Internet Performance Delivered right to your inbox

A Tale of Four Carriers: AT&T, Verizon, Level (3) and Sprint

AT&T is going to buy BellSouth and the ghost of Ma Bell looms large in all of our minds. No one is really surprised by this. As usual, this deal is multi-dimensional, with landline, Internet, and cell phone assets all tied up in a complicated set of overlapping bundles. But what does this really mean for the Internet?

More specifically: Do these acquisitions really have anything to do with the Internet at all? What will be the size and scope of these networks when combined? Who will be the winners? We can offer some pretty convincing answers to all of these as well as some wild speculation about the next acquisition to come.

Level (3) (AS3356) is buying Wiltel (AS7911). Verizon (lots of ASes but mainly 19262) is buying MCI UUnet (AS701, 702 and 703, and others). SBC (AS7132) bought AT&T (AS7018, 2686 and a bunch of others), and then renamed itself AT&T. Now they’re thinking about buying BellSouth (AS6389 and a ton of other ASes). Meanwhile, Sprint has bought no one significant recently.

So there are two useful questions about this:

  1. Why are these companies buying each other?
  2. Who is going to win and why?

Since I know much more about the Internet than I do about the landline business or the wireless business, I’ll make some projections from that perspective, although those other business will have to creep in eventually.

Renesys maintains customer-base rankings for all networks on the planet. We do this by measuring the total amount of weighted network space that is reachable through a given provider combined with the degree to which that provider contributes to the global reachability of that space. Needless to say, this is a somewhat complicated calculation and it doesn’t measure traffic directly. But the results are repeatable, objective, unfalsifiable and they appear to correlate very well with the traffic sizes we see on most networks. I’m going to use these customer base rankings to analyze these most recent combinations of networks. A quick note: we maintain these rankings at various continent-, country- and city-scales but in this article, I’m only commenting on the global rankings except where explicitly indicated.

As of now, here’s how things look, in terms of the global customer-base rankings:

  1. Sprint (AS1239) – 100%
  2. Level (3) (AS3356) – 98%
  3. UUnet (AS701/2/3) – 86%
  4. AT&T (AS7018, AS2686, AS6478) – 55%

“One of these things is not like the other ones…” AT&T is dramatically smaller, from an Internet perspective, than the other three. It’s almost not even playing the same game. Given persistent rumors about AT&T’s almost-legendary unwillingness to upgrade peering circuits, even when the congestion is customer-impacting, this may not surprise some people. I wasn’t surprised that it was fourth in this ranking, but I was surprised by the size of the discrepancy.

Fine. That’s where things are now. But where will they be? To find out, we added, without any analysis or discounting, the raw scores of every network being acquired to it’s acquiror. In the real world, the combined networks would never be quite as large as the combination of the raw scores of the two networks, if only because people who are multi-homed to the two discrete networks will almost certainly not become single-homed to the new network.

For example, Covad (AS18566) currently buys from Level (3) and Wiltel. I’ll be surprised if they keep only Level (3) as a provider moving forward. The smart money would be on them picking up another provider (Sprint? Verizon? AT&T?) to pick up the slack and continue to offer their network diversity of transit. There are other reasons why 100% of customers will not move over, but for simplicity’s sake, let’s assume they do.

The combined companies would the look as follows:

  1. Level(3) (AS3356 plus Wiltel AS7911) – 100%
  2. Sprint (AS1239) – 88%
  3. Verizon (AS701/2/3 + all Verizon ASes) – 82%
  4. AT&T (AS7018, et al. + SBC AS7132 + all BellSouth ASes) – 68%

The clear winner, on a size-of-global-IP-network basis from this set of acquisitions: Level (3). They leapfrog Sprint and move solidly into the lead, assuming they can keep enough of the Wiltel customers to make them worth their while. But let’s not discount Sprint, either. With no burden of acquisition and integration, Sprint is still sitting within striking distance of first place in the world. Depending on how well (or badly) these other acquisitions get executed, Sprint could do even better. UUnet with the Verizon merger treads water or loses a little bit of ground. Something will have to change here for this to make a huge difference to them. AT&T, with the speculative addition of BellSouth, improves their Internet position somewhat, but they are still lagging.

One obvious fact emerges from this: these purchases had very little to do with the Internet. From the perspective of backbone size and synergy, the only one that makes much sense is (3) buying Wiltel. The rest of them must be explained by looking at end-user, wire-line networks, and wireless voice & data networks. Sprint bought perennial laggard Nextel because it was the only way to increase its customer base significantly in the fight with Verizon and Cingular for the US mobile business. Likewise, AT&T’s proposed purchase of Bell South is all about putting Cingular back together under a single corporate owner (Cingular is a joint venture of SBC (now AT&T) and BellSouth).

The Verizon deal is the only one with a dominant Internet theme. Verizon is leveraging the revenues and competitive position of their landline business to buy a respectable, international Internet backbone. Verizon Internet Services is reasonably well-respected in the Northeastern US, but it’s never really been much of a network to crow about (ever since Verizon relinquished its interest in Genuity (AS1) and allowed that network to be bought and integrated by Level (3)).

Even given this rosy picture, Level (3) must not rest on their laurels. As pointed out in the recent report we helped prepare for, Level (3) has an Achilles heel: Asia. Level (3) ranks extremely well in most continents, with two exceptions: they are number 8 (recently risen) in Asia and number 22 in South America. South America is not a problem, because although the Internet is growing there, the rate is not explosive and the scale will not alter global rankings (or revenues). The same cannot be said of Asia. Sprint, in particular, has a sizable lead on Level (3) in Asia (mostly because of their service to China).

Level (3) has no infrastructure to East Asia at all. The ranking they have in Asia is explained by their service to some Turkish networks (that Renesys currently classifies as ‘Asia’ although it’s really a different market) and the fact that they are the #2 provider for Asia Netcom (AS10026) behind Global Crossing. They’re probably providing that transit service from California, but it still counts as providing transit into the Asian market.

So what is a poor US-based network to do? I have a suggestion: buy Global Crossing (AS3549). Global Crossing is ranked as the #6 IP backbone in the world on our customer-base score. Adding that base to Level (3) would give them a substantial lead over other networks. But even more importantly, Global Crossing is the number 15 network in Asia and actually has network assets there. GC has a market cap of just $400m (compared to Level (3)’s $2.5b).

I’m not an equities analyst, so I can’t really say whether it make sense from a financial point of view. But from a network point of view, it is intriguing, if only on the Asia factor. The other interesting takeover targets would be (or would have been) Teleglobe (AS6453) – already bought by VSNL; BTN (AS3491); Flag (AS15412) and AboveNet (AS6461). That last one is interesting, or crazy, or both.

It’s unclear exactly what is going to come from all of these mergers. Many will certainly fail (as did so many of the WorldCom mergers back in the day). Some will succeed (as did the Level (3) purchase of Genuity’s network, and many other network mergers). And there’s always the chance that someone will come from the back to win it all. Cogent (AS174) has executed a steady strategy of buying failing ISPs, integrating them relentlessly, and selling below the cost-basis of most of the rest of the market. They are the #6 carrier in North America, #11 in Europe and growing fast. We certainly live in interesting times.

Share Now

Whois: Dyn Guest Blogs

Oracle Dyn is a pioneer in managed DNS and a leader in cloud-based infrastructure that connects users with digital content and experiences across a global internet.

To current Dyn Customers and visitors considering our Dynamic DNS product: Oracle acquired Dyn and its subsidiaries in November 2016. After June 29th, 2020, visitors to will be redirected here where you can still access your current Dyn service and purchase or start a trial of Dynamic DNS. Support for your service will continue to be available at its current site here. Sincerely, Oracle Dyn