In 2004 it took an estimated 6400 megajoules to build a typical computer, including 17" CRT. This works out to 1778 kilowatt/hours, or about the average consumption of a house in the US for two months. Based on the 2009 US average industrial rate of $0.07 per kwh, assuming that only electricity was used and at 100% efficiency, $124.46 of the cost of the computer went to energy alone. I reckon this would represent somewhere around 10% of the total cost.
A gallon of gasoline has about 1.3x10^8 joules or 138 mj, meaning the computer would require about 46.4 gallons of gasoline to build. Using a rough average of the current prices, $2.70, this means about $125 worth of gasoline. Note that 1 gallon of gasoline has about 36.6 kwh.
Alternatively, the current average residential rate for electricity is $0.12 per kwh, meaning the cost to build the computer would be $213.36. Notice that a mere change of 5 cents to the cost of a kwh nearly doubles the end cost of the energy, which would most likely be reflected in the purchase price. It's important to recognize that energy and the cost thereof, from, gasoline, electricity, or beyond, is extant in all facets of our modern lives. In other words, if the price of gasoline goes up, the price of everything goes up. Of course most of our electricity is generated from coal and natural gas, so the price of gasoline doesn't seem directly related to building a computer. Unfortunately that's rather short sighted, as gasoline is required in order to move the computer parts to and fro, not to mention to transport the coal to the power plant to begin with.
Clearly if the cost of both gasoline and electricity were to rise by a nickel, we should naturally expect everything we buy to become quite a bit more expensive. It isn't difficult to see that this in turn would most likely have dire economic consequences. This is why there's so much buzz about energy, it should be obvious that the extreme consequences of demand outstripping energy supply readily justify extreme evasive efforts.
Imagine it was discovered that a meteor sufficient to absolutely obliterate Earth was headed straight towards us with a 100% probability of collision. Our dependence on gasoline is kind of like that. Buying a Prius would be like building a large bomb shelter: it would show that you probably realize there's some kind of problem, but that you nonetheless have absolutely no understanding of its magnitude. Do you know how much energy it takes to turn bits of iron buried in the Earth into a shiny new Prius? According to an average figure per car, not the Prius specifically, given by Toyota, around 22,519 kwh or 22.5 megawatt/hours , the rough equivalent of 615 gallons of gas. This means that driving a Prius 31,000 miles uses about the same amount of energy as building the thing to begin with! 22.5 mwh would power the average house for over 2 years, it's quite a lot of energy.
One implication is that by buying a used car instead of a Prius you are preventing the use of the equivalent 615 gallons of gas--buying a used 15 mpg beast and driving it 9,000 miles uses less gas than a new Prius with 0 miles on it, making the beast more sustainable and conscientious up to that point. I realize I always pick on the Prius but I don't mean to be too disheartening, the Prius is one of the better options available, even if I think it's not as extreme as it should be. Anybody who buys a new Prius with legitimate environmental concern is now obligated to drive that car into the ground.
New cars aside, one Mythbusters experiment showed a 39% increase in fuel efficiency from drafting a big rig--driving 10 feet behind it. Assuming every car could always draft in such a manner and increase efficiency by 39%, well then the yearly consumption of gas would decrease by a monumental 39%. That's a big assumption, but there's one way it could be realized, and that's with autopilot.
We can't all draft all the time because it is very dangerous to drive at almost any speed 10 feet behind anything, and the reason is simply biological: it takes a measurable and substantive amount of time for information to traverse the nervous system, this phenomenon is commonly referred to as reaction time. When you see brake lights, the light must activate an action potential in your retina, which travels into the brain. Once processed, another signal is sent down the looong path (compared to microscopic neural cells, inconceivably long) to your foot, telling it to press the brake pedal. If a truck moving 60 mph slams on the brakes with you at 10 ft behind, that reaction time is simply way too slow and it's game over. On the other hand, with autopilot brake lights aren't even necessary, the computer in each vehicle would be in constant communication with the cars in front and behind; the vehicles could be 10 ft apart, 2 ft apart, even physically connected like a train without any problem. I imagine the optimum arrangement would be a physical connection for a number of reasons. Of course, if all drivers were computers, the brakes themselves would hardly be needed, especially on the freeway. If you know the status of every car around you, about their planned movements, power characteristics and beyond, less wasteful air friction could be used to decelerate as appropriate, perhaps to allow a car to enter the train, which is itself a task much easier for computers than humans.
It isn't hard to envision that traffic lights would disappear with irrelevance as well, indeed I doubt it would make much sense to sit at an intersection when precise control and rapid, traffic-omniscient computer communications would allow cars of all headings to pass through synchronously. Sure, it will take a while to get used to constantly missing that other car by inches, but abandoning the familiar start/stop/wait process will give tremendous fuel savings, as it is the most inefficient part of driving and why the distinction between city mpg and highway mpg exists. Accordingly, the pace of society will see a new and considerable boost as not only the time between locations diminishes, but we are also free to spend that time doing something other than driving. Not only will we get places fast, ambulances, police, and fire trucks will be able to reach their destinations in the maximum possible time. If that weren't enough, we should expect that we are all a noticeably wealthier as our expenditure on gas shrinks, car insurance disappears, and as mentioned above practically everything drops in price along with energy cost. It's such a win for everyone it feels like cheating, but all of that is just the start.
The first thing people say when they hear of computers or robots driving cars is "but that sounds So dangerous, it would never be safe enough, I would never trust it!" Well, the bleakness of reality readily illustrates the absurdity of such a thought. Think of it this way: the autopilot system could have 5 million accidents a year and that would still be a huge improvement over humans driving cars! There were around 6.4 million car accidents in 2005. 100 people could die every single day in a computer driven car and it would still be safer, because 115 people are dying every day in the current system. One hundred and fifteen people sure seems like a lot, doesn't it? Well, consider that 3,303 people died in car accidents in the month of September, 2001. That month is and always will be bitterly remembered solely for the terrorist attacks that fell the Twin Towers, acts that meant the death of 2,819 people. There is no doubt that 9/11 was a tragedy, but so was 3,303 car accidents. Death by car accident and terrorist attack are fundamentally similar in that the victims of either are generally no less expecting nor deserving the outcome--incidence is practically random. Just because the first figure elicits strong memories and the next is unfamiliar doesn't make the prior any more tragic! Personally I'm inclined to think that every person is more or less equally valuable (namely, invaluable) and thus that each person's death is equally tragic. That being the case, the 2,819 terrorism related deaths on 9/11 are quantitatively about 85.347% as tragic as those due to car accidents in that same month. Alternatively, if we were to assume that only the death of a relative or dear friend qualified as measurably tragic then the majority of people would see that 2,819 random strangers and 3,303 random strangers are pretty close to each other, and we might expect to estimate their relative tragedy as similarly near. Objectivity aside, you would have to be colder than cold to somehow consider 3,303 lives lost any less tragic than 2,819 lives lost regardless of the details, these are all people that could have been you or I, yearning to be alive just like you and I: husbands, daughters, mothers, brothers... neighbors, friends and mentors; they were real people!
So it's established, there were two significant tragedies in the US in September, 2001, now what? Perspective: the 3,303 fatal car crashes in September was actually fewer than that for the two months post and prior, which makes 5 tragic months in a row. If you figure that anything over 2,000 deaths is sufficient to be labeled tragic, every single month in 2001 was a tragedy considering car accidents alone... 37,862 people died. Every single year from 1994 to 2008 has been a tragic year, with an average 37,500 fatal car accidents per year. 1994 is the earliest data I have, but I'm willing to bet the numbers don't improve much by going back further. Over the 14 years that span 1994 and 2008 562,712 people died in car accidents. If instead of happening over 14 years it happened in one day, that day would be about 199.6 times as tragic as 9/11, like the events of 9/11 replayed 199.6 times in one day. 562,712 is 2.5 times the total number of people that died from the atomic bombing of Hiroshima and Nagasaki combined. Know that those weren't the most devastating though--strategic firebombing of Japanese cities killed around 500,000 people, inconceivable yet still fewer.
Cars driven by people are as deadly, if not more, than world wars.
Then how dangerous would it be? Because an intelligent transportation system would need to be implemented everywhere all at once and thus a massive project, breadth and depth of testing at all stages is a certainty. To start, there has been decades of dedicated research on this specific issue, and the state of affairs is amazing (see DARPA's grand and urban challenges). Given the talent inevitably attracted to exceptional challenges (such as top engineers to NASA), a category for which this certainly qualifies, I presume each issue arising throughout development would be deftly handled. Finally, I would expect that some qualified organization would be intimately involved, dictating the requirements and governing the development to ensure safety and reliability, much as the FAA does with all things aerial. An autopilot system made properly as thus, I predict less than a hundred accidents per year from the very start, probably no deaths. With such a system the probability of dying in a car accident would go from frighteningly high to somewhere less than being struck by lightning. The current estimated yearly cost of car accidents is over $230 billion dollars, so... cha-ching! There's an extra $229.98 billion dollars floating around. Nonetheless we would expect the system to improve over time, transforming cars from most dangerous to safest form of transport.
Optimally the typical commuter car should be prepare for transition by being made small and ultra-light, with aerodynamics engineered in terms of chains of cars. The majority of cars should seat one passenger since most often a car carries only one person and any empty seats means wasted energy. With standardized interfacing and characteristics, other vehicle forms would fulfill the need for cargo haulers, high capacity vehicles, and so forth. Ideally vehicles would be public property, eliminating the need for a family to have multiple vehicles for commuting and family outings, but realistically this is the US and people want to own the things they use. Regardless, thanks to the reduced complexity and altogether more efficient vehicle design coupled with energy efficiency savings, a family could afford to own a number of vehicles which nonetheless add up to a fraction of the energy and materials cost of the present steel monstrosities, maybe able even to be stored in the same amount of space. Alternatively a sufficiently large platform could allow for modular passenger compartments; though the platform size would be less than optimum for single passengers, needing only one drivetrain would decrease materials consumption. The subsequent implication is that modular drivetrains could be used instead of modular passenger compartments.
The aforementioned efforts combined would make for an increase in efficiency so marvelous that domestic oil production would actually be sufficient for the first time since the 70's, when it peaked. Since we're making a whole new concept of car, it would make sense to complete the metamorphosis: ditch internal combustion for electrical, pave the road with solar cells, and oil becomes practically irrelevant for the first time since the second industrial revolution. Rather than carry around the really heavy main batteries, leave them stationary and build contact strips in the road so that cars can zip around like full scale slot cars. The relatively lightweight backup batteries would still be carried so that in the case of main power failure the vehicle could still maneuver and communicate safely. With the sum of these modifications, we should expect our busiest roads to give the impression of losing much of the normal traffic--in reality, the same road may have even more traffic, only seeming less because more cars fit in less space for less time. Each intersection would know about every car planning to traverse it from the earliest possible moment, and would assign each car a set of parameters with which it is to use for traversal, including possible alternate plans. Each car would then communicate with every other car assigned to the intersection around the same time to verify that everything works out, a sanity check independent of the intersection. For example, two chains of several cars each plan to travel east and north through the same intersection at the same time. The intersection may dictate that both chains enter the intersection moving 80 mph, the first at 5:00:00 and the other at 5:04:00. The chains verify together and find that they will pass within 6 inches of each other, but that this is an acceptable margin given the wind conditions and other factors. The plan is confirmed with the intersection and each car passes through, deviating a few hundredths of an inch from their predictions--these deviations would then be incorporated back into the prediction model which is distributed across the whole network. Suppose four very long chains travelling in every direction are approaching the same intersection. This time the intersection would probably dictate that the lead cars split and accelerate through such that at any moment there are 4, possibly 8 cars in the intersection, each one missing the other by a hair. Eventually it is expected that the traffic network will maximize efficiency of the whole system in unexpected ways. Maybe previously busy intersections will be used as though there were no crossing, or all but a handful of wide, long, straight thoroughfares will fall into relative disuse.
I am a driving enthusiast, I really love driving. Many days it seems my highest aspiration is to do laps around Laguna Seca in some kind of ultra performance four wheeled vehicle. But despite my pleasure in driving there is no way that I can call the present system workable. It's extremely dangerous, terribly slow, woefully inefficient, and absurdly expensive. The truth is that we have the technology to automate the roads, people have been working on it for decades and the resulting systems have proven reliable even in novel situations many humans might otherwise fail. It might not be perfect, but it's much better, and as I've shown we're so terrifically awful at driving that that's not saying much. The transition is ready to happen, and when it finally does our world will simply become safer, faster, better, and wealthier. The only downside is that it can't be done over night.
Once we finish automating our roads, what's the next revolutionary development? A space elevator. More on that some time.
"Energy Intensity of Computer Manufacturing" by Eric Williams, United Nations University
"How much does electricity cost? What is a kilowatt-hour?"
"How much electricity do computers use?"
Energy Content of Fuels (in Joules), other useful tables
"Weekly U.S. Retail Gasoline Prices, Regular Grade"
"Average Retail Price of Electricity to Ultimate Customers by End-Use Sector, by State"
"Energy to build a car?"
"Most and Least Fuel Efficient Cars "
National Highway Traffic Safety Administration's Fatality Analysis Reporting System
Wikipedia - "Intelligent Transportation System"
U.S. Dept. of Transportation - "Intelligent Transportation Systems Benefits and Costs, 2003 Update"