Why Isn't the Price of Broadband Obeying Moore's Law? | Hacker News Reader

Why Isn't the Price of Broadband Obeying Moore's Law?

technologyreview.com

18 pointsca98am7931 comments

Because the cost of digging up sidewalks, laying fiber, installing routers and switches has F* all to do with transistor density?

The cost of digging up sidewalks and laying fiber are a fixed cost that can be amortized over 10 - 30+ years. Cost of bandwidth upgrades after that is almost purely semiconductors/photonics.

Although strangely it does seem that the size of hamburgers does follow Moore's law.

And the Burger King quarter-tonner is particularly tasty.

I wish the author hadn't decided to use Moore's Law as a standard to measure the price against. It takes away from the point, or maybe is the wrong point entirely. I do feel prices are higher than they need to be for the consumer, and I very much feel this is because I as a consumer (at least where I live) have only one choice in the matter, cable or DSL. I wish we were debating that and not discussing why Moore's Law doesn't apply to this situation.

Have a look at the comments on that site. Very US centric. People are proposing all sorts of theories, while ignoring the data point from outside the US. And the article itself points those things out.

Because Moore's Law does not apply to broadband.

"Moore's law describes a long-term trend in the history of computing hardware." - http://en.wikipedia.org/wiki/Moore%27s_law

Because Moore's law speaks to transistor density, mentions nothing of price, let alone broadband price, and the last mile of the broadband market is dominated by the market forces of regulatory capture and cartels. Also, the dominant technology in the last mile is copper wire, not silicon.

I used to pay $30 a month for dial-up (14.4 Kbps), and now pay $40 for 7.5 Mbps, thats a doubling factor of 8.6 over ~20 years which is roughly 2.3 years per doubling, thats awfully close to Moore's law. The reason prices aren't decreasing is that new applications are coming out which consume the extra bandwidth and that people value the decreased time to load more than the reduced amount of money. It's the same reason that a processor still costs $200.

Actually, Moore's law is all about price -- or, rather, costs.

According to http://en.wikipedia.org/wiki/Moore%27s_law, its most widely accepted formulation states: The number of transistors that can be placed inexpensively on an integrated circuit has doubled approximately every two years (emphasis mine). Even the quote from the original Electronics Magazine article includes: "The complexity for minimum component costs has increased at a rate of roughly a factor of two per year".

Essentially, cost is the whole point of the Moore's observation. And the OP is asking the right question, which I've asked myself (http://berislav.lopac.net/post/495375785/is-there-a-moores-l...) and got no answer.

I don't believe that bandwidth should follow the same rate of progress as the transistor density, as there are different sets of constraints at work, but I'm also curious if there has been any recognizable pattern there as well.

I wish more people would take the small amount of time required to find out what Moore's Law really is. Since it only addresses increases in transistor budgets (due to density), it has nothing to do with price.

Another point that more people need to understand is that price = cost + margin. In a competitive market, that margin is determined by perceived value, so if you managed to convince someone that your widget is worth more than the next guy's, you can charge them a higher price... even if you originally bought the widget from the same supplier as the next guy.

The other issue to consider is how much price affects the perception of value.

Good point.

There is also a government enforced duopoly in broadband.

I doubt we would have seen the same innovation in chip design if the government restricted competition to the two dominant brands (Intel and TI, I think) when the industry was in it's infancy.

Cities do grant permits to place fiber lines to competing companies, but this is seldom done to run fiber into the home. The consumer market requires massive capital to achieve economies of scale that few companies outside of existing utilities can compete with. How do you compete with infrastructure that's been bought and paid for many times over?

Interesting. In Boston I have the pleasure of paying $45/month for 1.5Mbps internet service from Comcast. Sadly, living in the heart of Boston, I am out of range of Mifi and FIOS.

My Shannon–Hartley theorem applied to a pair of copper wires trumps your Moore's law.

Simple.

Because it doesn't have to (in the USA). The market drives price, not Moore, and people are willing to pay the relatively cheap cost to get internet in most locations.

And because the long tail to distribution in the USA is very long indeed, due to all the space, and that takes a lot of time and money which is sucked from the people who have it / can afford it.

What makes me curious is why transistor density does follow Moore's law. My theory is that it sets a benchmark to manufacturers.

It's well-known that Intel roadmaps are influenced by the expectations created by Moore's law, yes.

Now I'm really interested, could you show me a press release or a news article or something that mentions this? I've long suspected that this was the case but I can't really prove it.

There are a couple of references here: http://en.wikipedia.org/wiki/Moore%27s_law#As_a_target_for_i...

Because of the cost of digging up sidewalks etc (as mentioned: http://news.ycombinator.com/item?id=1694456)

But ALSO because of the massive teaming people calling their cable company when their wireless network doesn't work. Ever wonder why they give away wireless networks with data plans? It's to streamline the service they give to their buyers. It isn't just anybody that can setup a router or diagnose a DNS problem from a loss of connectivity one.

Perhaps because of increasing amounts of people wanting broadband, and it being viewed as a civil right, the prices have to be held in order to support the network expansion to the more rural areas?

Perhaps. But you should look at what the companies are really spending their money on. That argument only works if the share of (rural) investment in infrastructure is sufficiently high.

But US ISPs are not expanding in rural areas.

Why Isn't the Price of Sheep in Sudan Obeying Moore's Law?

I pay $2,500/month for full use of a 1gbps circuit at one of my datacenters. 10 years ago I would have paid about $250k/month for that sort of capacity (and another $100k for a router/firewall that could handle it).

Bandwidth has definitely gotten significantly cheaper, which of course doesn't solve the broadband problem. If you live in the suburbs, or on a farm, then you've chosen a lifestyle which is inherently wasteful & inefficient, too much so to facilitate rapid infrastructure upgrades. Move to a city if this matters to you. Here in SF I pay $33/month for a synchronous 100mbps ethernet connection.

Who do you use for the 100mbps connection?

http://web-pass.com/ .. We actually found our apartment complex by calling up every building on their list of lit buildings and asking if they had any condos or apartments for rent. Turned out that there are at least 3 other geeks here I know!

My friend John has a similar offer in Seattle too, at http://condointernet.net/.

Yeah, that's what I figured. Unfortunately, I'm not willing to move for an internet connection.

(For those who are wondering, this is an extremely special case -- most home bandwidth in the bay area is in the <10Mbps range, asymmetric. The OP is using a service that's available at only a handful of SF apartment buildings.)

It's growing, quickly. Others are entering this space, their main obstacle (besides the cost of running fiber) is that so many buildings have signed exclusivity agreements with comcast which have questionable legality. The economics make it easier to do in a city.. pay $20k to run fiber to a building with 1100 units, sell 25% of those units at $35/month, you gross $120k/year from that building. You probably have another $10k-$12k in equipment costs per building as well, but this sort of a network is quite simple to run.

This sort of connectivity will never be feasible in the suburbs without massive government subsidies (something I had hoped our current administration would see the value in).

Reply on news.ycombinator.com