Visit Your Local PBS Station PBS Home PBS Home Programs A-Z TV Schedules Watch Video Donate Shop PBS Search PBS
I, Cringely - The Survival of the Nerdiest with Robert X. Cringely
Search I,Cringely:

The Pulpit
The Pulpit

<< [ Rules of the Road ]   |  It's the Bandwidth, Stupid  |   [ Reach Out and Touch Someone ] >>

Weekly Column

It's the Bandwidth, Stupid: Why Owning Big Server Farms Can Be a Hard Way to Make a Buck

Status: [CLOSED]
By Robert X. Cringely

The thing about food chains is that a disruption anywhere along the line can have major consequences elsewhere in the chain. Kill all the plankton and the predator fish die — not because they eat plankton, but because they eat the fish that eat the plankton. Much the same thing is happening among Internet businesses, where the failure of advertising-based Web businesses is putting a major hurt up and down the line, bringing companies as huge and powerful as Cisco and Lucent to their knees. One class of operation that seemed for awhile immune to this problem was the big Web hosting outfits, but even they are now succumbing, going out of business or being acquired. Conventional wisdom says these bit factories, as Andy Grove calls them, just over-expanded in a cooling market, but the truth goes deeper. On top of the usual mistakes of Internet businesses, many of these hosting outfits have grossly overpaid for Internet bandwidth. They've been preyed on by the bandwidth providers, predominantly telephone companies.

Web hosting companies own data centers — large buildings filled with thousands of servers. These data centers are secure, often having armed guards, and they offer data-grade amenities like conditioned power, back-up generators, halon fire supression systems, and of course, massive Internet bandwidth. Lots of big companies like IBM and Intel jumped into the business over the past two years in competition with startups like Digex, Digital Island, and Exodus. It looked like a great business to be in, hosting all the big dot-com web sites — until those Web sites started going out of business, often leaving their bills unpaid. Now, having spent 18 months building massive capacity, some of the big companies are getting out of this business and the startups are being acquired, often by telephone companies.

The poster child for big Web hosts is Exodus Communications, headed by Ellen Hancock, late of Apple Computer and IBM. Exodus has raised and spent billions building data centers all over the world, but now Wall Street analysts are circling the carcass of Exodus like sharks, sensing vulnerability and each wanting to be the first to predict the end is near — that end being the point where it becomes clear that Exodus will run out of cash before it reaches profitability. Remember, these are the same analysts who were forecasting as late as October 2000 that the U.S. Web hosting market would grow from $2 billion in 2000 to $18 billion in 2004, a 75 percent compound annual growth rate.

The outward problem that Exodus and most of its competitors have is that dot-com customers have been failing, leaving bills unpaid, and new customers aren't taking their places. The Internet was growing so fast that it seemed for a long time that there would always be new customers — until suddenly, there weren't. And this profligate atmosphere led to a lot of bad business practices. The simple fact is that in a business climate that expects large losses, it is easy for those losses to get out of control without the CEO (or even the CFO) even knowing it.

When there is a large pot of money in the account — money from a recent funding cycle (or better yet a successful IPO) — immature companies get lazy and pay little attention to cash management. Just because you have the cash doesn't mean you have to spend it. What we have here is a lot of really bad management. Or maybe it is the Internet reality distortion field.

No matter what kind of business you're running, you have to look at theReturn on Investment (ROI) numbers. The rule of thumb in all markets except, perhaps, Internet fantasy markets, is that your business cycle must be longer than your payback period. If you are selling three-year Web hosting deals, your payback must be a lot less than three years. In the case of many Internet companies, this rule has never been recognized, nor has it been enforced by their funders or the public markets.

When you start a new business, you don't over-invest and you work towardprofitability. How many Web hosting firms spent one minute before last Fall working to reduce costs and increase margins? These firms spent billions on increasing capacity, and almost nothing on improving productivity.

Then there is the bandwidth issue. Few of these hosting outfits run their own networks, and most of them have been taken advantage of by bandwidth providers. These were server people, not datacom people. What did they know about bandwidth provisioning? As such, they often accepted bad deals form backbone providers.

Exodus, for example, not long ago bought Global Crossing's Global Center hosting business. Global Crossing is a bandwidth supplier, and they built the hosting business to use up excess bandwidth in their own early days. But when the network market grew enough and hosting became a less attractive business, Global Crossing wanted to dump Global Center. It is not clear that Exodus really needed the extra capacity, but they picked up some new customers, and a big part of the deal for them was preferred prices on bandwidth from Global Crossing. Getting bandwidth costs under control was the major incentive for Exodus to do the deal at all. But those cost savings won't be soon realized because Qwest (Exodus's previous bandwidth vendor) won't let them out of their current contracts.

This is a common story. Digital Island and Digex couldn't get network costs under control in time to avoid running out of cash, so both had to sell out. Companies like DataReturn and will probably never be able to get to a point that they have positive margins. They will always be at the mercy of those they buy data center capacity and bandwidth from — in both these cases, Level 3 Communications. And since Level 3 also competes with these businesses, don't expect anyone to get a break.

But this bandwidth problem goes further, all the way to the end customers, who are often demanding — and paying for — more bandwidth than they need. "The vast majority of our customers use less than a T3, many less than a T1," says a friend of mine who runs a number of very large hosting centers for a major player in the business. "Like anyone else we try to buy a reasonable amount of capacity that will cover demand. We'll get a new customer who 'has to have four OC-3s' for example. We'll compare them with a similar customer situation and realize they probably only need a couple of T1s. Their consultant, who is raping them on software development, is the one insisting on OC3s. For their site we'll then have to buy the additional capacity and charge them for it. We'll even suggest a lessor amount with contract flexibility to provide more bandwidth if and when it's needed. Their lawyers get involved and insist on the full bandwidth, don't share it, and don't suggest anything to reduce it."

This ego bandwidth contributes to the failure of Internet ventures that contribute to the weakening of the hosting facilities where those ventures live on the Net. It's a vicious circle.

But this doesn't mean that it is impossible to make money in hosting. What it means is that all the profits are being made at the far ends of the business. There are some very interesting and profitable Web hosting firms who can sell you a nice moderate sized web service for less than $25 a month. These firms don't offer a lot of options. Basically you get Apache and Linux on an intel server. By the use of some simple perl scripts, these firms have created a complete set of system management tools that allow their customers to do most of the actual work, including e-commerce. And this is plenty for the vast majority of Web sites.

At the other end are companies that see hosting not as an Internet business, but as an IT business. One of these outfits is Conxion hosts big companies like Microsoft and Oracle and lots of banks — companies that aren't subject to the boom-bust Internet cycle. These companies are more informed and more demanding buyers, willing to pay more to get more. As such, Conxion is profitable, debt-free, and has completely avoided the other pitfall of equipment leasing. They own what they own. The company views as its greatest strength, though, a deep knowledge of networking that has allowed them to negotiate toe-to-toe with the very bandwidth providers that are killing companies like Exodus. With the lowest backbone costs in the industry and nine data centers around the world, Conxion will outlast most of its competitors. Conxion CEO Antonio Salerno, a scrappy guy I've known since his days at Borland and Atari, sees this as his shot, and he will do what it takes to make his company succeed.

Another possible survivor is Akamai, run by George Conrades, another ex-IBMer. Akamai started as a specialist provider of streaming content from thousands of serves positioned closer to end-users. But in the last year, Akamai has added general-purpose distributed hosting to its mix with customers like The Motley Fool and even PBS Interactive, the folks who bring this very page to you. Akamai has done its share of profligate spending, as a recent $2 billion write-down attests, but now the company is making an earnest effort to grow out from under its cost structure. The new strategy places Akamai servers in host ISPs that provide free rack space and bandwidth, in exchange for other Akamai servers at other ISPs extending the reach and fault-tolerance of the participating ISP's own web pages. It is effectively a cooperative, though Akamai probably doesn't like to think of it that way, and once enough servers are distributed to ISPs, Akamai's costs should drop dramatically and the company will be profitable. I see it as a survivor.

And what of the rest? Some will be gobbled by their bandwidth suppliers, now their biggest creditors. And the rest will be aggregated by some old-line enormous company like, say, General Electric after the shakeout is complete. I have no particular knowledge of GE entering this business, but it just makes sense.

Going back to the food chain analogy, the ultimate power here lies with the backbone providers, which in this chain actually perform the role of plankton. In the end, you know, the plankton always wins.

Comments from the Tribe

Status: [CLOSED] read all comments (0)