The Holes in Netflix's Business Model

| About: Netflix, Inc. (NFLX)
This article is now exclusive for PRO subscribers.

Given the ionosphere-like valuation of this 1999-style darling of late, I thought I'd dig a bit into Netflix's (NASDAQ:NFLX) report and hype, and add a bit of cold water - not that I expect anything other than derision for doing so. (See Q3 earnings conference call transcript here.)

Let's start with the so-called "subscriber growth." It was impressive last quarter. Unfortunately, so are the free month trial numbers. And what nobody seems to be talking about is that the company went from a two week to a month-long "free trial" not long ago.

Well, that's a problem. Why would you extend your promotional reach - and cost - unless you detected a loss of momentum? You wouldn't. In fact if growth was exceeding management expectations you'd expect a retraction of the freebies, as they're an unnecessary cost. That didn't happen.

What we do not know is how conversion is holding up on a forward basis. And that can turn into a big problem - does anyone remember what happened with AOL and all their "Freebies"?

But that's not the big problem I see with the company and it's forward prospects. No, the real problem lies here:

Netflix represents more than 20% of downstream Internet traffic during peak times in the U.S. -- and is heaviest in the primetime hours of 8 to 10 p.m., according to a new report from bandwidth management equipment vendor Sandvine.

That's going to turn into a major issue.

Here's the deal. Contrary to popular belief, "unlimited" internet in the home isn't really sold as unlimited. I know, I know, that's what you were promised, and like most good Americans, you're going to sue if you don't get it.

That's ok. The language will shortly be changing, just as it has for mobile network users, where "unlimited" is basically gone now, and virtually all data plans have either a soft or hard cap.

And with good reason.

What nobody understands about the internet in the consumer space (but those of us who have run internet companies, including me, do) is that consumer connections are not sold with the expectation that you will hammer the link on a continual basis. That is, when you buy a 20Mbps downstream connection, the company does not have an expectation that you will pull 20Mbps any material part of the time. They expect the speed will provide a higher-quality experience, but not that you will use it to replace your cable television link and stream data to three televisions in the house - watching different things.

Let's look at a fairly-typical internet service offering in this area: Cox Cable (CXR) offers what they call "Preferred" service here that is $46/month. It is 12Mbps down and 1.5Mbps upbound. (They also sell a "Premier" service that is 25Mbps/2Mbps @ $62.)

The company also sells the service I have here at home. It's their "Business" internet service, and roughly matches "Preferred", being 15mbps down and 2mbps up. But here's the rub: It costs $200.

Now I need the business connection because I have externally-visible connections here, and while I can "cheat", their terms and conditions make clear that if they catch me cheating, that will be that.

The reason the business service is more expensive is that the company expects me to pound the line on a 24x7 basis - and I do. But notice the difference in price - more than four times as much money.

Why? Simple. It's the fact that they expect me to pound on the line.

So Netflix has a $9 monthly service plan that in effect, counts on you being able to push a $50 monthly service toward a traffic profile that the cable company would charge you $200 for, without paying the $200.

If and when you're forced to pay the $200 or stop pounding your connection as if it was a business link Netflix instantaneously implodes.

Let's talk about the costs in a typical ISP, whether it be a cable company, FIOS, or some other technology. There are roughly three cost centers associated with the physical infrastructure:

  • Local, "last mile": This is the connection between the ISP and you. In days of old ISPs bought banks of modem servers and line cards to provision this, along with circuits to local telcos. You dialed in with a modem over a phone line. Nowdays this comprises the DSLAM and backhaul connections associated with it, plus the circuit to your home, for DSL, and the various head end and neighborhood amplifiers and switches in the cable world that provide DOCSIS service. FIOS has a similar set of equipment. This equipment and expense is more or less traffic-insensitive - that is, it is "mostly" all hard cost that is invariant with the amount of use. This means that traffic between you and another customer on the same ISP is basically "free" for the carrier to handle, as it remains on parts of the infrastructure that do not, for the most part, have capacity choke points.

  • Switching and routing infrastructure. This is the equipment inside your ISP that nobody ever sees, but is very expensive and must be sized for peak load. In days of old these were typically CISCO 7513s and similar equipment from other vendors. Nowdays CISCO has even bigger iron, along with Juniper and other manufacturers, but their essential functionality and purpose is the same. This infrastructure aggregates, disaggregates and handles your traffic inside the ISP's infrastructure. It also handles local storage and services, such as a web farm for your personal web pages and email storage so you can have email boxes (if your provider gives you these things.) Your monthly service fee pays for this, but you never see any of it. Most importantly, however, this infrastructure has to be sized to meet the peak demand of all of the customers at their instantaneous peak usage - so the more "spikey" the load, the worse the deal overall - that is, the more that is bought yet rarely used. This reality is one of the reasons that high Mbps connections to end users cost more money - you might spike demand and overload this part of the infrastructure, and it's pretty-expensive stuff, especially when you need to add more of it, although when amortized over the entire user base it doesn't look so bad.

  • Long-haul interconnectivity. This is where the real nasty lies. Long-haul traffic is damn expensive. All ISPs try to avoid paying for this to the extent they're able. The way they avoid it is to pull connections within their own network's "reach" to public "meet points" where they can exchange traffic headed to someone else's customers. This is one of the places where games are often played, and where serious disputes erupt, although it is rare for customers to see them directly. The reason disputes erupt is that there are really big financial incentives for a carrier to execute what's called "Hot Potato" routing, where it will attempt to dump traffic as close to its point of origin as possible on someone else, forcing them to bear the long-haul cost, if it is headed for someone on the other guy's network. The other guy, for his part, wants you to carry the traffic as far as you can, and give it to him as close as possible to his end user, forcing you to bear the cost. The larger the flows involved, the more serious these disputes are, and they have frequently resulted in carriers refusing to peer with someone at a public meet point, demanding that you instead sign as a customer and pay to exchange traffic.

The problem Netflix is going to run into is that by purchasing bandwidth in a few places they are creating the conditions under which "Hot Potato" routing will come to the fore for those people it buys from. They're not going to find peers who will pick up long-haul costs - so they will either have to build their own parallel cross-country lines or get quite clever and build data centers to serve their data close to the point of consumption so that the "Hot Potato" scenario doesn't come up.

That's a flawed business model my friends.

Eventually Netflix is going to run into a problem with this. Management is, in my opinion, attempting to cost-shift this and deny it - they've got revenues up $59 million from a year ago but technology spending is only up $12 million. This sounds like excellent economy of scale - it's not - it's a model that has externalized costs and managed to get someone else to eat them. For now.

They're not going to get away with it. Their revenues are up 25% annualized but operating expenses are up almost 42%. The squeeze is on folks, and I don't believe it's anywhere near over.

Subscribers are up 52% year-over-year - which is outpacing operating expenses. But revenues are only up 25% while subscribers are up twice as much on a percentage basis. That too is a problem - people are paying less, on balance, to the company on a per-user basis. They claim revenue is only down 9.7% annualized on a per-customer basis as well in their financial release.

The company has a product that a lot of people like. I personally find the motion artifacts in their delivered streaming video to be unacceptable on anything bigger than a laptop screen, but that's me - I have fairly high standards, and my "minimum" these days is DVD-quality video. What Netflix actually delivers in terms of streaming content is more akin to an old VHS tape in my experience. When I last wrote about the company that drew a lot of fire, and that's fine - if you're not a quality hound or mostly watch old TV shows (which have moderately-poor or worse original content quality in the archive) you probably don't care. I do, and find it unacceptable.

But even ignoring the issues of resolution and delivered video quality (which, obviously, many people will and do ignore) the numbers are a different matter. Unless the company starts installing data centers near the nexus of all the population centers in America to provide the streams and interchange with "last mile" providers at or near those provider's hubs they are going to eventually run into one of two problems - either they will be forced to pay for their interchange or, if they (and others) manage to get the "net neutrality" folks to scream loud enough (and they may) the "all you can eat buffet" from the last-mile providers will disappear, instantaneously destroying their business model.

The cable company and other last-mile providers have an advantage here that Netflix never will - a near-zero cost of last-mile transport on an incremental basis. Disks these days are basically free, and in addition most of the last-mile providers have existing deals with the content owners for distribution. As such I expect a collateral attack to show up soon with cable providers installing disk farms and using their digital delivery systems to provide what amounts to Netflix's service at the same or lower price, undercutting their business.

At the same time, the company will be forced to either duplicate infrastructure that their competitors already own or pay for long-haul transport (either directly or via a "Hot Potato" charge to them), which will wind up impacting their margins as well. They can tattoo Netflix on COGS (Cost of Goods Sold) and once that realization sinks into the heads of the CableCo executives they will, I predict, do exactly that.

Add all this up and then top it with a P/E of 70 and I see a company that's overvalued by 50% or more at $172 a ticket. How far people will chase the parabolic blow-off is an unknown, but this much I will predict - this run will end in tears, as I see the same sort of business model problems that plagued the majority of the so-called "internet heroes" in 1999, and even those that survived got cut to ribbons in terms of share price.

Disclosure: Covered my short from the former $170 top at $151 pre-earnings and now find the stock hard-to-borrow and options premium somewhere beyond the orbit of Mars, which is a bummer - but I am actively looking for a good short-side play on this turd when I find a corner on it, as I fully expect it to flush.