turned a profit, have (by and large) passed the complexity of the very public patent battles with Akamai, have got their product lined up well, and have focused on their strengths extremely well.
I joked with a few Limelight folks I know that Rayburn's comments should have rightly bumped up their share price a few points last week. And to be fair I totally agree (acknowledging the bias I have from my closeness to several members of the team)—they really do seem to be sailing with a strong following wind at the moment, and it is great to see them properly challenging Akamai in the media space again.
Smaller CDN Vendors
The following notes come from Rayburn's slides, a link to which can be found at the end of this article.
Did $100M in revenue last year, 90% from performance/security. Has raised another large round of funding, number being circulated is $50+ milion. New money wasn't needed, but now allows them to grow even faster. Recently rolled out new security, image optimization and other services.
On a run-rate of $100M in revenue for 2017. Pricing deals based on level of performance shown over competitors. All focused on performance and security, no video.
Still focused on the media vertical, no performance solutions, but security revs were $200M in 2016 and they estimate that will grow 25%-30% this year. Media revenue was $170M in 2016, which was a 13% growth YOY with projection for 18% YOY growth in 2017.
And a final note from a question I asked across the floor: Since it was a topical news story last week, I asked Rayburn if he foresaw any significant risk vectors from renewed discussion in and around the FCC and net Neutrality regulation. In simple terms he felt that there was little adverse risk at this time. From my point of view net neutrality is a big "box of frogs" and it can be unpredictable, but given the current U.S. administration seems to broadly favor liberalization of the telecoms space, I too think the risk is low. That is not to say that that situation might not spin on a pin at any time—net neutrality seems to be a random hot potato that rears its head in un predictable ways from time to time—however for the moment I would agree that the risk of an adverse regulatory change which affects the CDN space is low, and it was good to hear that Rayburn felt the same.
So all in all the CDN space is actually in a good shape at the moment. The market is growing, albeit not as fervently as some would imagine in pure traffic terms. However, focused diversification is paying dividends in many cases. There was a fair amount of talk about the wide range of CDN brokers and QoE measurement companies that have burst out like a rash in the past couple of years too, with numerous new entries disrupting the ecosystem. Rayburn felt that this was an extremely good thing for the industry since increasingly CDNs actually have to focus on meeting SLAs against real benchmarks. Indeed Rayburn cited Limelight's new proposition where they actually offered service guarantees and commercial backing underpinning those—where they were to be measured by third-party QoE measurement companies—as something a brave and strong thing to put forward, and something that was most importantly going to be good for the consumer.
The full session webcast will be available soon on StreamingMedia.com, and Dan's slides are available here.