Dave Schaeffer on leadership and operations at Cogent Communications
Avi: Hi, and welcome to Network AF. Today, I'm here with my friend Dave Schaeffer, who's built one of the leading carriers that connects us all in the digital world, Cogent. You may notice my mask is on because he is graciously hosting me here in beautiful downtown D. C. Today, we're going to talk about paths into technology, science and technology, culture, a lot of business topics around capital and growth, and focus and execution, and learn a little bit about how the company was made and where it's going, which as you'll see is a lot like where it's been. Hello, and welcome to Network AF. I'm here with my friend, Dave Schaeffer, leader of Cogent, in beautiful downtown D. C. I'm also here for the World Science Fiction Convention, but thought that we would do this episode in person. Hi, Dave. If you could introduce yourself briefly.
Dave Schaeffer: Yeah. Hi. Thanks for joining me, Avi. I'm Dave Schaeffer, Cogent's founder and CEO.
Avi: So, I always ask people to start, how did you get into networking? I know I'll go back a little bit maybe, Dave, and how did you get interested in technology?
Dave Schaeffer: Woo. You know, it should be a short answer, but in this case it's probably going to be a little bit of a long answer, so I got involved... Well, first of all in technology going to school, as I was a physics major, but in business I had the first business I ran was actually a taxi cab company and related to the taxi cab business was the need for mobile radios for dispatch. And at that time, two major changes were occurring. One, Motorola, which dominated that market, went from being a manufacturer direct distributive model to establishing dealers, which I became one of those, and then secondly the frequency bands were also being allocated on a proprietary basis as opposed to a shared basis.
Avi: Per city? Per metro?
Dave Schaeffer: You could actually own them, but you did not have to pay for them if you met certain loading requirements, and as a result I started developing a number of SMR systems, or specialized mobile radio systems that eventually got rolled up to what became Nextel. I also, from that, expanded into a second technology venture, which was paging. I was a paging carrier up and down the East Coast, including covering your market, Philadelphia, where you were based, and this was in the early'90s. Had the good fortune to sell that business while there was still a business to sell. My third technology business was based on a regulatory complexity created when the government decided to reallocate Spectrum for personal communication services or PCS, creating additional cellular carriers. So, initially there were only two licenses granted, an A and a B license. The government decided to ultimately grant four additional licenses but in doing so, when they granted those licenses, the winners in an auction of those licenses had the obligation to relocate the existing licensees who were not communications companies, but industrial companies.
Avi: Who was using the frequency, the Spectrum.
Dave Schaeffer: That's correct. And I created a business that managed that process for those companies, created excess capacity, and aggregated and built a nationwide network through that regulatory catalyst. And actually, that company raised a significant amount of capital, and actually had public debt, but attempted to go public, and markets were not in our favor, and eventually was able to sell that business. And realizing that the network that had both the lowest cost and the most ubiquity would eventually win, I became convinced in the mid'90s that the internet would be the only network that mattered, and by building a internet network, most companies overlayed either the telephone network, the cable TV network, or a mobile phone network, and it was possible by the late'90s to build a much more efficient network capitalizing on two major changes in the marketplace. First, the ability to buy dark fiber rather than build all of your network infrastructure, which traditionally was the way in which carriers built their network, taking three to four years, costing$3 to$ 5 billion, and carrying huge amounts of excess inventory. By the late'90s, you could go out and purchase dark fiber, and with advances in wave- division multiplexing, that fiber would have virtually unlimited capacity. The second key here of technological change was the advances in enterprise LAN routing. So, traditionally carrier networks were built on SONET SDH, which had a number of expensive overhead protocols that were not relevant to the internet, and by taking land equipment and then deploying it on a global scale, you could drive down the cost of network construction. So, those two events gave us the ability to build a purpose built, facilities- based network, and I think Cogent even to this day remains unique in that. So, a long winded answer to your question, Avi, but how I got here was probably not the most conventional route.
Avi: Well, no. It's fascinating to me because when I talk sometimes to... I guess I'm old enough that sometimes I say kids, but people earlier in career about different paths, and when we get started, it was all nerd. It was all technology and before the great specialty. But today, it's like how often will you find a Linux kernel bug versus how often will you find a routing bug. And the analogy I often use is physics, like the people that I had physics with used to say starting from the basic laws of physics, how fast does the moon go around whatever, and in networking there's enough bugs and complexity that sometimes you just need to get back to first principles and say what should be. And I think that's fascinating. Some people like it more than others, but it's interesting to see the physics background. Do you remember when you developed the ability to sort of look at something and say this should exist but doesn't as a business opportunity? Because it seems like the story you told, all three of those, like, " Well, there's this thing in taxi, but there's this optimization here." But this is going to be this coming problem with people on the Spectrum that need to move, or as you said regulatory. You didn't say arbitrage, but complexity or opportunity. Do you think that's something that people are born with or is it family?
Dave Schaeffer: Well, probably a little bit of both. My dad was a cab driver and I'm very lucky I never had a real job. I started seven companies and the only job I've ever had is CEO. And in my mind, that's overhead all the people that really do the work, so I'm very conscious of trying to add value when I can. And you know, I think many entrepreneurs confuse developing a product with developing a business, which needs to be self- sustaining and able to grow. And then I think the second kind of opportunity comes from looking at the market and rarely does a company invent something that no one has ever seen before. I know Steve Jobs said, " I invented because no one else has seen this before," but whether it be his I guess borrowing the concept of the mouse from Xerox PARC, to the graphic interface-
Avi: Yeah. There's a new book about that I want to read.
Dave Schaeffer: Yeah. On the first Newton. I think the reality is everybody builds off of existing technology. And to me, the opportunity comes from taking any business and being able to do it faster, cheaper, and better. If you could do those three things, typically you'll succeed. And in running a business, there are really two fundamental questions that you need to ask yourself. One, do people need what I'm selling? And then two, will they buy from me versus someone else? Once you've answered those two questions, you've got the basis for a business. It's actually the third one that's maybe the hardest one for most companies and that is can I create a profit, an economic profit? That doesn't mean gap EPS profits. It means can I produce inaudible-
Avi: Well, you said that grow and make money.
Dave Schaeffer: At the same time.
Avi: And it's not just one or the other, right?
Dave Schaeffer: That's correct. You got to be able to do both. It's the old saying, I can sell a lot of dollars if I sell them for 50 cents apiece. So, it all starts with fundamental demand, your ability to meet that demand and differentiate yourself from others, and then the final question after you check those two boxes is is this self- sustaining. And it's why so many venture- backed companies never reach kind of escape velocity. They typically get bought by other companies because they're really a product or they're not self- sustaining. So, when I looked at the idea to build an internet provider, it seemed like the market was huge, the differentiation was substantial enough to give people a reason to buy from us, and then third, the economics would eventually work as long as we could reach scale. In building Cogent, we also had to be disciplined about where we deployed the network, so in any utility, and the internet is a utility just like any other, there's a natural monopoly. There should only be one, but the reality is for the internet, there's already three there. There's this overlay on top of the phone network, an overlay on top of the mobile network, and the more successful networks tended to be the overlays on top of cable. So, you had to say was my network good enough that I could overcome the disadvantage of not being an incumbent and having the system planned. And in our case, the way we did that is being very thoughtful about the end points on the network, effectively cream skimming the market, going after the highest traffic locations, and those really fall into two major categories. We connect just under one billion square feet of office space in North America where we thrive, 60% of our revenues, and that is as an end user ISP, but the average building we serve is 550, 000 feet, or the equivalent of about 12 football fields inaudible-
Avi: So, you wouldn't serve your own building here.
Dave Schaeffer: We actually do because we're here, but yes, this would not be a Cogent building if it was not for us physically occupying the building. And then secondly, we serve about 1, 350 carrier neutral data centers in 50 countries around the world that are really supermarkets for bandwidth. And the important reason for our success is we ignore the millions of other buildings that we could serve, but we concluded that we can't generate a high enough return on capital when we factor in our sales and marketing costs. So, we're very select about where we go.
Avi: You said a few different things in the last 20 minutes to unpack. One of the things that I struggle with is focus, and one of the things that I've enjoyed working with Cogent is seeing the focus, because I often... I maybe see too many opportunities out there and actually had some good friends before I started Kentik say, " Maybe you should just pick one. If you want to make an Akamai or a Cogent- sized thing, maybe you should not try to do five things at once." But is that a struggle? Do you see these other opportunities and say, " Well, all these other carriers think that they need to get out of capex jail by making managed services," or salespeople come in and say, " Hey, but we could sell this, or resell this." Is there any temptation to break the model that's worked so well?
Dave Schaeffer: No. Like Oscar Wilde said, I can resist everything but temptation. And you do need to be disciplined. And it's one of the biggest challenges for an entrepreneur because you start with a blank sheet of paper. You can literally do anything you want. Capital is fungible. And most businesses change course multiple times in their evolution. Actually, of the seven companies I've run, Cogent's actually the only one that's stuck to its original business model and did not pivot in midcourse. The problem is also exacerbated when you take venture capital because you then have a board that is very focused on the next new thing, so they're constantly looking for another shiny object and distracting the entrepreneur. Now, on the flip side, the entrepreneur needs to be able to stay focused, but when things aren't working, know when to pivot.
Avi: Don't ignore reality.
Dave Schaeffer: Yeah.
Avi: If you're getting the messages from God, it's like the person that is like, " Why couldn't you help me?" And he's like, " You could buy a lottery ticket."
Dave Schaeffer: You know, maybe it's too simpleminded, but I take the approach that when you go out to raise capital, the people you talk to are in the business of investing that capital, and if you talk to 50 people who really want to put capital to work and they all say it's a dumb idea, it probably is a dumb idea. I understand entrepreneurs need to be determined, they need to be focused, they need to have the tenacity to plow through no, and I agree with all that. But once enough people tell you something doesn't work, you probably should try to pivot.
Avi: Now, to be fair, I remember the telecom days, and hearing your pitch for the vision, and we'll come back and talk about it. It was crazy times when sometimes timing favors your plans, too, with what was happening macro. But I thought you were a little crazy at first, and so I say that with respect, and pleasure at seeing what Cogent has done. I think that I won't defend at least some of the VCs in this hour, but I think that there are some people that definitely get caught up, and it's true that VCs keep seeing the new hotness, the new hotness, the new hotness, but there's some folks that focus on the fundamentals. Now, it is true that in the private market today, most people would rather have you ignore profitability for growth. That's generally true in VC and right now even gross margin, as you said, selling the dollars for 50 cents, so it'll be interesting to see what happens. I'm glad for all my friends in the private and public markets who are taking advantage of this market, but it is a little crazy when I do the math trying to figure out how everything came together. So, when you decided to start Cogent, you did raise some outside money? How much and what year was that?
Dave Schaeffer: Okay, so most of my career I bootstrapped my companies. When I started my previous company, we needed the credibility.
Avi: The Spectrum arbitrage?
Dave Schaeffer: Yeah, the arbitrage business was called Pathnet. When I started Pathnet, I bootstrapped it. I quickly realized in dealing with major companies, Fortune 100 companies, that they would not-
Dave Schaeffer: ... dobusiness with a non- institutionally funded business, so I brought in external capital and grew that business by using other people's money. When I sold my position in that business, put in a management team, and then literally came in here... Well, came into our old office three blocks from here-
Avi: Right. I've been there.
Dave Schaeffer: ...on a Saturday afternoon I wrote the entire Cogent business plan that afternoon in August of'99 and realized that we needed to raise capital. Now, if you can remember back, the world was on fire much like it is today. We were actually a unicorn before companies were called unicorns. We raised ultimately about$ 500 million in our seed round.
Avi: We're about there now again. We're about there now again.
Dave Schaeffer: Off of a PowerPoint and at the time, Cogent was myself, and I was paying two 24- year- olds out of my pocket. One was an ex- banker from Morgan Stanley and the other one of them was from Anderson Consulting, now Accenture, who both wanted to work in a startup before going to business school. And we had told our investors that to build the network that we contemplated it would take about$2 billion. That$ 500 million was the down payment. Sometimes it's better to be lucky than smart. We raised the money. We started to hire a team and couldn't spend the money fast enough. The telecom market then imploded. It was a worse depression in technology than the Great Recession was in the financial industry.
Avi: Yeah. If Akamai hadn't raised a billion dollars in debt, it would have gone-
Dave Schaeffer: It wouldn't have survived.
Avi: It wouldn't have survived. Yeah.
Dave Schaeffer: And that's true of virtually every-
Avi: We got lucky. Again, we got lucky. There was arguments about whether-
Dave Schaeffer: Listen, Amazon raised $600 million in a convert right before the market blew up and wouldn't have survived if it hadn't. We had effectively a pool of capital, a good business plan, and a very small team. We sat down with our investor base and we literally laid out three paths forward. One was liquidation and dividend out the cash that we had raised minus what we had already spent, which was a small fraction of the money. Only about 10%. The second option would have been for the current investor base to pledge incremental capital because there was no external sources available. And then the third path, which is what we debated and ultimately chose, was to use our capital to buy broken businesses. So, a common mistake that many outsiders think of when they think of Cogent is they think we were founded to buy up distressed assets. That's actually not true. We were founded to build an all IP over DWDM network, protected at layer 3, using ethernet as an interface, serving the highest traffic locations.
Avi: You trusted that price would be the disruptive factor that would ensure, as you said, scale. Because to do that work, you need$ 2 billion, but it only works, and it's profitable with growth, but only at scale.
Dave Schaeffer: That's correct.
Avi: So, how do you get to scale?
Dave Schaeffer: Telecom is absolutely a business of scale, so in many industries, there's the concept of a backward bending supply curve. So, there are discontinuities where you have effectively diseconomies of scale. Marginal cost is always below average cost in telecom, so you had to get big fast. And we thought we had the right addressable markets. And in serving our customers, there were kind of two different ways to show value. For corporate customers, it was not price. It was quality of service for price parity. So, we went to the market offering in 2000 a 100 megabit, non- blocked, non- oversubscribed DIA product, on a month- to- month timeframe-
Avi: Direct internet access for those.
Dave Schaeffer: ...to that customer base for$1, 000 when a T1 was$2, 500. So, you paid less, you got more throughput. It was rain protected. The building was prewired, so it typically installed in one ninth the time. It was three times more reliable. And you got somewhere between 30 and 60 times the throughput for the same price point.
Avi: And no bills delivered on UPS trucks that you can't make out.
Dave Schaeffer: None of that. It's just literally plugging in with an RJ45 to a fast ethernet port and that model worked well for our corporate customers. Our corporate business generates only a small fraction, a couple percent of our total traffic, even though it's 60% of revenues. We have a second market segment that is totally different, which is selling bulk transit and data centers. In that market, there's only one dimension you compete on and that is price. We came to market at$ 10 a megabit when the average price for transit in a data center was$ 300 a megabit.
Avi: And that had recently been higher.
Dave Schaeffer: Yeah. It had been as high as$ 5, 000 a megabit. And we started at 10. Today, we sell sub 10 cents for many customers. The prices are going to keep coming down.
Avi: Do you have a prediction? How much per year, per decade, per five years?
Dave Schaeffer: So, over the past 20 years for Cogent, the average price we sell at has fallen at about 23% per year.
Avi: On the wholesale side or direct?
Dave Schaeffer: Wholesale side. So, this is measured services, metered, in a data center, where you're buying by the megabit. Prices are falling at 23% a year. The two technologies that allow that to occur are wave- division multiplexing and optically interfaced routing. If you look at wave- division multiplexing, that technology has improved since the mid'80s when it was first deployed. Actually, at the Lake Placid Winter Olympics was the first deployment of a test, commercial test of a four channel asynchronous wave- division multiplex system at roughly 80% per year compounded. Now, put that against Moore's Law, which is only 55% per year price performance improvement, and different than Moore's Law, we're much further from the physical limitations.
Avi: The limits. Right.
Dave Schaeffer: So, I think the transport side will continue to improve at this very rapid pace for decades to come. The second technology is an optically interfaced router. So, prior to Cogent, routers had to be front ended, usually with an ATM switch, so you had a Sonnet box, you had an optical transport terminal that talked to a Sonnet box, that talked to an ATM switch, that then talked to the router.
Avi: Right. That inaudible frame or whatever. Yeah.
Dave Schaeffer: By stripping all that complexity out and going directly in a LAN type architecture to IP directly over DWDM, we could ride the price decline curves in those routers. Now, the routing market has never been overly competitive. It's always been effectively a duopoly. And the price per packet forward for the past 30 years has declined at roughly 40% per year, so it's-
Avi: There are a couple discontinuities there around inaudible 12.
Dave Schaeffer: That's right.
Avi: Before everything went straight ethernet.
Dave Schaeffer: That's right. There's been step function, but if you just graph that out over time-
Avi: Right. But the last five years have been awesome. Yeah.
Dave Schaeffer: ...and I think we'll continue to see those types of improvements. If you look at the basics that are used in routers, they're typically still built on 19 nanometer traces. Today, new microprocessor fabs are being built on five or three nanometer traces. Today, seven is relatively common in the market. So, there's a huge amount of catch up that routing has to do to computing.
Avi: It's interesting because in the'90s I used to say the router would have whatever a four- year- old Macintosh would have. You're right, it's actually much more. It's stretched much more versus inaudible and things like that inaudible.
Dave Schaeffer: Yeah. And I think we will see a continuous improvement in routing technology for decades to come. So, with that, as long as the market for transit remains competitive, and yes, the competitive landscape has shrunk, other carriers have de- emphasized the product, but there's still enough diversity of suppliers that there's competitive tension. Prices should continue to decline at 23% per year. And we'll be talking about fractions of a penny a megabit probably over the next decade.
Avi: Yeah. No, it's definitely been interesting to see and we work in and with a lot of these folks, and everyone has their different take. What's fascinating about Cogent is probably the highest terabit per I'll say backbone architect of any company that I have seen, and I guess is there a secret to running lean effective teams? Is it that focus, that you don't bother them with other features? Is it something about the culture? What is it that enables such a tight inaudible lean team?
Dave Schaeffer: I think there's two major messages. One, you have to be relentless in driving down the cost of an interface routed bit mile. So, an internet service provider sells internet access, but the reality is we produce routed bit miles connected to other networks. That's our factory. So, everything we do is designed to drive that cost down. Two, it's about standardization. So, our network is about 60,000 route miles terrestrially, spanning every continent. In addition to the long haul network, there's about another 17,000 route miles, about 40,000 fiber miles of metropolitan fiber in 215 markets, about 1, 025 rings. Between those locations, we have just under 1, 000, about 995 optical transport nodes, and if you walked in to an amplifier site in the Ukraine, in Finland, in Sydney, or in Wichita, Kansas, the rack faces, the diagrams would look identical. It's all about standardization.
Avi: Is it all one single vendor?
Dave Schaeffer: One vendor. We are end to end a Cisco shop for routing and transport. Now, Cisco has never been the absolute best in any of those technologies, but the benefit of getting an additional half a generation of technology advance is eclipsed by the savings you get through standardization and operational simplicity. So, there's a lot of do it once, replicate it at Cogent, as opposed to bespoke design, and the product focus also is a big part of this. We basically sell three things. We sell internet connectivity, we sell VPN services on top of the internet, and we sell space and power in our 54 data centers that we operate. And that's it.
Avi: I didn't realize you had that many.
Dave Schaeffer: Yeah. 54.
Avi: I've been to the old PSINet one in-
Dave Schaeffer: Herndon?
Avi: ...Herndon. Yeah.
Dave Schaeffer: Now, our biggest one is actually the old Earthlink facility in Pasadena.
Avi: Oh, interesting. I didn't know that went up with you.
Dave Schaeffer: Yep. That's I think our largest. We have a pretty big one, I know you've recently relocated to Seattle. We have a big facility in Seattle. We have them in New York. We have them in Madrid. We have them in Paris. We have them in Frankfurt. Literally, around the world. We operate about 600, 000 feet of rays for space in those 54 centers with about 70 megawatts of power.
Avi: So, you're not competing for the Bitcoin mining market.
Dave Schaeffer: No.
Avi: People tend to want to be connected. It's to enable people to connect things to the network and take transit.
Dave Schaeffer: Yeah. It's for DR, it's for corporate customers who want to put a backup facility, but don't want to go into a carrier neutral at higher prices. Could be for some of our wholesale customers who look to colocate directly adjacent with our network. And these facilities also house the majority of our hubs, so we have roughly 191 Cogent central offices in those markets.
Avi: Like they're super POPs?
Dave Schaeffer: Yeah. They're POPs. Basically, they're the intersection of the backbone and multiple rings of fiber in a metro-
Avi: To go to the metro. Right.
Dave Schaeffer: Yeah.
Avi: So, you'll use the same ring for I'll call it backbone as you do for access to the buildings.
Dave Schaeffer: Absolutely.
Avi: But it's all wave, so it's not like it's stealing bandwidth from anything. You just get a better prism and you've got more bandwidth.
Dave Schaeffer: So, one of the innovations that we did was our network is entirely ring architected, so there's no single points of failure, and we used wave- division multiplexing to allow each building to be a logical hub and spoke, but on a physical ring, so you get the best of both worlds.
Avi: So, you don't have to break it out at the IP layer in that building.
Dave Schaeffer: That's correct. Basically, it's on a wavelength. Each building gets its own pair of wavelengths in two directions, both clockwise and counterclockwise, and if a fiber cut occurs, the restoration occurs through IP fast reroute, because we're able to redirect the traffic.
Avi: And you're still up to both devices in that POP. They each have their own wavelength. inaudible right.
Dave Schaeffer: That's correct. And they have them in each direction, so the network doesn't know that you're using a wavelength in the clockwise direction, but you're also using it in the counterclockwise.
Avi: You don't have any fancy Lemer glass, robots, fancy mechanical optical switches, et cetera?
Dave Schaeffer: No.
Avi: I remember when some people called it... I won't say the acronym, but it wasn't only CWBM course, but passive, and I remember trading a wave with a big web company that starts with a G back when I was doing packet traffic 1. 0, because they couldn't... They had put this off of the WDM equipment in, and it was going to take... I don't know, like 10KW. But it's optics. Why are you taking 10KW for prism?
Dave Schaeffer: Well, they were running inaudible-
Avi: Right, right. crosstalk Well, it's as you said. It's all the OC stuff. It's all the protective protocols and all that. But yeah, you overbuy. Yeah, if you're going to space, you have to be really careful, and maybe you want different protocols. But there's enough fiber. Or as Dave Rand at AboveNet said, QS should be quantity of service, right?
Dave Schaeffer: And that's always our model. There is no QoS prioritization.
Avi: Queuing. Yeah.
Dave Schaeffer: You throw bandwidth at every problem and all the problems disappear.
Avi: Yeah. No, it's funny, because I remember when I came to AboveNet, Dave was like, " Someone wants an OC, all we deliver is ethernet. Someone wants an OC, they can put it in their rack, but our endpoint is the ethernet." And then you took it one step further, like make ethernet come to the customer.
Dave Schaeffer: That's correct.
Avi: Yeah. Which is pretty cool. It's funny, with a network like yours, there must be something down all the time. Is there ever a time when no circuit is down anywhere?
Dave Schaeffer: So, with the breadth of network, we will typically have two to three fiber cuts terrestrially every day somewhere in the world. I mean, backhoes just seem to have a-
Avi: Backhoe, backhoe, digging deep. Make the backhoes go to sleep.
Dave Schaeffer: They find fiber. It's almost like they've got a divining rod directly to the fiber. But the ring architecture gives us a lot of resiliency. We've actually never had a backbone outage in 20 years of operation. We're very cautious about rolling out new code.
Dave Schaeffer: We have, again, a lot of redundancy built in the network. And a simple product stack. But yeah, listen, endpoint customers can go down. Typically, we're running about. 03 of a percent of port seconds that are down on our network in a given week. We actually monitor that, meet on it every week to try to improve it. And almost all of those typically are the customer, prem equipment either failing or someone unplugging it. I'm not saying we never have an outage, someone doesn't fat finger a config, but for the most part the network is very, very resilient.
Avi: No, we've seen that, as well. I mean, really, which I guess gets to the next topic, in the'90s if COVID had hit and we had tried to all ship each other webcams and use the internet, I'll just say it wouldn't have worked very well.
Dave Schaeffer: It would have failed. It was too much massive oversubscription. Remember, our local network is non- oversubscribed and non- blocked, so there is no statistical multiplexing in our-
Avi: But you don't peer with people at every... I mean, some people you peer more widely, but you don't peer in every metro with.
Dave Schaeffer: So, most of our interconnection is actually customer to customer. So, Cogent today runs BGP sessions with about 7, 600 ASs. I don't think there's anybody who's ever even approached that number. We're running roughly 25, 000 sessions.
Avi: When the internet gives you a customer, you peer with many networks.
Dave Schaeffer: So, when Cogent got started, we obviously had de minimis market share, and very beginning we had to buy upstream, and we were supplementing that with a very open peering policy. Today, we peer with less than 25 networks globally.
Avi: So, you flipped it from open to be my customer if you're not my peer.
Dave Schaeffer: Well, I would say that the strategy has stayed the same. The difference has been the bar to be our peer has gone up as our scale has gone up.
Avi: As your network has gone. Right.
Dave Schaeffer: And you know, I think we are the most interconnected network, so obviously we're running BGP sessions in over 1, 350 endpoint locations. Those data centers. And we're pretty flexible on where we'll peer. We can actually peer at any one of those facilities if someone wants to peer or we sell transit, which is in fact how most of our interconnection occurs through that.
Avi: Right. But there is going to be some, there is some oversubscription getting between networks, but it's 50X less? I don't know the exact number that it was in the'90s. I mean, when I started at Akamai, I remember asking the Zen question. If all we're doing is optimizing around the congested paths, then at some point will we have done all the optimization? The internet will suck equally for everybody. But I think we just got fortunate that one, things have evolved, and as you say, the costs have come down so much, and two, most of the networks that I've seen and work with had ordered enough in advance when COVID hit that even before the supply chain stuff started getting bad, they had either built or they had equipment ready, so it was enough capacity prebuilt that when more things went online, things did okay. So, how has that been for Cogent?
Dave Schaeffer: Let me address the oversubscription piece.
Avi: Oh, sure.
Dave Schaeffer: So, clearly we don't dictate to our customers what size ports they buy, but we price our service in such a way that most of our customers generally utilize 20 or 30% of their port capacity in kind of an average.
Avi: Right. So, you're willing to give people that kind of-
Dave Schaeffer: That's correct.
Avi: You can't buy 5 megabits and get a 100 gigabit port. Right.
Dave Schaeffer: Yeah. We try to be very flexible. If we have in excess of a exabit of capacity to other networks, so very, very broad reach for the geography and in scale, the 25 networks we peer with, because these are non- monetary agreements, they tend to be stingier than we are in terms of-
Avi: I saw that AboveNet, too. We were like, "Well, why don't we just do 10X, not 2X the capacity?" And they're like, " No, that's okay. We're good. We'll just do 2X."
Dave Schaeffer: Yeah. They kind of want to ration out bandwidth. Usually because their networks can't support it. And at the end of the day, I think what most telecom companies and cable companies won't admit is the internet is their worst nightmare. It drives up the load on their network by orders of magnitude and concurrently drives revenues down by orders of magnitude. So, if you were a phone company, you love selling 64 kilobit POP circuit that people used basically 1, 400 minutes a month at 64 kilobits and life was great. You send them a bill for 25 bucks and they had no choice, and you were great. Where conversely, if you're a cable company, you would take a six megahertz channel, deliver video over it, it was shared between all of the end users asymmetrically, and you charged them 80 bucks a month and life was good. Now, the customer is demanding hundreds of megabits of end user capacity.
Avi: Oh yeah. And the content is moving to the internet instead of necessarily from your network, where-
Dave Schaeffer: The revenues are going down while your capital expense is going up. So, legacy providers who still dominate the internet, whether they be telephone, cable, or mobile phone, all hate the internet. In other words, if John Stankey could go to bed tonight at AT& T and wake up in the morning and the internet was gone, he'd be a happy man.
Avi: I would say that, I'm not going to name names, some non- eyeball carriers do still have that let's just double the ports and not 4X to 10X. Some eyeball networks do better than others. Or the IP engineering people, I had Elliot Noss on and we had to edit out a few things that were said about Tucows and his experience too, so sometimes the IP engineering team is doing the right thing, but I understand what you're saying.
Dave Schaeffer: I'm talking about the corporate.
Avi: Right. If you took a rational person and made them the CEO of the company, and they look at the inputs and say, " What helps my company and what is a threat?" Then yes.
Dave Schaeffer: Right. But if you built the right network and you built it around the products that people want, you can actually make money as an ISP.
Avi: Well, I think that they're... Well, so I had Elliot Noss on and I was like, " Why?" He called me, I don't know, 2016, and was like, " I want to start an ISP. Do you want to help?" I'm like, " Who bit you? What got into you?" He's like, " No, no. We're going to make money. It's a mitzvah and we're going to make money, and why not?" And for him, he's been working on solving billing challenges, which is one of the problems that you solved early on, but it's still a huge problem for a lot of companies. Yeah, I would say some, again, some of the incumbents, some of those folks do have very modern networks and IP sits underneath, and then some of them are straddled with fiber plants where they've got DB loss from side to side because there's 40 splices. Yeah. It's an interesting market. And also, as we talked about, how many products do most of those companies have? They've got consumer, they've got business, they got many different products, and again, it can be hard to navigate. And it's interesting working with companies that have focus versus working with broader companies and seeing that. So, with COVID, how did that affect I guess I'll say the people culture side and the business side?
Dave Schaeffer: I'll start with our Cogent team. We put their safety first. We have about 1, 100 employees. In March of 2020, they all worked remotely. And they continued through October of'21 to be remote in the U. S. and in some parts of the world they are still remote. As vaccines became available, we implemented a mandatory-
Dave Schaeffer: ... mandatoryvaccine policy. And we have required employees to come back because there is a huge benefit for people collaborating face to face and the culture of the company is a interesting mix of mature processes with startup drive. That's really what we try to maintain. In our customer bases, there's really been two very different tales from COVID. For our corporate customers, demand fell off.
Avi: The access, building, using the offices.
Dave Schaeffer: They don't know what the future is going to look like. They know their employees are working from home. They upgraded their VPN concentration ports. But they're not willing to make massive rearchitecturing of their businesses and networks, particularly migrate away from MPLS to a SD- WAN or VPLS solution until they have clarity on what the new work environment's going to look like. For our wholesale business, we saw an acceleration because people stayed home, they have more time, and they streamed more, so we have thousands and thousands of access networks around the world that buy upstream. Their consumers were spending more hours a day. They needed more bits.
Avi: We've seen that in the Kentik data also. Yeah.
Dave Schaeffer: Yeah. And we've been accommodating that. Over 100 PTTs around the world, for example, buy their upstream from Cogent. And then on the content side, the streamers, and the application providers are seeing unprecedented demand. There has been a wave of new entrants, whether it's Disney, or Amazon, that are broadening out the streaming market and that has helped that business. It's actually... We've had the best two quarters over the last two quarters in the company's history in that hosting side.
Avi: I haven't read your quarterly filings or the transcripts, but you consider that the same business, the access business for service providers and the content side? That's all wholesale?
Dave Schaeffer: All wholesale.
Avi: Or do you break those out? Okay.
Dave Schaeffer: We treat it inaudible-
Avi: Okay, so it is that. Okay.
Dave Schaeffer: We coined the term netcentric and it just means we're selling bulk bandwidth to someone who's reselling our bandwidth, either as an access provider or with the content.
Avi: Companies were sneakernet was never an option.
Dave Schaeffer: That's correct.
Avi: Right. Their network is their business. Delivers the revenue. We have similarly, we have traditional enterprise and then we have netcentric, as well.
Dave Schaeffer: Yeah. That's a logical dichotomy.
Avi: Yeah. Yeah. That makes sense. So, what was the reaction to mandatory vaccination?
Dave Schaeffer: You know, we lost a few employees. Not many. A few. We had a lot of counseling to do, a lot of cajoling, and it's a combination of making sure coworkers are safe, making sure the work environment is safe, and quite honestly making sure our employees are safe, and it's a balancing act between people's personal liberties and public health safety. And we also-
Avi: We are in a society.
Dave Schaeffer: Yeah. We're social animals. And we also had to weigh the benefits of getting people together versus the isolation of being on our own, and we have a big sales organization. That sales organization typically has churn. Salespeople come and go. Some of them very quantitative, some succeed, some do not. And to train new salespeople, even inaudible a very robust online classroom training materials to be done remotely, there's no substitute for a mentor on the next desk, so we just had to get people back to support that growth of the sales organization.
Avi: I'm told that I shouldn't say osmotic learning because people don't know what I'm talking about, but it is funny-
Dave Schaeffer: That it just absorbs into.
Avi: Yeah. It's funny how I try not to use dictionary words. CEO, you need to speak simply, and clearly, and everyone can understand you, but yeah, it's interesting in networking how... In many professions, how much we are not that far from the middle ages of padawans and apprentices. And not about power structure and do my work for me for free, but know it is known. In the Game of Thrones, " Oh, it is known. Did you not know that? Did you know that that network doesn't have that? Or that if you turn on that protocol that way, there's that problem?" And there's ways of doing it constructively, which is not like, " You are an idiot," but more like, " Hey, one might think this, but actually..." So, yeah, that's definitely interesting.
Dave Schaeffer: I'll tell a story. I remember early in my coaching I had some exceedingly talented routing engineers who knew far more about routing than I did. I consider myself an amateur hacker. And you know, maybe as more of a classically trained person I came into a problem and white boarded the Konigsberg Bridge Problem. They didn't know what the hell I was talking about. But they knew how to fix the problem. So, I'm thinking of it more the way Euler thought of the problem. They're thinking about it like, " Oh, I've seen this, and this is how I need to..."
Avi: Right. Well, there is a lot of pattern matching, but that reminds me of a story. Probably why I don't teach. I had a great teacher who was a doctor, who got into image processing, and when I went to Temple University, the first day I met this guy Paul Follett, and I might Bob Stafford who did networking, and I was like, " Ooh, AI image processing. That's the really interesting stuff. Networking, yeah, yeah." And then of course I got into networking more, and so he taught a class on realtime systems, and he asked a question on the exam why do tires appear to be going backwards. Now, we had studied mind twists and sampling rates and all that, and so, but a lot of the people in the class thought it was unfair that he'd asked a question that we didn't... We never talked about tires, right? And he was asking to make the leap from computers to brains and your eyes sampling, and so sometimes that from first principles, again, the classic physics education of let's reason and understand is at war with the-
Dave Schaeffer: Practical, real world concerns.
Avi: Well, yeah, or as again, I talk with academics, and I was like, " You have to understand the internet is a bunch of garage mechanics banging on routers with wrenches." And like yes, there is science here, but it is not primarily science. It is definitely like I want to do car talk for the internet. It's like what sound is your router making. But you sort of see that, right? It's like you log in, or API, whatever you do, and this thing is like that. You're like, " Aha. I think that's something we need to formalize not just because of COVID, yet for us as we strive to be even more diverse, the most diverse set of folks is early in career." And yeah, how do you get those people and how do you train them, both about networking and then what level. Right now, for Kentik we are... We gave up our office because we couldn't use it when the term was up, and we're doing a lot of collaboration, but we'll see how that goes. What percentage of the company is local and you said, you mentioned you have multiple offices.
Dave Schaeffer: So, we actually have 40 offices around the world, mostly sales oriented. We also have about 90 people in our field services organization that either can work in our office or some are so remote they work from home. But here in D. C., there's about 200 of us out of the 1, 100. Probably another 100 or so in Northern Virginia, so about 300 of the 1,100 is in this area, and the rest of the company is truly global. From Singapore, to Stockholm, to Seattle, to L. A.
Avi: I won't make you name them all or ask when's the last time you visited all the offices.
Dave Schaeffer: Yeah. My goal pre- COVID was to get to every office once a year.
Avi: Oh, every year.
Dave Schaeffer: Every year. And I did pretty well. I would typically get about 35 of the 40 done in a year.
Dave Schaeffer: That was pretty much my pace for the past decade. And then COVID hit and I haven't been to an office outside of D. C. in two years.
Avi: I just last week was the first week I was at home for three months. It felt like three years ago. And it started in the middle of that was KubeCon, which was mostly vacant. Very safe, with lots of space between spaces, between booths, and then a couple weeks ago in Vegas for re: Invent it was almost full. Even as Omicron was firing up. So, it's interesting. inaudible has type- 1 diabetes, so we try to stay safe, and if people are coughing on each other stay away, and we have to decide whether we want to go to the masquerade at the World Science Fiction convention with all the people.
Dave Schaeffer: And we have pretty strict protocols here in the office and really do try to keep people safe.
Avi: So, even with mandatory vaccination, people still mask in the office?
Dave Schaeffer: Absolutely. If you're in an open space or if you're walking around, you have a mask. If you have a private office with a door on it, you can take your mask off while at your desk. Absent that, masks all the time.
Avi: Awesome. I can tell you that there are definitely... We're do an off site in January and then Omicron hit. We're still planning to do it. Number one, as we talk to people, they want to feel safe, be comfortable that they're protecting that. So, two last questions. What percentage of Cogent hires are I would say early in career? You mentioned sales and engineering. Do you prefer to hire senior people and bring them in, or is there a thought... Do you try to weight one way or the other?
Dave Schaeffer: It's kind of a tale of two different workforces. Cogent's operations team have average tenure of about 12 years.
Avi: Right. I'd heard longer. All the people I know have longer.
Dave Schaeffer: Yeah. I know people that have been here 20 years. We have very little turnover on the operations side, so we just don't do a lot of hiring. Occasionally people retire, someone moves up, and then we'll bring in young engineer-
Avi: Is the NOC a traditional place for technologists to come in?
Dave Schaeffer: Typically, we're looking for people who have interests but not necessarily the exact skills we need and we'll train them. And with some kind of engineering computer science background. But now on the sales side, it's a very different sort. There, I would say the average hire is in their late twenties. We hire some older people. We have huge turnover, 70%, 80% a year in the sales force. Now, again, we have salespeople-
Avi: I know a lot of people that started their career in sales at Cogent.
Dave Schaeffer: Yeah. But you know, it's sales is brutal. It's very quantitative. And it's very demanding. Cold calling, which is how our sales model is organized is probably the hardest job you can imagine.
Avi: Recruiting, number two. Sales, number one.
Dave Schaeffer: I think that's right.
Dave Schaeffer: And for that reason, people say they can do it, train them, and few can, and it's very hard, so there the average hire is probably late twenties. Second or third job selling, but not necessarily-
Avi: So, not right out of college.
Dave Schaeffer: Not right out. We do hire right out of college. Generally not. It's usually you sold cars, you sold insurance, you sold... I've had people who sold cemetery plots. We've had some people sell life insurance, medical supplies.
Avi: Something. Yeah.
Dave Schaeffer: Or technology. And then we train the technology component of it and some succeed an some don't. But on the operations side, we have a very stable workforce.
Avi: Okay. Yeah. No, I've definitely seen that, but it's as we get larger, that's just something we've been thinking about is what is it that it takes to train people as we started, and we just had a big network of people that had a lot of experience, but if you hire from your network, then you don't have diversity of thought, and opinion, and wisdom, and backgrounds, so we're working on that. I guess the last question, if you could go back to taxi Dave, or earlier on, any lessons or advice if you could do so?
Dave Schaeffer: I think everyone kind of has to find their own path. I think I'm often asked by salespeople what makes me successful. And your basic life skills. It's the tenacity to stick with something. It's the flexibility to change when you need to change. It's organizational skills. Time is your most precious asset and you have to organize your time, you have to set goals, and you don't always reach them. I start every day with still a white notepad with a to- do list.
Avi: I'd say I won't take a picture of it, but I see it. Yes.
Dave Schaeffer: Yeah. And you know, I've got literally decades of old notepads that I've worn through and-
Avi: I used to use index cards and now I use VI inaudible-
Dave Schaeffer: I know I could be more high tech than that, but it's old fashioned. And I just tell people manage your time, focus on what you want to get accomplished, set realistic goals, and success will come. It's that simple.
Avi: Cool. Well, thank you very much.
Dave Schaeffer: Thank you for hosting, Avi.
Avi: Thank you for sharing. Absolutely. Well, thank you for physically hosting me. I know that Cogent's story is fascinating and you've got a lot of customers out there, and people want to look at Cogent, it's cogent. com. Any other ways, are you with the social?
Dave Schaeffer: I'm not a very social guy. People can reach out. I talk to customers all the time. inaudible
Avi: Okay, so being a customer, the best way to get connected. Well, thank you very much. It's been a real pleasure.
Dave Schaeffer: Thank you very much, Avi.
Dave Schaeffer, founder and CEO of Cogent Communications, discusses his path in networking and the innovations that have propelled their ISP to success. In this episode Avi and Dave touch on:
- The differences between developing a product and a business
- Focusing on the product and knowing when to pivot to new ideas
- Raising investor money and the dynamics it creates
- The decreasing costs of internet service and Cogent's role in creating affordable access
- How COVID-19 has impacted culture and operations at Cogent
- Hiring practices when looking for early career vs senior employees
- Dave's advice for his younger self