« Fighting marine debris: the dustcarts of the sea | Main | Electric Road Trains »


Feed You can follow this conversation by subscribing to the comment feed for this post.



Please get your units straight. Watt is a unit of power, not energy. Watt-hour on the other hand is a unit of energy, like Joule. It's wrong to say "..consume 60 watts of electricity per hour". Just lose the "per hour" and it's better already.

Otherwise, another great in-depth article on an important issue.




So if we figure embodied energy in a product does that me the manufacturing equipment runs for free? Why not just say this manufacturing equipment and processes are inefficient?

kris de decker


Jan: thanks, corrected. Also good to hear that at least someone agrees this is an important issue :-)

Jon: manufacturing equipment and processes are inefficient and can be improved substantially. But energy efficiency has its limits and can never solve this issue by itself (see the article). We also have to address technological obsolescence. If products are so energy-intensive to manufacture, it is insane to replace them after two or three years. With servers and network appliances embodied energy is already of lesser importance, because they are only replaced after 5 to 8 years.



Just to add to the energy units.

This bit doesn't make sense "(around 180 watts, or, using the appropriate unit of measurement, 648 kilojoules)" since the comparison is between rates of energy consumption. Similarly, "Manufacturing one kilogram of electronics or nanomaterials thus requires between 280 kilowatt and 28 megawatt of electricity" doesn't make sense either here you want to be using amounts of energy rather than rates of use. These were the two errors I noticed, but I think you should go through the article again to check for other instances in what is otherwise a very interesting piece.



Actually, before publication I replaced all "watt-hours" by "watts" because I thought that would be more clear to most readers, and anyway 60 watt-hours = 60 watt x 1 hour.

But you are right to correct me. Article now makes clear distinction between units of power and units of energy. Thanks for the advice.

Gert van Loo


Nice article but how much worse does it get if you add the design energy?

I am one of the engineers designing these chips.

We are a team of 60 people designing about 1 graphics chip per year. For that we use 2/3 building, about 3 managers, 3 sales persons, 4 administrators, 2 IT persons. 2 LCD screens/person. 1 computer/person and computer farm of about 74 very high end Linux boxes (highest possible speed with 16-64Gbytes of memory each) with all the networking file servers etc. etc. all of which run 24/7.

Maybe you could from that info estimate how much more energy is required on top of the production energy.



Wow, Gert, many thanks for sharing this.



Excellent post. This is the first article that I've ever read covering this hidden problem. I recently retired from the semiconductor manufacturing field, and this has been an environmental concern of mine for a long time.

The direct manufacturing alone is astronomically energy consumptive. On the fab line, and in the analysis support labs, the machinery runs 24/7, including vacuum systems, chillers, ovens, SEMs, and all manner of connecting and contributing mechanical systems -- all of these equipment components running non-stop every day of the year.

Second on the consumption list is the vast use of support materials, e.g., the water, acids, metals, gases, etc. Not to mention consumables such as cleanroom accouterments. And not only can the direct usage be accounted for, the amount of energy expended for the waste products involved such as disposable safety apparel, chemwipes, oils from vacuum pumps, chemicals, etc., discarded daily is enormous as well.

I just shake my head when I hear someone comment that electronic equipment is so small now that it hardly makes an impact on the environment at all. You are correct in stating that the smaller processors get, the more difficult and consuming it is to manufacture them (and I've seen it all the way back to SLT chips). And I'm very surprised that the environmentalists have their heads completely in the sand about this topic, if they've ever even thought of it. They worry about smoke-stacks? If they really took a look at the process details of semiconductor manufacturing their heads would explode. It would be very interesting to see a study published on the overall collective costs of a computer from design stage to finished product.



I'm researching a state-of-the-art snapshot of green ICT technologies for a university of applied sciences and an ICT R&D cluster and stumbled upon this article. Very, very interesting. Also I'm a huge proponent of Free Software and totally in agreement with your conclusion. Thanks for the great read!

Pangea Joel


Although i did appreciate your article's breakdown of manufacturing cost of ICT components, the value of the article was severely detracted from by your misleading and fallacious conclusion:

"No environmental benefit would appear and the energy savings realised by digital technology would merely absorb its own growing footprint."

i just read the Smart2020 report referenced in your article.

Appendix 1 defines the Climate Group's methodology of calculating "embodied carbon."
Appendix 2 clearly does break out explicitly and quantitatively the embodied carbon from the manufacturing processes for the ICT components.

So your concluding claim that the 1.4Gt consumed by ICT in 2020 should be multiplied by 4 and thus reach 7Gt and thus virtually erase almost all the 7.8Gt f energy saving benefits from ICT is fallacious and misleading.

The energy savings of ICT stands.

I am at a loss to understand why you felt compelled to distort the Climate Group's report you referenced.
Are you so motivated by your ideology that you failed to read the Smart2020 report you referenced?
Or were you hoping that no one would drill down to the source material?

Kris De Decker


If I link to my source material, that is because I want readers to verify my claims. Thank you for doing that. However, you are wrong to say that I was trying to distort the report I referenced. The truth is that I did not read the appendices. I am very surprised to learn that they contain information on embodied energy of electronic products, because I read the main report from A to Z and nowhere it is suggested that the production phase is included.

So at first, I thought you were right and I had to change the conclusions of the article, but then I did what you did: I drilled down to *their* source material. And then I found out that my conclusions still stand, more than ever - and that instead their report is truly misleading.

To start, check their source for the embodied energy of laptops and desktops in Appendix 1. They write: "The calculation of CO2e as part of the manufacturing proces of the components is calculated based on publicly available data (101) or company data".

If you then check footnote 101 this is what you get: "IVF Industrial Research and Development Corp (2007), preparatory studies for eco-design requirements of energy-using products (EuP): Lot 3-PC's (desktops and laptops), final report".

I pasted this into Google (contrary to me, they don't link directly to their sources) and after clicking through some documents I found their source:

http://www.ecocomputer.org (see "documents")

Go to page 120 of the final report and read this: "The main computer and monitor manufacturers supplied the data presented".

So, above they say "our calculation is based on publicly available data (101) or company data". But it turns out that this so-called publicly available data is also supplied by the same manufacturers. So all their information on embodied energy is solely built on the claims of manufacturers. In other words: they refer to themselves.

My figures are based on independent research. Now let us compare both outcomes.

My final conclusion (the one you challenge) is based on the study that says that 80 percent of the energy use of a desktop machine is due to the production phase. However, the study on which the Smart 2020 report is based comes to the opposite conclusion: they say that for a desktop PC, five to six times more energy is required during the use phase than during the production phase. For a CRT-monitor this is even six to thirteen times.

They also have figures for more modern configurations (see page 157 and further), but I select a desktop + CRT screen to compare with my own figures. For all configurations, they conclude that the use phase strongly dominates total energy use.

Why do they arrive at a diametrically opposite conclusion than other reseachers?

Firstly, they assume a life expectancy of 6.6 years for desktops and 5.6 years for laptops (see page 142-144). The life cycle analysis that I based my conclusion on assumes a life expectancy of 3 years (a very common assumption). This already makes a huge difference, as I conclude in the article that longer life expectancies are the clue to address the footprint of digital technology.

Secondly, for all their figures, they refer to Appendix 2 of this document (page 282). There, I expected to find the information on how they calculated the embodied energy of integrated circuits. However, IT IS NOT THERE. They give the weight of integrated circuits in a laptop or desktop (very interesting information, see below), and they convert that into a figure in MJ. But, nowhere in this appendix they give any source or formula for this conversion. We just have to believe it.

Now go to Appendix 3 and you will find out that the study to which the Smart 2020 report references to as a source for embodied energy is written by AeA Europe, EICTA and JBCE, which are all organisations representing manufacturers of digital technology.

So in conclusion the Smart 2020 report might contain embodied energy of production but these figures are ridiculously small and based on nothing else but their own claims. I love appendices, by now...

Very interestingly, they do give figures in Appendix 2 for the amount of integrated circuits in a computer, data I could not find before: 165 grams for a desktop and 78 grams for a laptop.

If I multiply these figures by the information collected in my article (20 kWh embodied energy per 2 grams of microchip) the results are 1,650 kWh of embodied energy for a desktop (without the screen) and 780 kWh of embodied energy for a laptop.

Above I calculated that only the memory chips of a laptop have an embodied energy of 360 to 720 kWh, which is already enough to power a laptop non-stop for 500 to 1,000 days (or 12,000 to 24,000 hours, well above its life expectancy).

Thanks a lot for challenging my conclusions. But, the energy savings of ICT do not stand.



Another way to do this is to find the cheapest competing brand (people pay a lot for a brand name), subtract out all the employee benefits costs, raw material costs, salaries, maintenance, vendor service costs, and profit (without profit companies go bankrupt), and you'll be left with the total energy cost. Divide that by the cost of energy and you'll have the kWh burned.

Take the iPhone ... cheapest competing product (without a plan) can be had for $200. Subtract all that gunk I mentioned out and I bet the number ends up somewhere around $20, maybe more. At $0.05/kWh that device then took 400kWh. It takes an average household (14MW/yr) 10 days to consume that much energy. By buying that iPhone you just consumed as much energy as an average family uses for 10 days. That's alot.



What about Apple Inc. and their Eco report? The impact of the aluminum unibody manfacturing process and use of renewable energy sources?

Brent Norris


Awesome, let's go back to building stores out of bricks so we can drive to buy more things! Just kidding, great article, great work. Very comprehensive.

I wonder how energy intensive self-assembled, tablet, 100% powered by the sun will be?



I work in a UK science and technology institution, in the ICT department and measured the power usage of the whole department. I worked out that lights, monitors and P.C's only used about 10% of annual energy consumption, with the energy hogs being fan coil units that run all day, everyday for the entire year.

I think if you use digital technology effectively - Digital cameras, MP3 downloads, webconferencing, it can save energy or use less than conventional methods, but the problem is, the old methods are not becoming obsolete. they are still here.

The biggest problem we have is infrastructure, which is much, much more difficult to alter and make efficient, both practically and bureaucratically.



One can simplify matters a great deal by realizing that your $700 computer did not take more than $700 of energy to make. That would be your conservative, worst case scenario.

Ivor Cogdell


A very good article. To extend the life of Motherboards, use a modular, plug in, approach to all micro-processors, not just the main one. All internal and external connections should be standard (not always the case, in my experience). To extend the usage of individual memory cards, design them with a memory slot on top, so another one can be plugged into it when needed.
I think computer recycling should be given much more publicity.



Hello, excellent article.
After the financial crisis following 2008, many people started finally understanding that infinite growth is not sustainable, that the principle on which our economy stands are not acceptable anymore and so on (I hope you already have heard of this, in Italy it is a big topic now).
I was wondering about the 2020 estimate: in 2020 the ICT will save 7.4Gt. Now, EVEN IF your assumptions or data were not correct, and thus say that ICT costs a bit less than 7Gt you calculated, say 5Gt obtaining an overall saving of 2.4Gt by 2020. EVEN IF it saves something on the short term, market laws require infinite growth! It means that then, say 10 years later, the global market has doubled, thus the ICT costs 10Gt CO2 and saves 4Gt. They may say "we doubled our CO2 saving", but they actually doubled the CO2 emissions: 14Gt. What I mean is that the 2020 report is not only deceiving (I didn't read it, but I understand from the comments that references were bad), but also ridicoulus: if the market requires growth, the environment can never be saved.

What do you think about that?

Adrián A. García


First of all, congrats on this blog, it is really fantastic.

Now to the point. Perhaps an energy-saving solution for our everyday computing needs could be computers able to do most of what we usually do with them but at the same time using a fraction of the materials and energy that we currently use on a typical computer. And by this I mean a truly significant reduction of them.

Take the Raspberry Pi as an example: http://www.raspberrypi.org/faqs. A 25$ computer with the size of credit card made of very few electronic components and able to do multimedia tasks. In time more capable computing platforms based on the Raspberry Pi philosophy will surely arise, rendering our current PC solutions obsolete.

What I try to say is that if people becomes aware of how important it is to reduce the energy footprint of electronic products, great progress could be made in a relatively short period of time. Because, let's be realistic here, it's doubtful that people suddenly stops buying and using electronic appliances, most of them designed from the beginning to last no more than their warranty.



Here's my thing. To me it's misleading when you compare the energy to manufacture of heavy simple "normal" goods/lb, vs. Electronics/lb. You get all of these figures, but no indication of the "Utility"/KWhr.

Keeping electronics (or any thing) longer is an excellent idea, it is antithetical to our Marketing driven economy. How would anyone accomplish the same utility of a computer (Word processing, publishing, design, etc.), with some "normal" low-tech device?

While the Moon was reached with mostly slide-rules and T-squares, almost every goods/articles in use today would be UNPRODUCABLE with those limitations.

In Summary, to get tens of orders of Magnitudes of savings, it would be much better to reduce population - the real driver behind our profligate energy consumption.

Tom M



You have '20 kWh embodied energy per 2 grams of microchip' I think your calcultion is off -

The book - "Energy in Nature and Society: General Energetics of Complex Systems" - states it takes 21 MJ/chip with each chipe weighing 2g, the equivalent of 20 GJ/kg of produced electronics.

20 GJ/kg = 5.55 KWh/g
41 MJ/2 g = 5.7 KWh/g

So about half your number.

5.55 KWh/g * 165 g = 916 KWh
5.55 KWh/g * 78 g = 433 KWh

Tom M


We could also estimate maximum embodied energy by manufacturing cost.

"A 16GB iPhone 5S costs Apple Inc. $213 in materials, while a 16GB 5C comes in at $156, according to a report published Monday by UBS AG."

213$ assume that was pure energy usage -> .12$/KWh ->
1775 kilowatt hours for the most expensive iPhone.


Or the 25$ raspberry pi above 25$ * 1 KWh/.12$ -> 208 KWh

Kris De Decker


Hi Tom,

Thanks a lot for the additional information. Do you know where Vaclac Smil (the author of the book you mention) got his numbers?

It would make sense if the energy intensity would be roughly half my number by now. As I write in the article: "The International Technology Roadmap for Semiconductors (ITRS), an initiative of the largest chip manufacturers worldwide, aims to lower energy consumption per square centimetre of microchip from 1.9 kWh today to 1.6 kWh in 2012, 1.35 kWh in 2015, 1.20 kWh in 2018 and 1.10 kWh in 2022."

My number of 20 kWh per 2 grams of microchip was based on this information and on the 2002 research of a 32 MB RAM chip (in 2009 still the most up-to-date study available).

However, Smil's book was published in 2007, two years before my article, and then the numbers don't correspond. So I wonder to which study he refers.

Manu Sharma


Apple publishes life cycle energy analysis for each device and if I remember correctly the manufacturing and usage energy ratio for Macbook was 60:40. If that's correct than the claim about embodied energy for memory chip might not be correct. The data should still be on Apple's site.

Francesco Pasa


I think the main problem is obsolescence, and that is a huge problem to solve, because the entire IT market revolves around selling new products when the old ones are just good enough for the same tasks. Software is also big part of the problem, because most programs become heavier on the hardware every year (especially OSes and browsers).

I believe the solution is rather simple: (1) Design software to run on old hardware (or low-power like the raspberry) and support it for longer, so that we can use the same hardware for longer. (2) Make hardware modular and replaceable. If my laptop's processor could be changed, I would use it for 10 years instead of just 5 (and I believe I'm one of the few who use a computer for 'so' long), saving the rest of the hardware. Make broken parts easily replaceable. That is an issue also for the raspberry, which is a one-piece board with all component surface-mounted. You cannot adapt your hardware by just replacing the needed part.

I think the problem is more one of mentality (to be cool you need the latest gadget) and of market.

Have you tried looking on the apple website? Mike Bernares-Lee in his "How bad are bananas?" writes about a life-cycle assessment report from apple on his products.

Great article, keep up!

Jonathan Sundqvist


There are actually two LCA industry studies, one done by dell: https://www.one-report.com/download.html/2011/shared/library/0692-00006713.pdf and then LCAs done by apple (http://www.apple.com/environment/reports/). The carbon footprint of a laptop is between 300-400 CO2eq and manufacturing represents between 50-65% of that footprint. So your number of 80% is way off to be frank.

You might be interested in reading this too: http://greenelectronicscouncil.org/wp-content/uploads/2013/12/slateswkshp/GECTabletsWorkshopDec2013_LCA_Teehan.pdf



I think we should research LCA of quickly depreciating goods and parts:
- batteries, in remote controls etc..
- accumulators, now that smart phones and watches need daily charging, and we have electric toothbrushes etc..
- SD cards, USB memory sticks- not only electrical wear but also mechanical damages increase turnover
- embedded electronics (greeting cars, wearables...)
- headphones
- ....

Vikas Kashyap


A decent article. To expand the life of Motherboards, utilize a particular, connect to, way to deal with every single miniaturized scale processor, not only the principle one. All inward and outside associations ought to be standard (not generally the situation, in my experience). To broaden the use of individual memory cards, plan them with a memory opening to finish everything, so another can be connected to it when required. I think PC reusing ought to be given substantially more attention.



I think the most drawback is degeneration, which may be a vast drawback to unravel, as a result of the whole IT market revolves around commerce new product once the recent ones square measure simply ok for constant tasks. software package is additionally massive a part of the matter, as a result of most programs become heavier on the hardware each year (especially OSes and browsers).

I believe the answer is quite simple: (1) style software package to run on recent hardware (or low-power just like the raspberry) and support it for extended, in order that we will use constant hardware for extended. (2) create hardware standard and replaceable . If my laptop's processor might be modified, i might use it for ten years rather than simply five (and i feel i am one in every of the few UN agency use a laptop for 'so' long), saving the remainder of the hardware. create broken elements simply replaceable . that's a difficulty additionally for the raspberry, that may be a one-piece board with all element surface-mounted. you can't adapt your hardware by simply commutation the required half.

I think the matter is additional one in every of mentality (to be cool you would like the most recent gadget) and of market.

Have you tried counting on the apple website? electro-acoustic transducer Bernares-Lee in his "How unhealthy square measure bananas?" writes a couple of life-cycle assessment report from apple on his product.

Great article, keep it up!


The comments to this entry are closed.