Facebook’s newly launched internet.org coalition to deliver mobile internet to everyone is more than its attempt to connect to its next billion users. It wants to reduce the cost of mobile data for the 5 billion people who are not online by making mobile data cheaper to deliver and apps more efficient.
Facebook has signed up top wireless chipmakers Qualcomm and MediaTek, as well as device and equipment makers Samsung (Samsung has a big chip business too), Nokia and Ericsson, to take part in this coalition. In a call last night with Jay Parikh, the head of infrastructure at Facebook, I came away thinking two things — one, that Facebook doesn’t seem to have much beyond a plan to make this happen, and two, if it really wants to shake things up it should try to bypass carriers as much as possible.
Facebook’s plan is probably to work with the carriers as much as it can, but there are plenty of challenges to that approach.
Where’s the there on this plan?
Let’s start with the obvious here — this reads at the moment like Elon Musk’s Hyperloop plans, only there’s no detailed PDF to detail how this will happen. The announcement and accompanying white paper spent a lot of time discussing the problem and big ideas for solutions without ever specifying the details needed to understand how Facebook and coalition members will make this happen.
The paper cited Facebook’s previous experience with the Open Compute Project — a customer-led coalition that helped forever alter the $55 billion server market, as an example of Facebook creating change in an entrenched and inefficient market. OCP began in 2011 when Facebook launched some servers it had built for itself and it has since managed to put the large server makers on their guard. Facebook’s latest data center doesn’t even have servers from the established vendors Dell, HP or IBM anymore.
But to equate Open Compute to disrupting the carrier business is like saying because you own a dog you know what it’s like to have a child. And when I told Parikh that, he said that he’s aware of the challenges and we even discussed them at a high level, but he was unwilling to go into details about how Facebook will move forward. The company does have some trials in unspecified countries at the moment, but Parikh declined to name them.
One potential example could be Facebook’s partnership with Nokia in Mexico, where Facebook offered to defray the data costs associating with accessing Facebook on Nokia’s Asha phones sold through Telcel. As a result data usage was up 40 percent and device sales we up 20 percent. Facebook and Nokia are taking the model on the road to Airtel in India and Africa. This model helps bring in carriers, device makers and Facebook in lowering the cost of Facebook at least for end users. If one assumes Facebook is a good gateway drug for more internet usage, then this might help advance connectivity. At the least, it helps Facebook find users in an area where finding money for data plans can be tough.
It’s possible that Internet.org’s biggest role will be financial, not technical, mobile analyst Chetan Sharma told us. “There is probably a big subsidy element to all of this,” Sharma said. “Facebook may just put a bunch of deals in place where it pays carriers to make a portion of the mobile internet free.”
Spectrum, core and towers, oh my!
But if we start with the larger problem of lowering the cost of mobile data for the carriers at an infrastructure level (rather than just shifting those costs), there are several elements to deal with: from the right equipment to the laws of physics. A mobile network has highly specialized gear in data centers and out at tower sites that communicates data over licensed airwaves. I’ve long argued that to lower their costs, the operators need to take a page from the webscale players’ book and build out cheaper gear that’s open.
Efforts such as Verizon’s involvement with the Open Network Foundation are helping carriers find ways to cut their core equipment costs, while vendors such as Alcatel-Lucent and others are trying to reduce the cost of the Radio Access Network (or RAN), which makes up the lion’s share of all mobile infrastructure costs.
Here Facebook’s coalition partners might come in handy. Ericsson and Nokia are both driving RAN costs down by making tower siting and gear deployment more flexible, but Nokia has also developed a technology called Liquid Radio, which eventually aims to remove all of the processing functions from the cell site and stick it in the data center. This kind of webscaling of the radio network is called Cloud-RAN, and it’s an ideal fit with Facebook’s own compute philosophy. But it’s still years away from being implemented in a commercial network.
There’s another huge cost for carriers — that of the airwaves they use to send data. For example, Verizon alone spent $10 billion to acquire the 700 Mhz licenses to build its LTE network, but research group iGR projects that all U.S. carriers will have spent a total of $37 billion on building their LTE networks by the end 2017. In developing countries, spectrum auctions are a large source of government revenue, carrier expense and controversy. When I asked Parikh about escalating spectrum prices as well as government intervention adding to costs of delivering data in the developing world he acknowledged the problem but didn’t offer up a solution.
Shannon’s law is very different from Moore’s Law
And finally we get to the physics. Zuckerberg can talk about building more efficient mobile networks as much as he wants, but natural laws are far less mutable than manmade laws and there’s a big one governing data rates over the airwaves. Shannon’s Law describes the maximum amount of usable data that can be transmitted over any communication channel, whether wireless, wireline or even plain speech.
Even as we move into the more sophisticated stages of LTE-Advanced we’re going to max out the amount of data we can squeeze into the airwaves at about 30 bits per hertz per second. There’s a capacity limit in the mobile airwaves, and carriers are using strategies like Wi-Fi offload and spectrum re-purposing to get around that limit.
But the basic thing to note is that while computing is governed by Moore’s Law, which follows an exponential curve, Shannon’s Law describes a far more incremental improvement in capacity. That’s one reason why the cost of storage falls so rapidly, while the cost of cellular data is relatively constant.
Where Facebook could disrupt everything
So given all these challenges on disrupting the cost of delivering cellular data (I didn’t even tackle the app piece), I would still love to see Facebook actually make something happen here. And given its coalition members and its experience with Open Compute, it has many of the tools to do something really disruptive — remove the carriers from the equation. Facebook isn’t planning to do this, but if it did so, it could follow a similar path that Google is with its Google Fiber efforts — making a big, bold bet on infrastructure to show the world how different (and better) things could be.
Facebook doesn’t have to eliminate carriers entirely, but through technologies such as white space broadband, microwave, peer-to-peer networking and laying its own fiber it could build out networks that might take over some of the function that current mobile networks offer. And like Xavier Niel, the man who built Free in France, it could do so more cheaply by building its own core and RAN equipment as it did with OCP — only this time working with Ericsson, Nokia and Samsung.
The coalition document mentions technologies, such as white spaces broadband, that uses unlicensed spectrum to deliver broadband wirelessly to stationary places (as opposed to true mobile usage where people walk from access point to access point.) However, when I asked Parikh if Facebook was going to focus on wireline access at all, fixed mobile or mobile, he did tell me that mobile was the focus, so perhaps the company is thinking of a new way to use white spaces as a truly mobile technology (given the 1-mile radius for white spaces, though that’s a lot of towers).
Bypassing — or reducing reliance on the carriers– can be done. In the U.S., Republic Wireless is doing it, and David Morken, the CEO of Republic Wireless’s parent company Bandwidth.com emailed me to say that Republic’s cellar data usage is only 11.5 percent on the network and Wi-Fi data is 88.5 percent. That’s one reason Republic offers unlimited service for $19 a month.
In France, Free offers a mobile service that takes advantage of Wi-Fi as well, and it is struggling to meet demand, but is also forcing French carriers to adjust their business and costs. Now both the U.S. and France have huge existing wireline infrastructures that provide backhaul, which can be missing in developing countries.
But if Facebook’s internet.org coalition can focus beyond the cellular world and use white spaces broadband, its existing fiber and maybe even new peer-to-peer ad hoc networking technologies that my colleague Kevin Fitchard has covered, then we might see the start of something new and something totally disruptive. Kind of like what Facebook did with Open Compute.
And that would be worth all the hoopla.
Kevin Fitchard contributed to this piece
This story was updated on August 27 to change the percentage of traffic going over Wi-Fi on Republic’s network.