AT&T begins cannibalizing its 2G and 3G networks to boost LTE capacity


Verizon(s vz)(s vod), Sprint(s s) and T-Mobile(s tmus) are all turbo-charging their 4G networks, but AT&T won’t be left out of the party. It’s begun building a new LTE network in the PCS band, the swath of digital communications spectrum where carriers have traditionally run their 2G voice and 3G data services.

AT&T PCS LTE screenshotGigaom’s favorite mobile signal spotter Milan Milanovic spied AT&T’s new network in his NYC stomping grounds last week (Milanovic caught the first glimpse of Verizon’s massive new LTE network in NYC as well). As Milanovic’s screenshot shows, AT&T has a 5 MHz-by-5 MHz network running in the 1900 MHz PCS band — band 2 is PCS — which previously contained only HSPA and GSM signals. Commercial devices are already connecting to the network.

AT&T actually revealed its plans to cannibalize — or as they say in industry speak, refarm — portions of its networks for LTE in an August FCC filing related to Ma Bell’s planned acquisition of Leap Wireless(s leap). AT&T SVP of Network Planning and Engineering William Hogg wrote that AT&T was in the process of deploying LTE in PCS and would launch the service commercially in Baltimore, Dallas, New York City, Philadelphia, San Francisco and Washington, D.C., by the end of this year.

But AT&T began laying the groundwork for this move well over a year ago. In March of 2012, AT&T started culling all 2G-only PCS phones from its New York networks in order to make way for a shutdown of GSM in the PCS band. At the time, AT&T said it would repurpose that GSM spectrum for its 3G and 4G networks, which was a bit confusing since AT&T calls its HSPA+ service 4G. But it’s now obvious it had LTE intentions for the band as well.

New York City NYC

In September of 2012, FCC documentation revealed two hidden bands on AT&T’s version of the iPhone 5 that didn’t appear in Apple’s(s aapl) official spec sheet: PCS and cellular. At the time, no GSM operator had plans to deploy LTE in either band. Apple included them probably because AT&T asked it to.

Since then PCS and cellular support have been showing up in a lot of devices, including the current generation of iPhones and iPads, the latest Galaxy phones and tablets and the new Google(s goog) Nexus 5. The end result: as these new PCS LTE networks go live some of AT&T’s most popular phones will automatically connect to them — even phones from a generation back. It also means that AT&T almost certainly has plans to repurpose some of its 850 MHz cellular spectrum for LTE as well.

Unlike Verizon’s forthcoming 4G monster, Sprint’s planned Spark network and T-Mobile’s recent doubling of LTE bandwidth; AT&T’s new PCS LTE network won’t boost the speeds of its current networks. The 10 MHz configuration AT&T is using is only half the size of the 20 MHz network its already running in the 700 MHz band, which means it will support only half the theoretical speed. But I doubt AT&T is too worried.

AT&T's network coverage with LTE markets marked in black

AT&T’s network coverage with LTE markets marked in black

The new network will give AT&T a significant boost in overall LTE capacity. Between the two networks, AT&T will be able to support a lot more connections and maintain its current speeds even as its network grows more crowded. AT&T already holds the LTE bandwidth crown; this new network will help it keep that crown longer — at least until the other carriers’ new souped-up 4G systems come online.

As AT&T continues to connect more customers to LTE devices, its HSPA networks will less use. Then it can start adding that freed-up bandwidth to the 10 MHz it’s already using.  We’re also going to see a lot more 2G and 3G cannibalization in the coming years. T-Mobile was the first, shutting down most of its GSM network capacity to make room for LTE. Then came Sprint, which canned its old Nextel network so it could use its spectrum for Spark. Verizon has said it will start replacing CDMA with LTE in 2015.

2G and 3G networks will stick around for a while. Even as phones all migrate to LTE, the old networks will be needed in rural areas and to support machine-to-machine communications. But as carriers migrate their voice and SMS services to IP and the vast majority of traffic flows to LTE, those old networks will become scarcer and scarcer.

NYC image courtesy of Flickr user Joey Parsons



AT&T is still accepting 2G GPRS/EDGE only devices on its network, and you figure they’ll support those for at least 5 years if not longer. I don’t think they’d want to anger customers by kicking them off or forcing them to upgrade their 2G radio modules to 3G/4G LTE so quickly.

2G is going to be around a while, even if it only lives in the guard bands of WCDMA (3G).

Kevin Fitchard

Hi Willie,

I’m gonna have to disagree this time because this new network is essentially a fallback network for its main LTE system at 700 MHz. LTE is more efficient and faster than HSPA+ so why not back your LTE network up with another LTE network?

But to your last point, you’re definitely right that AT&T will have to keep some HSPA+ capacity alive for a long time for roaming, M2M and customers who don’t buy LTE phones. But it can cut down considerably on the carriers it devotes to HSPA+ once VoLTE becomes widespread and use that capacity for LTE.


I have an opportunity to upgrade from an HSPA+ -only device (N4) to an LTE device on AT&Ts network, at no cost. I probably won’t have this opportunity in the future. So for how many more years can we guess that AT&T will robustly support HSPA+ ?

Kevin Fitchard

You got plenty of time, Hortron. Your N4 will break long before AT&T shuts down a substantial portion of HSPA+


AT&T prides itself on maintaining a faster fallback network in areas LTE may not perform well or be available. They are competing against Verizon in this aspect. They have mentioned before that when LTE isnt available on Verizon, customers are limited to a slower 3G network than AT&T’s network. So as long as VoLTE isnt being massively deployed, AT&T has no real reason to kill off HSPA+, not to mention the in-bound roaming they have to support for a few more years.


Ok Kevin, so where does this put AT&T in terms of overall spectrum? Between Spark, T-Mobile, and Verizons brand spanking new #2 LTE network seems like AT&T missed a step or two

Kevin Fitchard

Hi Raymond,

If you’re asking when AT&T is going to get to 20×20 MHz carriers, the answer is not for a while. It’s build at AWS, PCS and possible cellular and has plans for WCS. None of them will get it a 40 MHz network — at least not until it starts using carrier aggregation. But I’m not so sure raw speed over a single connection really matters that much when you’re already offering 15 Mbps speeds. What’s more important is overall capacity so you can support devices at decent speeds, in my opinion at least.

Sure AT&T won’t be able to say “fastest network” when these new super-LTE networks comes online, but with the exception of Verizon’s network that could take a while.


AT&T has limited AWS now. They are only deploying that in limited areas, since a vast chunk went to T-Mobile. If AT&T wanted to keep it, they wouldn’t have put most of it up as collateral for the merger. Merger failed, T-Mobile now is stronger because of it.

Ironically, AT&T and their WCS is similar to Sprint in many regards. Both have 800 type spectrum, both have PCS 1900 spectrum and both have a high frequency spectrum 2300 for AT&T and 2500 for Sprint. AT&T has 700Mhz and Sprint does not, but the upcoming auction of 600Mhz may prove to be a winner for Sprint if they bid, putting them on par with AT&T.
Wouldn’t it be in AT&Ts best interest to keep an eye more on Sprint, to see how they are aggregating their spectrum and learning from their mistakes and triumphs, to deploy their own aggregated spectrum across the middle, high and higher frequencies? This will work with AT&Ts advantage simply because AT&T can hold off merging them together as they still have a very fast, capacity rich and fully covered area in HSPA+ and LTE that can afford to wait another year before upgrading it again. Once that is all put together, it could potentially blow Verizon out of the water, as capacity lies in the higher spectrum bands, something Verizon doesnt have. AWS 2100 doesnt count since its paired.

Kevin Fitchard

AT&T did, but it only sold the actual real estate. It’s radios are still on those towers.


Where your wrong is where the older networks are needed for Machine-to-Machine. LTE is an all IP based Network, meaning that an older system can upgraded and ultimately be more compatible with Machine-to-Machine networks. This is due to lower Ping times, High Bandwidth, and a more reliable connection on a stronger frequency.

Kevin Fitchard


You’re right about the benefits of LTE, but I’m not wrong on the resilience of old networks. An LTE module costs a hell of a lot more than a GSM module today. And there are a lot of old 2G modules in the field that carriers are committed to support for years and years. Phones get traded out every year or so, but a tracking module on a fleet truck does not. Sure they’ll eventually be replaced, but we’re talking 10 year horizons, not 18 months. Supporting those M2M applications takes minimal capacity so AT&T can shut down most of its 2G networks, but it will need to keep a modicum of GSM alive for a long time.

Comments are closed.