Breakfast Bytes Blogs

It's Earth Day today. So let's take a look at what it would take to make everyone's Earth Day wishes come true—carbon neutrality, sometimes called Net Zero, by 2050.

As it happens, I'm not especially worried about carbon dioxide. As I pointed out in my post, All Models Are Wrong; Some Are Useful, all the catastrophic forecasts are based on models that have proven to have zero predictive power. They can't even hindcast correctly and explain why in the 20th century, the temperature of the earth warmed to the hottest it has been recently (the dustbowl in the 1930s), then cooled until the early 1970s, then warmed again until the late 1990s, and has been flat since. The earth has been warming since the Little Ice Age (around 1650), and every doubling of CO2 raises the temperature by about 1°F. But in this post, I'm going to ignore all that and just focus on what it would take to decarbonize the world's economy. That is, if we convert all our energy use to sources that don't produce any carbon dioxide, what would it take?

One thing that everyone agrees on is that using less energy is good. Using a lot is wasteful of everything. I don't know how big a contribution to world energy demand a cut of IC power makes. After all it is not steel production, but less is good. And datacenters are big power hogs, and there are going to be more and more of them.

Cloud hyperscalar datacenters are a major part of the glue that holds society together. But they are estimated to use 90 TW/year, about 34 coal-powered plants. This is an area where Cadence and the semiconductor industry can make a major contribution. The semiconductor manufacturers reduce power with every new node as we race to the end of the silicon roadmap. Cadence can reduce the power on-chip starting at the architectural level with tools like Stratus for high-level synthesis, Joules at the RTL level, down the lower-level tools like Voltus and system-level thermal analysis with Celsius. A lot of the power saving comes from increasing accuracy so that guard-banding due to excessive pessimism does not contribute to wasted power. The amount of electric power in the world that is consumed by semiconductors is only going to increase, it seems. With our help, individual chips might consume less power, but there are going to be so many more of them. Just think about cars, where the semiconductor content is forecast to go from 10% today (and just the radio if you go back a few decades) to 50% in 2030 or so. For obvious reasons, we don't want to waste energy, either in the power consumed by chips, but just as importantly, in the power consumed by HVAC systems to get the heat back out of datacenters.

Net Zero in 10,000 Days

There are 10,115 days between today, 22nd April 2022, and 31st December 2049. So that is the time we have to convert everything. How much are we talking about?

The World Energy Council’s business-as-usual 2050 estimate of fossil fuel use in 2050 is about 193 PWhr/yr. So if we are going to go net zero by 2050, we need to convert 193 PWhr/yr in 10,115 days. Let's even up the units a little and get rid of the weird hours/years/days in that sentence. There are 8,760 hours in a year (ignoring leap years), so that is 22TW of new construction before 2050. That turns out to be around 2 GW per day of new non-carbon-emitting power stations (2.17 GW, to be more precise).

I think that the best hope to do this is to build a lot of nuclear plants, and that makes the math simpler. We can look at wind and solar later, but, as I pointed out in my post How the Electricity Grid Works you still need backup power stations to stand behind solar and wind for when the sun doesn't shine and the wind doesn't blow, or some other storage technology that hasn't been invented yet.

According to the US Government, "A typical 1,000-megawatt nuclear facility in the United States needs a little more than 1 square mile to operate". So if we do it all with 1,000 MW nuclear power stations (1,000 MW = 1 GW), then we need to build two new nuclear power stations every day until 2050 (and find 20,000 square miles to put them on). Of course, that assumes, unrealistically, we already have some under construction so that two come online today, and two tomorrow, and so on. One big challenge in the US is permitting. Since the Nuclear Regulatory Commission was created in 1975, they have never approved a new commercial power station that was not already in development. The most recent denial was in January of this year, when they denied Oklo Power's application to build and operate the company's Aurora compact fast reactor in Idaho. Some plants are a lot bigger than 1 GW. For example, Diablo Canyon, which California is shutting down, is 2.2 GW. So maybe we only need to build one nuclear plant per day on the Diablo Canyon scale.

So what about wind and solar?

First, there is the capacity factor. For wind, that is about 1/3, so the actual power generated is about 1/3 of the label on the can. So to get our 2 GW we will need to build wind farms with about 6 GW of nameplate capacity every day. That is about 3,000 turbines each generating 2MW every day between now and 2050. A total of over 30 million windmills by 2050.

Solar? According to the National Renewable Energy Laboratory (NREL), solar generates about 9 W/m² of panels averaged out over a whole day (including the night), depending a lot on the location: Minnesota (snow on the panels) versus the Mojave desert is unfair competition. The nameplate number is about 150 W/m² in the middle of the day in a desert, but I think the NREL numbers come from what has actually been installed in the US. Let's assume that we'd put grid-scale solar in good places and go with a more optimistic number, say 25 W/m², meaning that our 2 GW per day requires 80 sq km or 30 square miles of solar panels per day. A total by 2050 of over 300,000 square miles of solar panels. Plus, all the transmission infrastructure to connect them to the grids.

Note that these numbers are world energy consumption, not just the US. So we are not looking at all these windmills, panels, or nuclear plants in just the US. France gets about 3/4 of its power from its 56 existing nuclear plants. That's almost the same number of plants as the whole US, with about five times the population and a zillion times more land area. France could build about 20 more nuclear power stations and be at net zero, and they have actually announced plans to build about a dozen more. US energy consumption is about 16% of world consumption, so you can reduce the above numbers by 5/6ths to get to the US numbers. So only 500 turbines per day, or only 5 square miles of solar panels. And one nuclear plant per week-ish instead of one per day.

These numbers are looking at world total energy use, not electricity only. But there are two big trends that mean that the amount of electricity we will need is much more than any sort of business-as-usual analysis. They don't change the overall power usage, but they bias it to electricity. To get rid of gas/petrol use in vehicles, we need all vehicles to be electric, and this is estimated to require about a 60% increase in electrical power generated and delivered. The next big transition is converting all natural gas heating in domestic and industrial settings to electric. I can't find a good number for this, but a lot of the world is cold in winter (not Silicon Valley, we barely need heating here) and currently consumes more energy as gas than electricity. Then there is converting industrial plants from gas and coal to electricity. I'm not sure if that is even possible. For example, coal, in the form of coke which is pure carbon, is used in steelmaking. My educated guess is that we will need at least 3 times as much electricity by 2050 than we use today.

Moving away from the US, here is a report on what Net Zero would mean for the UK. One conclusion is that the electricity grid needs to be 2.7 times as big to cope with the switch to electric vehicles and heat pumps, so close to my guess of 3. Here is a quote from the Executive Summary:

The cost to 2050 will comfortably exceed £3 trillion, a workforce comparable in size to the NHS [1.5 million staff] will be required for 30 years, including a doubling of the present number of electrical engineers, and the bill of specialist materials is of a size that for the UK alone is comparable to the global annual production of many key minerals.

I'm a bit skeptical it will cost that much. I remember the UK converting from town gas to natural gas, which required a mobilization and training of people to change every burner on every gas unit (boilers, ranges, and so on). I was a teenager, so I didn't pay much attention, but it appeared to happen without too much problem.

I tried to discover how many power stations there are in the world today. The two datapoints I found were 62,500 in 2012 (Washington Post quoting a GE report that is a broken link) and 30,000 in 2020 (World Resources Institute). So let's say the error bars are pretty big, but it is at least the same order of magnitude as the 10,000+  2 GW stations that the calculation above produced. I think that the numbers depend on how many of the smallest plants get included. For example, the US Energy Information Administration reckons that in the US, there are 11,070 plants...but that is all plants of more than 1 MW (tiny, much less than a single modern wind turbine). I fell into a similar trap in my post Moss Landing. Where I said that, I was surprised to discover that there were 6 pumped storage schemes in California. The smallest has a capacity of 25 MW. I actually don't really know what that means since that is not a capacity, which would be measured in MWh. 25 MW for three months is a very different capacity from 25 MW for 2 hours. The largest is 1500 MW on the same unclear scale but 60 times larger.

Summary

So I think it is clear that, whatever the good intentions, the world will not be carbon neutral by 2050. If we just consider the US, it is clear we will not be building a new nuclear plant every week between now and 2050, or equivalent numbers for wind/solar.

There is a saying that the best time to plant a tree is 20 years ago. And the second-best time is now. If you want to be carbon neutral by 2050, the best time to start was 20 years ago. Or at least start now.

Sign up for Sunday Brunch, the weekly Breakfast Bytes email.

.