His essential argument is:
- We have some agreement on two goals: 30% reduction of CO2 output by 2025, 80% reduction by 2050.
- Some countries will not make much reduction, and some countries, like China and India, will expand their CO2 output quite a bit as their huge populations pass through their own industrial revolution.
- Some portions of our western economies will not reduce their CO2 output easily. (I think this is a minor point.)
- The former goal might be achieved through conservation and improved efficiency.
- The latter goal requires that CO2 output from two sectors, transportation and electricity generation, be reduced to zero. Still more will be required, but this is a baseline.
- Once transport and electricity have been reduced to zero CO2 output, conservation in these areas will not improve our CO2 outputs. This is, for instance, why France doesn't bother subsidizing more efficient electric appliances, as many other countries do -- France's electricity is close to zero CO2, so improved electric efficiency doesn't reduce CO2 emissions.
- Therefore, reworking the economy to reduce transportation and electric consumption does not help towards the 2050 goal. To the extent that it costs money that could otherwise be spent on zero-CO2 electricity and transport, it frustrates progress towards the 2050 goal.
Our transport sector currently burns 146 billion gallons of gasoline and diesel every year. In 2050, assuming an increase of 2%/year in transport miles and a fleet efficiency increase from 17 to 23 MPG, it will consume the equivalent of 248 billion gallons of petroleum. If we replace those vehicles with electric vehicles getting 3 km/kWh, those vehicles will consume 3 billion megawatt hours per year. The Nissan Leaf gets 5 km/kWh, so I think an estimate of 3 km/kWh average may be reasonable.
So, the big question raised by Gates' insight is, what can deliver energy like that? To my mind, there are two contenders, wind and nuclear.
The first problem is generation. And the second problem is storage, to cover variations in production as well as consumption.
Here is the generation problem:
The US consumed an average of 470 gigawatts in 2008. The EIA predicts annual increases of 2%/year, so that the average might be 1038 gigawatts in 2050, for the same uses we have today.
The additional 3 billion kWh per year needed to run the electric car fleet, if spread evenly through the year, amounts to 350 GW, which isn't really so bad when thought of in the context of total electric generation. So the grid in 2050 will have to deliver an average of 1400 GW.
1400 average gigawatts could come from 1 million 5 megawatt wind turbines spread over 1.2 million km^2 (at 1.2 watts/m^2). Right now, the US has 1.75 million km^2 of cultivated cropland, so switching US electricity and transport to wind would require a wind farming sector nearly as physically large as our crop farming sector. This is conceivable. After all, 150 years ago most farms had a wind turbine for pumping water. However, 150 years ago that turbine was not the majority of the capital on the farm. These new turbines will cost about $5000/acre, compared with the $2100/acre that farm real estate is currently worth. From an economic standpoint, wind farming would be a much larger activity than crop farming.
The turbines have a 30-year lifespan, so the cost is more than just the initial capital expense. By 2050 all of the turbines installed in the next decade will have worn out, and we'd be into a continuous replacement mode. Cost? $5 trillion in capital outlay for the turbines, another $5 trillion for the infrastructure, and around $160 billion a year (present dollars) for worn turbine replacement.
Here's the storage problem:
The morning commute in any major US city lasts for about 3 hours, with most of the activity in that last hour. The evening commute is longer and more centrally distributed. If we have east-west transmission lines capable of moving most of the commute peak power, we can smooth the U.S. commute peaks into two with four-hour wide centers. Even assuming this transmission capacity, electric consumption during commute hours would be about 500 GW above average.
The current thrust of electric-car research is to improve the batteries so significantly that the cars can be charged overnight and the batteries can provide all necessary power for daytime use. Per vehicle, that's about 18 kilowatt-hours per car, which sounds possible. There is a problem, however: there simply isn't enough material to make these batteries for all our cars. [Edit: I was wrong, there is. Lead-acid batteries require 240 kg lead for 18 kWh. Lithium-ion batteries require 8.5 kg lithium for 18 kWh.]
- Lead-acid batteries would require 60 million metric tons of lead for the 254 million U.S. cars. World production of lead is around 4 million tons/year, and total reserves are around 170 million tons.
- Lithium-ion batteries store 75 watt-hours per pound, and can use about 60% of that (although a five-year life is a goal rather than a deliverable). 18 kWh would require 400 pounds of battery per car, which is physically possible. The U.S. fleet would require 2 million tons of lithium. Total recoverable worldwide lithium is 35 million tons.
Another way to achieve this goal is with nuclear reactors. Thousands of them. A nuclear electric infrastructure would have five big advantages over a wind infrastructure:
- It would cost far less to build.
- It would last 60 years or more.
- It would not be weather dependent.
- It would not require secondary storage (still more cost).
- It would have far less environmental impact (no lakes with tides, no dead birds).
However, if we are to scale up the existing fleet of 104 reactors by over an order of magnitude, some things are going to have to change.
- Nobody really knows how much it will cost to build the next American reactor. We know that it costs the Koreans and Chinese $1.70/watt, and we know that it used to cost about that much in the U.S. If we build thousands of reactors, the cost will drop back into this range or below.
- Most of the new powerplants will have to be cooled by seawater or air, but not fresh water as is most commonly done today. We do not have enough fresh water to cool thousands of plants. Quite the contrary, by 2050 electric power and waste heat from reactors will be used to desalinate seawater for residential use, as is already the case in Florida and some California municipalities.
- Typical reactor sites will have a dozen or more gigawatt-class reactors, rather than the two or three as is common today. Far from being "extra large", gigawatt reactors are right-sized.
- Either very large new deposits of uranium will be discovered, or most reactors will be breeder reactors.