Wednesday, February 17, 2010

100 Mbps for 100 Million Homes by 2020?

If Federal Communications Commission Chairman Julius Genachowski gets his way, the FCC will set a goal of 100-Mbps service delivered to 100 milliion American homes by 2020.

Genachowski says his preferred approach to a national broadband policy would require ISPs to offer minimum home connection speeds by 2020. The “100 Squared” initiative might in fact be too modest a goal, he suggests.

"We should stretch beyond 100 megabits," he adds.

The proposal is part of the FCC's national broadband plan, due for initial public comment in March 2010.
The goal is sure to come under some scrutiny by service providers, in part because there is not currently any way to provide bandwidth of that magnitude on a national basis while pricing service at rates most consumers would pay.

There is not enough usable wireless spectrum to provide that kind of coverage and usage, and fixed access networks are not completely or primarily subject to Moore's Law. While chipsets and processors do get faster, the cost of digging trenches does not get less expensive over time. In fact, construction cost is the dominant cost element for any optical network providing service directly to end users.

 "One hundred meg is just a dream," says Qwest Communications International Inc Chief Executive Edward Mueller. "We couldn't afford it."

Few customers now buy 50-Mbps services where such speeds are available, in large part because the cost is in the triple-digits range. Proponents might argue that the goal is 100 Mbps for not much more money than people now pay for 4 Mbps or 7 Mbps service, but it is hard to envision how even "free" opto-electronices could support such a value-price combination.
 
In other words, even if all the active elements actually were provided for free, could service providers actually build ubiquitous networks offering 100 Mbps or faster speeds, and price in middle-double digits? So far, the answer appears to be negative.

About 60 percent of the cost of building an FTTH network is construction work, ducts and cables, not to mention cabinets, power supplies and other network elements. Still, in some dense areas, it might be possible to do so, since the construction and cable might amount to about $1200 per home passed. Again, keep in mind we assume totally free opto-electronics.

In suburban areas the business case is marginal, at best, since about $2400 might have to be spent on construction and passive elements.

Since the FCC goal only calls for connecting 100 million homes out of possibly 113 million, we can safely assume the cost of most rural networks of such capacity need not be considered.

Of course, opto-electronics are not "free." But the point is that construction costs, were nothing else an issue, would still be a tough proposition, if the goal is very high speed access at prices most consumers would pay.
American consumers will be paying more for broadband in the future, if for no other reason than that most mobile plans will require it, and those charges will be paid for on a "per-device" basis, not "per home."

What seems improbable is that U.S. consumers are willing to increase overall broadband spending by an order of magnitude (10 times) to have 100 Mbps or faster service on a fixed basis.

One can of course argue from history. Prices for lower-speed broadband services have declined over time, while the prices for the faster tiers have remained stable, but speeds have increased. The issue is how much price compression is possible.

"In order to earn a return for investors, you have to be conscious of what consumers will pay. I don't know this is something consumers will pay for," Piper Jaffray analyst Christopher Larsen says. "It's a nice goal, but it's a little on the over ambitious side."

Having a "stretch goal" is fine. Firm mandates, though, might run smack up against stubborn consumer willingness to pay and the fixed costs of building access infrastructure.

No comments:

"Tokens" are the New "FLOPS," "MIPS" or "Gbps"

Modern computing has some virtually-universal reference metrics. For Gemini 1.5 and other large language models, tokens are a basic measure...