Blog Post

How Much Will Google's Fiber Network Cost?

Google on Wednesday announced an audacious plan to build what is essentially the most cutting-edge broadband network in the U.S. While it’s being misportrayed in certain segments of the media as an ISP effort, in reality it’s nothing more than an experimental network, much like Google’s early efforts to provide municipal Wi-Fi in the city of Mountain View, Calif.  It will be a trial-only network, not Google’s entry into telecommunications services. And while the planned network won’t be cheap, in the end it will be worth the price.

The idea behind the network: provide bandwidth and see if it fosters new user behavior and thus innovations.  I admire Google for creating a real-life laboratory that will provide intelligence to predict not only the future of the web, but also help it develop new products to stay relevant. By announcing this network, Google also showed why it’s quite distinct from its onetime peers such as Yahoo (s YHOO) and AOL.

When I said that Google’s plan was audacious, I said so because of the cost. For starters, Google wants to offer 1 gigabit-per-second speeds to some 50,000 to 500,000 people. At 2.6 people per household, that roughly translates to between 20,000 and 200,000 homes. Our friend Ben Schachter, Internet analyst with Broadpoint AmTech, estimates that it will cost Google between $3,000 and $8,000 per home, or roughly $60 million to $1.6 billion, depending upon the final size and footprint of the network. If Google reaches, say, 100,000 homes, it would cost the company about half a billion dollars.

The folks from Calix Networks, a company that sells gear for FTTP networks, have developed an equation that allows them to calculate the cost per household depending on population density, which, according to them, is the single most important factor in calculating the cost of FTTP connection per home. These costs are quite varied, in some cases as much as $4,000 to connect a single home. Google’s final tab will depend on where it decides to build out the network.

The end cost is also determined by the kind of technology the company uses. While Passive Optical Networking technologies have come a long way — newer versions of Gigabit PON (GPON) have a range that extends up to 40 kilometers — there is a better than good chance that Google will opt for an all-Active Ethernet approach. Mike Fox, business development manager for the carrier networks division at telecom equipment maker Adtran, says that both technologies can get the job done. It just depends on how Google wants to build its network. Fox helped me break down the comparative costs of the two technologies.

A typical PON is made of two pieces: an optical line terminal (OLT) at the service provider’s central office, and an optical network terminal (ONT),  which is used to terminate the fiber optic line and is typically outside the customer’s premises. A single ONT costs about $330, according to Fox. Since PON is a shared fiber technology, OLT costs are calculated in terms of ports and are about $80 per port. So the total per household is about $410. The fewer the number of ports, the higher the per-home costs.

Of course, when taken together with other costs such as fiber and construction, things start to add up pretty fast. In comparison, an Active Ethernet-based network is pretty much like a corporate Ethernet network. It’s a giant switched Ethernet infrastructure. It costs over $600 in electronics for an Active Ethernet-based network. This is closer to what Google has in mind, according to some experts I spoke with. But that’s not all.

Mike Day, chief technology officer of ADC Telecommunications, told me that this network is going to be a lot more expensive than somewhat similar ones. It would need a really fast switching fabric that in turn would be connected to the Internet backbone at astonishingly fast speeds. This seamlessness is what will bring true speed to the homes on the test network. Think of this as building a smooth Autobahn from the home to the backbone. Day said the network would need to overcome some major design challenges such as different data center architecture and a different style of servers that don’t become a bottleneck and are able to leverage the 1 Gbps speeds.

Is spending this much money — even for Google, which has about $25 billion in cash — a good idea? I think so. Just as car companies spend their R&D dollars on Formula One Racing teams to get a better idea of what new features could be included in their commercial vehicles, a company such as Google needs to explore the outer limits of broadband. (I will explore more on this topic in a different post.)

In addition, “Google has a secondary motivation here and that is to also push the FCC to accelerate its examination of using TV white spaces for wireless broadband,” says Jeff Heyman, broadband and video analyst for Infonetics Research. He points out that if “Google can make this endeavor successful for a number of communities, why couldn’t they do so for even more using white spaces? This FTTH initiative, in other words, could be a proving ground for Google as infrastructure provider.”

Is that likely? Perhaps not — but as someone who hearts broadband, I hope Google’s desire to push the limits brings about the change we so badly need.

34 Responses to “How Much Will Google's Fiber Network Cost?”

  1. We are currently building FTTH for about $500 per home passed. We are an over builder so we only pick up about 30% of the homes passed. Ironically over 50% of our customer take the cheapest package. We just updated that speed from 768k to 5mb after failed attempts to sell 768k/5mb/25mb speeds. We could give gig-e speeds today but now one would pay an extra $10.

    Dan Jackson

  2. Longview, Texas remains in the Google fiber spotlight. Still as the only community over the entire East Texas region vying for fiber-to-the-home, Longview is gaining national recognition. Longview has many assets to be attractive to Google-abundant resources: electricity and other utilities; diverse workforce; raw land for development and expansion; new public school faculties and several private schools; transportation hub of air, rail, interstate access; excellent universities and colleges; temperate climate; recognized medical facilities; and receptive and supportive City and County Governments – not to mention dark fiber.

    • Google’s fiber will light up select communities. Longview is hopeful to be the city that welcomes Google to Texas. With speeds of 1 gigabit per second Google will be responsible for the Web 2.0 to Web 3.0 transition.

  3. Asheville NC sees the cost to install the last mile as cost prohibitive for companies. It must start somewhere so apps that consume that bandwidth can be developed. IF your town is vying for the opportunity as mine is you can use our tools to keep up with how your town is doing.

    Jason Hill

  4. Japan already has a 1 gig fiber optic network and it costs about $50 per month for the end user.

    The cost of a personal computer was ridiculously unrealistic for the first 15 years but it emerged against all odds. These evolutions are inevitable.

    Asheville, NC is now in the top 5 running for the install and more than willing to evolve the experiment.

  5. Longview, Texas has been in the Google fiber spotlight. As the only city in the entire East Texas region vying for fiber, Longview is gaining traction and a lot of local support. In a recent press release local businessman Shawn Hill said, “What’s most important now is that residents get involved by the masses.” Hill feels this will help demonstrate to Google how badly Longview wants to be a host community for their fiber project.

  6. Why don’t they ask Occam Networks and Zhone about costs in addition to Calix and Adtran? When looking for the best solution, or a healthy mix of solutions around some GOOG standard, why play favorites?

    Occam and Zhone, FYI, are closer to breakeven (vs. Calix) with sustainable margins (~40%ish) selling the type of equipment that Google wants/needs to deploy for the project. I don’t place much weight on Calix study numbers because it seems that from their S-1, that they have been buying marketshare for years…

    Opening the discussion to others would also get GOOG a good feel for the price per port deployed at various densities in rural communities, not just big cities.

  7. In order for Google to fulfill its goal to offer true network neutrality and network speeds up to 1 Gbps, it is critical that they carefully consider their technology choice. Selecting a technology that is fully standardized will enable them to offer a true neutral open system. Telco Systems has helped many municipalities over the past 10 years succeed in offering its residents this type of network using Active Ethernet.

    There are important advantages for municipalities to deploy a high-speed FTTH network. Municipalities like Grant and Mason counties in Washington decided to use Active Ethernet when they built their network several years ago. As a result they have seen an increase in the number of small businesses and home-based businesses, larger businesses establishing facilities in the area (Microsoft, Ask Jeeves, Intuit, Sims, trucking companies and online engineering firms), as well as an improvement in the efficiency of local city administration. These deployments have been well documented by the FTTH Council and are proving to be not only profitable but also beneficial to the community.

    I hope that the Google activity will in fact serve as lobbying activity in the 14 states which have limited or banned this activity by city government. Incumbent telephone companies are reluctant to invest in upgrading their networks, so it is only through competition that we can hope to achieve open access and improve the lives of every citizen.

    Irit Gillath, of Telco Systems

  8. Small rural ISPs could offer better than my 512K/256K fixed WiFi bandwidth if they could get cheaper backhaul. How ’bout Google offering that? Replace several 40mi. T1 links to the nearest big-city with access to the fiber that already passes thru this valley, please!

  9. Om, I share your enthusiasm for the Google investment in raising the bar of broadband expectations for America. It seems that the nation’s Economic TeleDevelopment has been in hibernation for way too long. Perhaps we can now create the public policy momentum to move beyond accepting “good-enough” network infrastructure. I’m hopeful, once again.

  10. Google’s ‘high speed network’ has been in place for YEARS… This is merely a packaging exercise…. ESPECIALLY embarrassing to the brain-dead telco’s that have written rules to put themselves in a box they can’t escape and Google is now 2-3 YEARS ahead of the telco slugs!! HA!

  11. In news that will probably not be heard in the US. Reggefiber the Dutch Fiber network with several hundreds of thousands of homes connected in an active Ethernet network will move to 1Gbps as of this year. XMS an ISP on their open network will offer 200/200mbps in the town of Zeewolde as of today.

    The costs of a fiber network are probably more in the range of 500-1500 dollars for the passive infrastructure. IN the Netherlands Regge is paying roughly 800 euro per home connected for direct buried cable.

  12. wonder if they’ll use the fiber optic cable from corning that can turn/bend around corners (instead of having to only turn at right angles with a piece of hardware… and i’m curious if they’ll choose a couple of buildings in a downtown area (as @benoit suggested), or choose bigger geographic area in the ‘burbs…

  13. Forget current-gen solutions (especially GPON), there’s way too much cost & baggage built-in to those solutions to accommodate things which won’t be needed here: RF over Glass? POTS? Connectivity to STB’s… forget it. This is where Google can really showcase what a clean network approach can do – they don’t have legacy crap to worry about.

    A few additional thoughts I scratched out yesterday after reading the announcement:

  14. Good and interesting analysis. I am intrigued by this:

    “…there is a better than good chance that Google will opt for an all-Active Ethernet approach.”

    Why is this? One of the attractions of passive architectures is that there is less downtime and risk–i.e., the more passive, the less stuff there is that can break. You seem to imply that it will be active. Why?

    • Dial Tone

      A PON network uses splitters to deploy services. Meaning that from a single port in a CO you will split upto 32 or 64 times. When you “split” the single, you share the bandwidth. So, on a GPON it is roughly 2.5G down by 1.25G up. Split that 32 times and you about have 70Mbps to the home. GE-PON (or E-PON) is better, but I am not sure of the specs, but I think it is around 10G. But that 10G would still be “shared” I think.

      “Active Ethernet” is a dedicated “home-runned” fiber from the customer prem to the CO/Node. Making the link between A & B more “future” proof and only needing to change out the electronics on each end to speed it up. Which in turn does increase the cost to deploy though.

      As far as equipment failures, on PON when a port or splitter goes bad, you could loose upto 32 to 64 customers. On active it would be just 1. I do know that all equipment can fail so it’s hard to say you can’t loose more with either technology.

      So to get 1G to each home today, Active is the way togo.

  15. The “new behaviors” implied must be both television/video related as well as apps related. In order for these things to succeed we need data throughput. It can be incredibly frustrating to want to watch a video and sit through a slow load or erratic behavior. If something like YouTube is ever going to replace television — I believe it will be content delivered through Wave-like tools — then a new network is required. And, the majors are not going to build it without competition.

  16. I think the cost estimates that you quote are highly over-inflated.

    Two things to keep in mind: Verizon is deploying exclusively to single dwelling units (homes) and only very recently have they started to deploy to multi-dwelling units in NYC. The cost differentials are huge. Depending on Google’s chosen location, the costs could be significantly lower. The other thing to keep in mind is that Verizon is deploying using a two-step model where they only pull the last few meters and connect a customer when he signs up to the service. That means a truck roll for each customer signing, which inflates the overall cost per home connected.

    The second thing to take into account is that Google is clearly not looking to deploy from scratch. The whole RFP to communities thing means they want municipal involvement, and that in turn means ducts. It’s even explicitly stated in the FAQ.

    With available duct space (from municipal infrastructure) the costs could drop considerably as the amount of digging needed suddenly goes down.

    I would think that if they get the right choice of municipalities (and I’d be surprised if they didn’t already have something in mind) they can deploy for less than $600/home all things considered.

  17. Partho Choudhury

    I especially like the analogy with Formula One being used as an incubator for newer ideas to put into commercial vehicles.

    The super-high bandwidth rollout just announced by Google is just one of those rabble-rousing ego-trips that they will continue having for some time (they have a ton of money in hand to afford as such)…..they also bet heavy on the 700 MHz wireless spectrum auction, but had no intention whatsoever to run a wireless service….they just wanted the American consumer to get a taste of what a truly open wireless access system would look like, and now that the consumer had tasted blood, I hope better sense prevails. Moreover, Google never cared about the nitty gritty of running a wireless service (since they had no intention of doing so in the first place). They only wanted open networks in order to ensure that their own services are not restricted/outright blocked by AT&T or VZ….same case here – they want to scare the hell out of VZ, Comcast and their ilk (but they wont actually become an ISP – that would mean too much regulatory headaches, like the one in which they had to roll back the ETF for Nexus One under pressure from the FCC last week).