[wwweston]

That would miss gowld's point, which seems to be that a good first order approximation of the energy expended in the manufacture of a new machine is probably captured in a fraction of how much its retail price is.

You would definitely not expect that a machine selling for $1000 would have used more than $1000 of energy in its manufacture. Even accounting for energy costs all the way down the chain (materials arguably don't have any other cost other than energy and attention involved in collecting/isolating/refining them), my uninformed guess would be that with something like a computer, it's something like a bit less than half, with the rest of it being amortized human attention at various stages plus some profit margin, though I'm sure there's more refined and accurate models available.

So, throw out $480 as an energy number for our $1000. That means old hardware that's less energy efficient to the tune of $40 a month will out-impact the manufacturing cost of the new machine in a year.

As an exercise, contrast this with older automobile. If you've got a well-functioning 10-20 year old vehicle, it's probably somewhere between 80% to 50% as energy efficient as the higher efficiency choices you can buy new off a lot today. But the sticker price of a vehicle will tell you that it probably $10-20k of energy to produce. Because that number is high, from today, it will probably take your used vehicle longer than its remaining lifetime to exceed energy use involved in making the new car (and most new vehicles won't get you where you're going any faster either).

[jhayward]

> You would definitely not expect that a machine selling for $1000 would have used more than $1000 of energy in its manufacture.

This is a misconception. Energy is not a transportable, fungible quantity the way dollars are. It is entirely possible that the device one buys for $1000 would require more that $1000 of energy to make, if manufactured in a modern economy with high environmental standards.

The major forces in the global economy over the last 3 decades have been this imbalance in labor, energy, and environmental compliance costs. The $1000 retail price in the US does not contain the largely externalized costs that its place of manufacture may have permitted.

[sitkack]

Buying items in highly competitive industries from far away places is the most efficient way to turn western dollars into pollution.

What the parent didn't enumerate is that the energy cost has to be calculated as the point of use. You can do this for say solar panels made in China. If you assume that 100% of the purchase price is translated into energy costs, zero physical resources used, etc. That 100% of what you pay, went into energy, used by the nastiest sources you can get an upper bound on energy costs to create an item are. In China, $/KwH is 2.5-5 cents US. Probably lower if you are in some direct use of coal scenario. Using the lower bound, say a $100 solar panel made in China consumes energy costing 2.5C/KwH. It could have used at most 4MWH of electricity.

[wwweston]

> Energy is not a transportable, fungible quantity the way dollars are.

You mean energy and dollars aren't both transmitted up/down wires or moved around with mass that represents stored potential?

> It is entirely possible that the device one buys for $1000 would require more that $1000 of energy to make

This.... doesn't sound like a business model that will last long. Can you give a concrete example?

[jhayward]

Electrical power does not cross oceans. Dollars do. A factory in China running on a local coal plant's energy does not pay the same cost that a factory in Texas would pay.

[simonh]

If I go out and burn enough coal to release 1000 Joules of energy into the atmosphere, will that cost me the same as 1000 Joules of electricity consumed by my electric oven? Of course not, it would be orders of magnitude cheaper. You can’t make simplistic assumptions about the cost of energy in a product based on its final price because the costs of energy in different forms vary enormously - by many, many orders of magnitude.

[wwweston]

The costs involved in the manufacture of a product at any single point in time are exactly what I can make assumptions about from the price of the product, nothing simplistic about it, given that prices are a result of the negotiation of a lot of details, including whether the optimal industrial input is some raw mass of coal that's burned in some managed way or Watt-seconds of directly supplied electricity. It's not going to be a perfect signal (demand matters, and any single estimated or registered price reflects a certain degree of imperfect judgment/optimization), but energy inputs are going to be a bounded factor.

If what you're saying is the actual energy expenditure may be what we're concerned about if we're speaking about environmental impacts and might be more properly modeled by something more complex, that's a worthwhile point. But it's going to be much less a matter of 1000J from a given mass of coal vs 1000 J of directly supplied electricity -- this factor will disappear behind whatever market allocations/optimizations are available -- and much more a matter of general industrial energy costs circa 1995 vs 2015, combined with the relative efficiency of manufacturing processes at both points in time.

What would we expect on those two fronts? Personally, I'd expect energy prices to rise with economic growth and occasionally fall with recessions, absent some large new source coming online or state-imposed costs for use. I'd also expect process efficiency to increase as well. Which would lead me to, again, see energy expended as reasonably estimated by some bounded factor of a final product price.

[yellowapple]

> So, throw out $480 as an energy number for our $1000. That means old hardware that's less energy efficient to the tune of $40 a month will out-impact the manufacturing cost of the new machine in a year.

That seems like an obscenely high difference in monthly energy cost (if we're going for an apples-to-apples comparison, in contrast with the article's posed comparison of a desktop workstation v. a consumer laptop). For reference, I run multiple desktops, multiple laptops, a full-size fridge, lights, fans, and an Echo, all mostly 24/7, and per PG&E my total monthly power bill (near SF) is less than that (and most of my hardware is on the older side).

We're more realistically talking (from my experience, running a lot of the sorts of older desktops the article mentions) a difference closer to $4 than $40. Even $10 (which would still be a pretty high estimate) would extend your estimate to 4 years until break-even.

[wwweston]

Of the several objections to my comment people have registered, this seems like the best one: on reflection it seems likely that an 85W laptop's daily use is likely to be around 1 kWh (maybe 2 kWh if driven near capacity 24 hrs), which is on the order of $5-$10/mo. So for things to come out something like I'd speculated, either older workstations would need to use much more power (5-10kWh) or the manufacture of something new would have to involve much less energy than I'd guessed.

[crygin]

In addition to the sibling comment's point that externalities of energy production are unevenly distributed, the difference between an old machine and a new machine is that in the manufacture of the old machine the energy/pollution/labor have already long since been expended, and the amortization of that expenditure across more years of use is likely to compete well against even a very efficient newly manufactured piece of equipment. This is most dramatic in e.g. the purchase of cars, where it's typically more environmentally friendly to drive an older, even significantly less efficient car, than to purchase a new car.