The Next Decade of Solar Energy

Bill McKibben has a new book on solar energy coming out in a couple of weeks, and the New Yorker has published a nice article, drawn from portions of the book, that includes this remarkable passage:

But here’s the current prediction from the I.E.A.: by 2026, solar will generate more electricity than all the world’s nuclear plants combined. By 2029, it will generate more than all the hydro dams. By 2031, it will have outstripped gas and, by 2032, coal. According to the I.E.A., solar is likely to become the world’s primary source of all energy, not just electricity, by 2035.

Sounds impressive, right? But to see just how impressive it is, let’s add some context. (And let’s hope I’m not merely duplicating material that’s in the rest of McKibben’s book.)

IEA projections

By “current prediction from the I.E.A.”, I think McKibben means the Stated Policies Scenario in the International Energy Agency’s most recent World Energy Outlook, published in October 2024. There’s a chart on their web site that shows this scenario’s solar generation through 2035, alongside the other major electricity sources. Here’s a plot of the same numbers, adding in a little more history and the three minor sources:

Sure enough, just as McKibben says, this projection shows solar soaring past nuclear, hydro, gas, and coal, in rapid succession, over the coming decade. A couple of the crossing dates are slightly different from the dates McKibben gives, but the discrepancies are too small to quibble over. The IEA Stated Policies Scenario unambiguously puts solar in first place among electricity sources less than a decade from now. What’s more, it puts wind in second place.

But electricity generation isn’t merely a race to see who comes in first and second. We also want the total amount of electricity to grow, so poorer regions can improve their living standards and everyone can replace polluting combustion fuels with electricity. At the same time, we want to reduce the world’s use of the more damaging sources of electricity. To assess both of these forms of progress, it’s useful to plot the same numbers on a stacked area chart:

Here we can more readily see that in this scenario the world’s total electricity use continues its steady climb. And we can see that as a fraction of total electricity, the share provided by fossil fuels declines from well over half today to about a third in 2035. That’s real progress! Yet either chart also shows that the total amount of electricity from fossil fuels declines only modestly by 2035, with coal dropping by about a third while gas hardly changes. Deep decarbonization of global electricity is a goal for subsequent decades.

Meanwhile, although the IEA projection has solar well out in first place by 2035, solar then accounts for only a quarter of all electricity generation. That’s a spectacular increase from just 8% in 2025, but not enough by itself to provide all the projected growth in total electricity generation—let alone to reduce fossil fuel use.

A reality check

A separate question is whether the IEA’s Stated Policies Scenario is a realistic projection of what will actually happen. I’m not qualified to assess it in any detail, but here’s one reality check. BloombergNEF publishes detailed projections of new solar capacity additions, based on economic and other factors in each regional market. And unlike the IEA, they’re being paid by their commercial clients to get these projections right. Jenny Chase, their solar guru, has generously shared their projection from the third quarter of 2024:

Here we can see that solar’s era of exponential growth has ended. BNEF projects that annual installations will continue to increase, but only linearly, and more slowly than they did in 2023 and 2024.

To translate these new capacity installations into amounts of electrical energy generated, we need to apply a capacity factor: the percentage of peak power that the installed panels generate, on average, over an entire year. For existing solar installations, as of 2024, the worldwide average capacity factor is 12.3%. (To get this number I divided the 2024 world solar generation, 2129 TWh, by the number of hours in a year to get an average solar-generated power of 242 GW. Then I divided by the total installed solar DC capacity as of the middle of 2024, 1960 GW, which I obtained by adding up the annual values in the chart above and an earlier version of it that goes back to 2007. Note that DC capacities are always larger than AC capacities, so DC capacity factors are smaller than AC capacity factors. I'm using DC because BNEF does.)

If we apply the 12.3% capacity factor to BNEF’s predicted new capacity over the next decade, we get an estimated total solar generation that’s slightly above the IEA Stated Policies Scenario:

This near-agreement makes me much more confident that IEA’s solar projection is close to what we can actually expect.

IEA vs. EIA

On the other hand, there’s another highly reputable agency that has projected much lower numbers. The US Energy Information Administration published its most recent International Energy Outlook back in October 2023, with projections for a “reference case” and six “side cases”: high and low economic growth, high and low oil prices, and high and low zero-carbon technology prices. Here I compare¹ the IEA and EIA projections for world solar generation, with EIA’s reference case in red and shading to indicate the full range of the six side cases:

At least one of these two projections is going to turn out to be badly wrong.

Just as troubling as the huge gap between the two projections is the ridiculously narrow range among the seven EIA cases. I genuinely admire anyone who’s willing to stick their neck out with a quantitative projection of what the future may hold. I have even more admiration for a projection that comes with an associated range of uncertainty. But I see no point in suggesting that the uncertainty range is far smaller than everyone knows it actually is.

But with or without its side cases, I don’t believe the EIA solar projection—and I don’t think this is merely because I dislike it. To highlight what’s wrong with it, I’ve calculated the new solar capacity that it implies would be added each year, again assuming a 12.3% capacity factor:

Here the black bars are from the BNEF data shown above, while the red bars are the annual capacity additions implied by the EIA reference case generation data, assuming that the DC capacity factor remains at 12.3%.² The tops of the gray bars indicate the actual capacity additions in the two years that have ended since EIA’s International Energy Outlook was published.

The enormous gap between the two gray bars and the red bars below them shows that EIA failed badly at assessing the state of the solar industry as of October 2023. That’s one reason to distrust EIA’s solar projection. Then, looking ahead to the coming decade, EIA projected a slowdown in the rate of new installations, even relative to the modest 2025 peak in its own projection. Relative to the actual 2025 installation rate, that predicted slowdown would be an abrupt crash, from around 600 MW to just 200 MW installed each year.

I'm open to the possibility that the current solar boom won’t last forever. Perhaps we should expect a slowdown as solar penetration levels increase and the intermittency of sunlight makes further installations less profitable. But I can’t think of why a slowdown would occur so suddenly, at a time when solar still provides less than 10% of global electricity. (In the EIA reference case, solar would reach only 13% of global electricity generation by 2035.)

So I’ll stick out my own neck now and predict that world solar generation in 2035 will be closer to IEA’s Stated Policies Scenario than to EIA’s reference case. Even if I’m only barely right, that would put solar in first place among electricity sources by 2035—if IEA’s projections for the other sources turn out to be correct. But what about those other sources?

Other electricity sources

I think almost everyone would agree that nuclear and hydro will grow only slowly over the next few years, so solar should easily pass both of them sooner rather than later, as McKibben says. As for the fossil fuels, I don’t know enough to form a strong opinion. Let’s just say it wouldn’t surprise me to see coal plateauing (as EIA projects) rather than falling (as IEA projects), and/or gas rising gradually rather than plateauing (as both IEA and EIA project). Then, if the growth of solar is on the low side, it might not pass coal or gas until after 2035. For coal, it could take as much as another decade or so. But I think it’s more likely that solar generation will pass both gas and coal by around 2035.

Incidentally, I think the IEA projection for wind energy is almost certainly too high. I’ll be surprised if wind generation exceeds either gas or coal by 2035. The wind industry is currently facing several challenges, one of which is competition from the booming solar industry:

Total energy

So far I’ve been talking only about electricity—not total energy. But the final claim in the McKibben quote above is that “According to the I.E.A., solar is likely to become the world’s primary source of all energy, not just electricity, by 2035.”

I’m unsure of how to interpret this claim, and I can’t find anything similar to it in the IEA World Energy Outlook. But let’s consider some ways in which it might or might not be accurate. I’ll start by comparing solar to natural gas.

Tables A.12 and A.13 of the IEA World Energy Outlook project that, in the Stated Policies Scenario, world production and consumption of natural gas in 2035 will be 4,422 billion cubic meters. In energy units, according to the conversion factor on page 356, that’s about 44,000 TWh, which is more than four times as much as the projected solar electricity production (10,750 TWh) shown in the first chart above. If solar provides less than a quarter as much energy as natural gas, it sure wouldn’t seem accurate to call it the “primary source of all energy”.

Comparing fossil energy to renewable electricity in the way I just did is unfortunately all too common. It’s also grossly unfair, because I’ve credited solar energy only for the electricity it produces, while crediting gas for the heat we get when it burns. If you use that heat to generate electricity at a power plant, you typically get less than half a TWh of electricity out for each TWh of heat you put in—thanks to the inevitable thermal losses from heat engines.

The traditional remedy for the unfairness of this “direct method” of comparison is to divide the solar electricity generation by the average efficiency of fossil electricity generation, to obtain solar’s “primary energy by the substitution method”: the amount of fossil fuel that you would need to burn in order to produce the same amount of electricity. Nowadays the average efficiency of fossil power plants is about 40%, or 0.40. So solar’s (projected) primary energy by the substitution method in 2035 would be 10,750 TWh / 0.4, that is, 10,750 TWh times 2.5, or about 26,900 TWh. And that’s still considerably less than the 44,000 TWh projected to be coming from gas. (Alternatively, we could multiply 44,000 TWh by 0.4 to get the amount of electrical energy the gas could generate, 17,600 TWh, and compare this to the 10,750 coming from solar. Again, gas comes out well ahead.)

I suppose you could try to argue that the efficiency factor of 0.40 is too high, because some gas is burned for purposes (like heating buildings) that could be served by electrical appliances (like heat pumps) that provide efficiency improvement factors of more than 2.5 (in some cases). But I don’t think it would be reasonable to argue that the average efficiency factor for all natural gas use should be as low as one fourth, especially because combined-cycle power plants have efficiencies somewhat above 40%.

Instead of comparing solar to gas, we could compare it to oil. Again, in the IEA WEO Stated Policies Scenario, oil tends to come out ahead—though the calculation is complicated by the ambiguities of what efficiency factor to use, and whether “oil” should include natural gas liquids, biofuel admixtures, and/or non-energy uses of liquid hydrocarbons such as production of plastics. If you split oil into sub-categories like gasoline, diesel fuel, jet fuel, and heating oil, then each of them separately is projected to provide less energy than solar by 2035, once we apply an efficiency correction as described above. But I’d be surprised if McKibben secretly had that kind of splitting in mind.

More likely, I think, is that McKibben simply made a mistake when he tried to compare solar electricity to total energy coming from oil or gas. Or perhaps he got the talking point from someone else who made a mistake. (Or perhaps I’ve made a mistake in the calculations I’ve just described—in which case I hope someone will tell me!)

There is one more sense in which solar will be the world’s primary energy source in 2035: the same sense in which it has been the world’s primary energy source for billions of years. Each year, the sun delivers just over one billion TWh of radiant energy to earth’s surface and atmosphere,³ warming our planet to a temperature that can sustain life. We tend to take this huge energy influx for granted; neither IEA nor EIA nor anyone else includes it in their official global energy statistics. Even when we design buildings with passive solar heating features, nobody tries to add up the absorbed solar energy on a society-wide scale—probably because there’s no consistent, feasible, or useful way to do the accounting.

I’m quite sure, in any case, that these long-standing uses of solar energy were not what McKibben had in mind when he said solar would become our primary source of energy by 2035. Still, we should remember that all those solar panels we expect to have by 2035 will be intercepting less than 1/10,000 of all the sunlight that hits earth’s surface.⁴ There’ll be plenty of room to expand them further until solar electricity really does provide more total energy than fossil fuels.

---

Update, 19 August 2025

Now that McKibben’s book has been published, I’ve corrected the first sentence of this article to indicate that McKibben’s New Yorker article is not actually an excerpt from the book (as I originally said) but rather is “drawn from” it.

The passage in the book that corresponds to the passage I’ve analyzed here is rather different. It reads:

By 2026, according to the International Energy Agency, solar will generate more electricity than all the world’s nuclear plants combined. By 2028 it will generate more than all the hydro dams. By 2030 it will have outstripped gas, and by 2032 coal. If we manage to stay on the IEA’s net-zero-by-2050 curve, then solar energy will become the earth’s largest source of all energy, not just electricity, by the 2040s.

Notice that the exact dates when solar is projected to surpass hydro and gas are different in this version. But again, these details are too minor to quibble over.

More importantly, the final claim in this version, comparing solar to other sources of total energy (not just electricity), is completely different. Here (in the book version), we’re talking not about the IEA Stated Policies Scenario, but rather about the (far-fetched, in my opinion) Net Zero Emissions by 2050 Scenario. Also, the date when solar is projected to achieve first place is not 2035 but “the 2040s”. This is a far weaker, and less notable, claim than the one in the New Yorker article.

---

¹It’s a little unfair to compare an IEA projection from 2024 to an EIA projection from 2023, but these are the most recent versions from each agency. There is, though, a 2023 version of the IEA World Energy Outlook, in which the Stated Policies Scenario projects solar electricity generation in 2035 at 8,748 TWh (see Table A.3a, page 267). This is substantially lower than the corresponding 2024 projection of 10,747, though still far above the EIA reference case projection of just 4,531 TWh.

²The EIA International Energy Outlook includes projections of generation capacity, including solar. I haven't used those projections here because they seem to be for AC rather than DC capacity, making comparisons to BNEF’s numbers difficult, and because they seem to assume a much higher capacity factor for future installations than for past installations. Let me also note here that there is some minor ambiguity in assigning capacity additions to particular years, because new capacity added mid-year contributes only partially to generation for that year. So the height of any particular red bar on this chart could be adjusted up or down a little, if an adjacent bar is adjusted to compensate.

³Not counting another 450 million TWh that gets reflected back out to space, mostly by clouds.

⁴At 20% efficiency, the panels intercept five times as much radiant energy as the electrical energy they generate, or about 55,000 TWh if IEA’s projection for 2035 is correct. The solar radiation striking earth’s surface is about 750 million TWh per year (less than a billion TWh because some of the billion gets absorbed by the atmosphere). Dividing 750 million by 55,000 gives about 14,000, which I’ve rounded to 10,000 to be on the safe side.

The Energy-Economy Connection Is Complex

I’ve said it in earlier articles, but perhaps not clearly enough: There’s no simple connection between energy use and economic growth.

Perhaps this chart will help:

I’ve plotted U.S. GDP per capita (adjusted for inflation) alongside primary energy use per capita, from 1800 through 2022.  As always with a dual-axis chart, the relative scales are arbitrary (though I haven’t introduced further arbitrariness by starting the vertical axes at nonzero values).

How should we describe the relation between these two data sets?

Obviously there are correlations, especially in the short-term fluctuations. For instance, the economic shocks of 1929, 2008, and 2020 all triggered dips in energy use. When you lose your job, you travel less and buy less manufactured stuff and try to use less heating fuel. But these fluctuations have little to do with the long-term trends.

Equally obvious is the observation that both GDP/capita and energy/capita have increased over the long term.

But the similarities pretty much end there. Most importantly, the overall shapes of the curves are completely different. The rise in GDP/capita has progressed pretty steadily over the full 222 years, with only brief interruptions, resulting in a factor-of-23 cumulative increase. Meanwhile the rise in per-capita energy consumption was confined to the 80-year period between the 1890s and the 1970s, and this increase was by a mere factor of 3.

During most of the 19th century, Americans grew wealthier without increasing their per-capita energy use. That was possible because most energy use in the early 1800s was fuelwood burned in fireplaces for home heating. The fireplaces were inefficient, sending most of the heat up the chimney. By late in the century, more efficient stoves and furnaces were providing most of our heat, and this efficiency gain canceled out our increasing energy use for manufacturing and transportation.

The 20th century brought a multitude of new and expanded energy technologies. The added energy uses outran the efficiency improvements until the 1970s, causing per-capita energy use to increase. Since the 1970s, a wide variety of efficiency improvements have approximately compensated for further added energy services.

I think it’s also fair to say that we’ve been adding energy services more slowly since the 1970s (though it’s hard to quantify “energy services”). By the 1980s, the market penetration of cars and large home appliances was saturating, and Americans shifted more of their spending toward entertainment and other services that aren’t energy-intensive.

Another factor in recent decades has been offshoring of some of America’s most energy-intensive industries (such as metals production) to other countries. The energy data I’ve plotted above doesn’t include energy used in other countries to manufacture goods we import—and to be fair it really should. I don’t know of a comprehensive data set that applies this correction, but Our World in Data has published an analysis by Viktoras Kulionis for a shorter time period, from 1995 through 2020:

The big picture is that the correction is significant, but not nearly as large in percentage terms as GDP growth over the same time period. An accompanying article by Hannah Ritchie points out that the “decoupling” of GDP growth from energy use—even after correcting for offshored production—has occurred in many other rich countries besides the U.S. [I should note that the Kulionis data set does not appear to have been published elsewhere, let alone peer reviewed. But there’s no reason to disbelieve the broader conclusion, which is consistent with what we know about industry’s share of total energy use.]

In trying to draw a close connection between GDP and energy use, some authors point not to historical data but to comparisons between countries:

The overall correlation is quite strong, though I would add a couple of caveats:

• The spread in energy use at any given GDP value is pretty big. Even if we ignore the more extreme outliers, it’s easy to find pairs of countries, like Canada and the U.K., with about the same per-capita GDP yet whose per-capita energy use differs by more than a factor of 3. These variations are often associated with differing climates, population densities, or other characteristics that wouldn’t vary so greatly within a single country over time.
• Although the plotted energy per capita values span a breathtaking factor of 1000, this is badly misleading. At the low end, all the values are much too low because these data exclude energy from traditional biomass fuels. And at the high end, there are a handful of small countries whose energy use is connected less to local lifestyles than to trade-dependent industries (oil, aluminum).

When interpreted in light of these caveats, the data basically tell us that any low-income country can expect to increase its energy use several-fold as it industrializes.

What many people really want to know, of course, is whether further economic growth in the U.S. (and other rich countries) will entail increased per-capita energy use. The answer is that nobody knows. But the pattern of the last several decades suggests that “no” should be our default answer. Those who wish to argue that the pattern will soon change bear the burden of proof.

Solar Thermal Power in the U.S.

What ever happened to solar thermal electricity generation?

As recently as around 2010, many experts thought the future of solar power would be huge arrays of mirrors, focusing sunlight to heat fluids to run steam engines. To replace fossil fuels we would need to carpet the world’s desert plains with these facilities, sending the power to less sunny places over thousand-mile transmission lines.

You can read all about solar thermal power (sometimes called concentrated solar power) at the SolarPACES web site, which has a map of solar thermal generating capacity by country, plus links to detailed data on individual projects. Wikipedia also provides a good overview and a convenient list. The United States and Spain have led the way in developing and deploying these technologies.

At least for the U.S., though, I think we can summarize much of the story in a single chart:

This chart (inspired by a similar chart from EIA) shows the annual electric energy produced by all large-scale solar thermal power plants in the U.S. To put the scale in perspective, the total energy generated by all these plants during any of the last few years—about 3,000 GWh—is comparable to that of a single medium-sized fossil power plant, or roughly a third of what’s produced by a typical nuclear power reactor.

Besides the relatively modest overall scale, I’m struck by the long stagnation during the 90s and 00s, followed by a burst of new construction in the early 2010s.

Here is a map of the locations of these facilities:

Most of these power plants use the parabolic trough design, shown in this old photo of one of the SEGS (Solar Energy Generating System) mirror arrays:

(This photo has been copied so widely that I haven’t been able to nail down its origin. It probably came from a SEGS owner or operator or contractor.) The trough structures are aligned north-south and pivot during the day to focus sunlight on a pipe full of fluid that runs along the focal line. The fluid then provides heat for a nearby steam turbine connected to a generator.

The Ivanpah and Crescent Dunes plants, on the other hand, use the “power tower” arrangement, with thousands of pivoting flat mirrors that reflect sunlight onto a fluid reservoir at the top of a tower:

(Photo of one of the three Ivanpah generators from EcoFlight.org.)

Two of the solar plants—Solana and Crescent Dunes—use molten salt to store some of the thermal energy for use during the hours after sunset.

Three of the plants—Genesis, Ivanpah, and Crescent Dunes—are located on federal (BLM) land. The rest are on private land.

As the map shows, all but one of these solar thermal power plants are located in the desert Southwest. The exception was Florida’s Martin Next Generation Solar Energy Center, which was also unique in being part of a “hybrid” plant, providing supplemental heat to a primarily gas-fired generator. Nevada Solar One and Ivanpah also burn small amounts of gas, as did the SEGS plants, for which the amount of gas wasn’t always small.

Details

Here are the plants’ nominal power capacities (in megawatts):

• SEGS (trough): 400 MW (total for 9 units at 3 sites)
• Nevada Solar One (trough): 75 MW
• Martin (trough): 75 MW
• Solana (trough): 280 MW
• Genesis (trough): 250 MW
• Ivanpah (tower): 392 MW
• Mojave (trough): 280 MW
• Crescent Dunes (tower): 125 MW

The total comes to 1877 MW, of which 1402 MW is still operating as of late 2024. Because the sun doesn’t always shine, the average total power over days and nights and seasons is currently about 340 MW. Again, that’s comparable to one medium-sized fossil plant or about a third of a typical nuclear power reactor.

The plants that are no longer operating are SEGS and Martin. The nine SEGS units had a good run, but began shutting down in 2015. The last of them, SEGS IX, was scheduled to retire in October 2024, according to information submitted to the Energy Information Administration. The sites of the other eight are all now occupied by photovoltaic solar farms: Sunray 2 and 3 at the site of SEGS I and II; Resurgence I and II at the site of SEGS III, IV, V, VI, and VII; and Lockhart Solar at the site of SEGS VIII.

Meanwhile, the solar portion of the Martin hybrid plant in Florida seems to have been a failed experiment. According to monthly generation data submitted to EIA, it worked pretty well for a few years, then much less well for a few more years, and was apparently shut down in October 2022.

The latest Google satellite images appear to show that Martin Solar’s mirrors have been dismantled, leaving only the support structures behind. I can’t find any news reports or other official announcement of Martin Solar’s demise, but I would imagine that operating a solar thermal plant in a non-desert climate was challenging—even in the Sunshine State.

The other big disappointment on the list is Crescent Dunes. It has never come close to generating its expected output for more than a few months in a row, and it has been shut down completely for 45 out of the 107 months since it began operating:

Still, Crescent Dunes has now operated for more than 12 consecutive months, and it sounds like its remaining problems might be solvable.

Impressions

Solar thermal electricity generation is clearly a technology that can work, at least in desert environments with plenty of days of full sunshine.

The challenges in non-desert environments are formidable. Even a thin cloud layer diffuses sunlight too much for mirrors to focus well. And because heating the fluid enough for efficient electricity generation takes time, even occasional cloud cover during a partly cloudy day takes a disproportionate bite out of the electricity output. The map of successful U.S. solar thermal plants confirms that this technology is proven only in desert settings.

For similar reasons, at the latitudes of U.S. deserts, solar thermal plants don’t work well in winter. This is especially true of parabolic trough systems, whose mirrors can’t be tipped toward the low winter sun. At southern California’s Genesis plant, for instance, the average generation in December has been only 18% of the average in June:

The good news, of course, is that we now have an alternative technology that’s better in almost every way: photovoltaic (PV) panels. They work fine under diffuse light. They don’t require any warm-up time before producing full power. They work great in deserts, but they also work pretty well almost anywhere else in the U.S. On average, across the U.S., they generate nearly half as much energy in December as in June. They require far less labor to maintain and operate, and somewhat less labor to install. They take up about the same amount of space to generate a given amount of electricity (substituting the quantum inefficiency of the photovoltaic process for the thermodynamic inefficiency of the steam engine). Best of all, they’re made in factories where steady improvements and economies of scale have brought their price down to almost miraculously low levels.

During the 2010s, after it became apparent that falling PV prices would make PV cheaper than solar thermal electricity, many experts still believed solar thermal would retain one advantage: its ability (when so designed) to store heat for several hours into the evening, generating when electricity demand tends to be highest. But over the last few years even this potential advantage has been erased by the falling prices of lithium batteries—a much more versatile (and lower maintenance) means of storing a few hours’ worth of energy.

And so it’s photovoltaic power, not solar thermal, that has now grown to become a significant component of U.S. (and worldwide) electricity generation. Here’s the same chart as at the top of this article, with the scale shrunk roughly 50-fold to make room to show PV:

In 2023 the U.S. got 3.9% of its electricity from utility-scale PV, 1.7% from small-scale PV, and 0.07% from solar thermal. As the chart shows, the first two are growing very rapidly while the third is in gradual decline.

Here’s a map that tries to show all utility-scale (1 MW and above) solar farms in the 48 states:

The dots are sized by the amount of electricity generated in 2023, but many of them are overlapping or too small to see (for a zoomable version click here). Of the more than 5000 dots shown (more or less) on the map, nine represent solar thermal power plants—all in the Southwest (can you find them?). The PV farms, meanwhile, are abundant not only in the Southwest but also in Texas and the Southeast. Florida alone now has more than 80 solar farms with capacities comparable to that of the defunct Martin Solar. And an increasing amount of PV is being installed in northern states.

There are no more solar thermal power plants currently planned or under construction in the U.S., and the list of those in development around the rest of the world is dwindling. Academic research into solar thermal technologies continues, but the topics of the latest articles on the SolarPACES site suggest that these efforts are shifting toward industrial heating applications rather than electricity generation. (This seems at least superficially promising, because when you use the heat directly there’s no thermodynamic efficiency penalty.)

Perhaps it’s unfortunate that the U.S. chose to invest several billion dollars in solar thermal power plants that came online in the mid-2010s, just as it was becoming clear that this technology had no future. But hindsight is always 20/20, and personally I’m glad that Americans (and others, of course) have been willing to make risky investments in new energy technologies. We need to plant the seeds and then wait to see what will bloom.

(Corrected 26 November 2024 to say that on the map of US solar farms, the number of dots representing solar thermal power plants is nine—not seven—because the dataset separates out the three Ivanpah plants.)

Powering the Next Utah Olympics

It’s now official: Utah will again host the Winter Olympic Games in February 2034.

As we look a decade into the future, it’s natural to wonder how our state will change between now and then. One area of major change—not just in Utah but worldwide—will be energy.

When Utah last hosted the Winter Olympics, in February 2002, 96% of the electricity generated in the state came from coal. Another 2% came from fossil gas, and just 2% came from renewable sources, mostly hydro and geothermal.

(Chart from ember-climate.org)

Twenty-two years later, the changes have been profound. By February 2024, coal’s share had dropped to just 36%, while gas had grown to 45%, and renewables to 19%. Among the renewables, the biggest share was solar power, at 14% of Utah’s total generation (even in February!).

So where will our electricity come from in another decade? A naive extrapolation of recent trends might suggest that coal will vanish from the mix entirely, while solar skyrockets and gas probably stays about the same. Indeed, last year Pacificorp (Utah’s largest electric utility, dba Rocky Mountain Power), announced plans to shut down its two big coal-fired power plants in Utah by 2032. But this year they back-pedaled on that plan, saying they want to keep running the Huntington plant through 2036 and the Hunter plant through 2042. These plants have been running less and less of the time in recent years, and it’s reasonable to hope that they’ll run still less going forward, as renewable generation increases.

(Map from here)

Meanwhile, the medium-sized Bonanza coal plant, which provides power to Utah’s scattered municipal utilities, has also dropped plans to shut down before 2034. So at this point it sure sounds like a substantial portion of the electricity that powers the 2034 Olympics will still be coming from coal. Our state legislature has made it clear that that’s what they want.

Solar power’s share of electricity should continue rising in Utah (like just about everywhere else) for the next several years. New solar farms are currently under construction in several parts of the state, sometimes with battery storage to spread the energy over more hours of each day. By 2034, though, price cannibalization during sunny hours may make further solar development unprofitable. I don’t know how to guess whether solar’s share of Utah’s electricity in February 2034 will be closer to 20% or 50%.

To better understand Utah’s electricity situation, we need to remember that Utah trades electricity with neighboring states. Our largest coal-fired power plant, the Intermountain Power Project, sends most of its power to Southern California. Despite the best efforts of our legislature, I expect that facility’s coal generation to wind down as scheduled in 2025, replaced in part by two new gas-fired generation units. Much of Utah’s solar power also goes to California. Meanwhile, although Utah doesn’t have many good wind generation sites, we do import some wind power from Wyoming and it appears that the amount will continue to increase.

A big question mark for 2034 is geothermal power. It has provided a small fraction of Utah’s electricity since the 1980s, growing slowly to reach 1.6% in February 2024. Now, however, it seems set to grow much more quickly, because Fervo Energy, a startup company that uses advanced drilling and fracking to extract heat from dry rock, has begun work on what they promise will be a 400-MW geothermal facility in the southwestern part of the state, near Utah’s older traditional geothermal plants (whose capacity totals just 84 MW). Again this power will be going to California, incentivized by that state’s mandate for round-the-clock low-carbon (so-called “clean firm”) electricity.

There are some folks who believe Fervo’s drilling technology will make geothermal electricity the Next Big Thing, cheap enough that even states like Utah without low-carbon mandates will choose to buy it. I hope they’re right, but my bet is that geothermal power will not contribute much to Utah’s electricity consumption by 2034.

Utah has no nuclear power plants, and probably won’t have any by 2034. But just over the state line in Kemmerer, Wyoming, nuclear startup company TerraPower seems to be on schedule to build a 345-MW advanced sodium-cooled nuclear plant in time for the 2034 Olympics. Pacificorp is a partner in that project, so some of the power may make its way to the Olympic venues. Meanwhile, by 2026, Pacificorp plans to convert its two coal-fired generation units at the Naughton plant in Kemmerer to burn gas instead. At the huge Jim Bridger coal plant farther east, they’ve converted two of the generation units to gas just this year, and they’re promising to install carbon capture technology on the other two units by 2028 (I’ll believe that when I see it).

In summary, I expect Utah’s electricity-related carbon emissions to be substantially lower in 2034 than they are today, and of course even farther below the embarrassing levels of 2002. There’s still time to get our electricity-related emissions down to near zero by 2034, but our elected officials have shown no interest in that goal and it won’t happen unless they have a change of heart.

On smaller scales, Utahns might still claim that Olympic venues and other individual buildings have zero carbon emissions. A building might have solar panels on the roof, or the owner might subscribe to a program that invests in enough solar or wind power to offset what the building uses. I think these efforts are helpful because they increase the pace at which renewables are deployed. In the long run, though, they contribute to power price cannibalization so they may become more of a bookkeeping gimmick—shifting blame for carbon emissions from some utility customers onto others—than an actual contribution to the total amount of solar and wind that ultimately get built in Utah. (Rooftop solar has the further benefit of reducing the amount of land used by utility-scale solar.)

Besides electricity, there will be emissions from direct burning of fossil fuels. The bulk of the carbon footprint from any Olympics is surely from burning jet fuel, and that won’t change by 2034.

What about ground transportation? Utah is ahead of the national average in its share of electric cars, although that share stood at just 0.9% (of the entire fleet, not just new sales) at the end of 2022. How far it rises by 2034 will depend mostly on national trends: probably to at least 10%, and perhaps as high as 30%?

Utah used droves of buses to carry spectators up to the mountain ski venues in 2002. I’ll be surprised if they can round up enough electric buses to make the fleet fully electric by 2034, but it seems like a worthwhile goal. Now is the time to start phasing in electric school buses and transit buses that can be repurposed for this task during the Olympics. (The greater benefit locally, of course, will be cleaner air for our children and the rest of us to breathe.)

Finally, there’s the fossil gas that we currently use to heat nearly all buildings in Utah. Will any of the indoor Olympic spaces be outfitted with electric heat pumps by 2034? It’s certainly a possibility. The University of Utah has already begun investing in ground-source heat pump HVAC systems, though I don’t know whether its stadium and Olympic athlete housing are among the buildings that have these systems. This technology is cost effective for any large building or campus, and Utah has plenty of folks who know how to do it.

[Revised on 26 July 2024 to add the discussion of individual venues or buildings claiming zero carbon emissions. Revised 28 July 2024 to indicate that the conversion of two generation units at the Jim Bridger plant from coal to gas has already been completed. Revised 29 November 2024 to use more accurate wording in a few places.]

Energy Isn't an End in Itself

Ezra Klein of the New York Times has a recent opinion column advocating for a future in which people everywhere use many times more energy than Americans use today. The column is largely inspired by J. Storrs Hall’s book Where Is My Flying Car?, along with other writers who have been promoting what they call “energy abundance”.

Before I explain how wrongheaded Klein’s column is, let me say what I like about it.

I absolutely agree that “Where’s my flying car?” is a legitimate question. I would even call it a fascinating question—though not, ultimately, a very important question.

I also agree with Klein’s incisive criticisms of Hall’s sociopolitical theories. Rather than review those here, I’ll just refer you to Klein, who’s far better than I am with words.

Finally, I agree that poverty is a huge problem in most parts of the world and that bringing everyone out of poverty will entail a pretty big jump in global energy use.

But by “pretty big” I mean something like a factor of 2, not a factor of 20. The world per-capita energy use rate is currently about 2.5 kilowatts. Ending widespread poverty throughout the world will roughly double that number, bringing it up to the current level in western Europe and Japan, about 4 to 5 kW.

Klein is trying to envision a world in which per-capita energy use doubles three more times, then grows even further by unspecified amounts. The first of these additional doublings would bring the world up to the current per-capita energy use rate in the United States, 9 kW. The next doubling would bring the world up to about 20 kW per capita, the current rate in a few small industry-heavy countries such as Iceland and Qatar. The third doubling would take us to 40 kW per capita, a level of energy use that no country on earth has ever experienced. A recent report from a think tank called The Center for Growth and Opportunity, cited favorably by Klein, defines 40 kW per capita as “energy superabundance” and advocates for such a goal. Yet Klein refers to this goal as “fairly modest”, touting the idea of using still more energy.

So what’s wrong with aspiring to use enormous amounts of energy? Just two things: There’s no known reason to do so, and there’s no known energy source that’s free of unwanted side effects.

Before explaining the first of these points in some detail, let me briefly address the second. The only known energy sources that can conceivably scale up to a significant percentage of 40 kW per capita—that is, 400 terawatts globally—are wind (maybe), solar, and nuclear energy. (Sorry, geothermal resources are too limited to sustainably contribute tens of terawatts.) Wind or solar generation on that scale would take over vast stretches of earth’s surface that many people would prefer to reserve for other uses. Nuclear energy on that scale, whether fission or fusion, would produce enough excess heat to vie with the effects of anthropogenic greenhouse gases on a regional, if not global, scale.

So even “clean” energy sources have unavoidable costs. The question then becomes whether there are benefits that make these costs worthwhile.

Why use more energy?

Merely using more energy provides no intrinsic benefit. I could add massive amounts to my personal energy use just by opening the windows in January and cranking up the thermostat. As Amory Lovins famously said, “People don’t want raw kilowatt-hours or lumps of coal or barrels of sticky black goo. They want hot showers, cold beer, comfort, mobility, illumination.”

Energy is a useful abstraction because we can often substitute one energy source for another: coal for wood, natural gas for coal, wind for natural gas. When we make such a substitution, we can estimate how much of the new source we need by equating its energy content to that of the old source. But real-world complications quickly arise in the actual machinery that captures the energy and transmits it and puts it to use. Energy gets lost as waste heat, in steam turbines and automobile engines and chimneys and incandescent bulbs and open windows. Different technologies can produce the same benefit with vastly different amounts of waste.

So although it pains me, as a physicist, to say this, energy per se is not as fundamental to human needs as Klein and others seem to think.

Klein seems to believe that energy itself is more important than what we use it for. He writes that we can choose among three possible goals as a society: use less energy, or use the same amount, or use more. In fact none of these goals make any sense. It’s wrongheaded to treat energy as an end in itself. The sensible goals are hot showers, cold beer, and so on.

OK, so what about those kinds of goals? Klein lists some particular energy uses that would supposedly become possible if we just had more energy. Much of his list is standard fare among “abundance” advocates: desalination to obtain fresh water; indoor farming with artificial light; capturing carbon dioxide directly from the air. Like his predecessors (at least all those I’m aware of), Klein makes no attempt to do the math to determine whether any of these activities will ever be practical—or desirable—on a scale that would add more than a few percent onto global energy consumption. Desalination, for instance, already provides most of Israel’s municipal water supply, yet adds just 5% to that country’s electricity use.

To this standard list Klein then adds nanotechnology, but here he seems to be misreading Hall. Yes, Hall complains ad nauseam about the slow pace at which nanotechnology has developed in recent decades, supposedly depriving us of all sorts of technological miracles. But nanotechnology isn’t a big energy consumer, and even Hall never suggests that the reason why it hasn’t advanced more rapidly has anything to do with limited energy supplies. Klein implies that it does. Maybe he has some convoluted, unwritten reasoning to back up the claim, but I suspect he’s just confused.

Finally there’s transportation—specifically, aviation and space travel. Hall claims that these technologies have been held back, since the 1970s, by “ergophobia”—fear of using energy. Klein suggests (again without doing any of the math) that if we just had enough energy, virtually everyone on earth would be using it to fly hundreds or thousands of miles a day. Is any of this plausible?

I don’t think so. It’s true that aviation and space travel both require a lot of energy per passenger, but neither is being held back by fuel shortages or even by fuel costs. Less than one percent of the total cost of a Space Shuttle launch was for rocket fuel. For commercial jet flights the fuel cost percentage is higher (typically about 20%), but still not dominant.

The more important role that energy plays in limiting aviation and space travel is indirect. The kinetic and gravitational energies maintained during flight make it intrinsically dangerous—and it’s perfectly rational to be “ergophobic” about that danger. There are ways to mitigate the risks of flying, and the safety record of commercial jet travel is a miracle of the modern world. But risk mitigation comes at a high cost, both in dollars and in convenience: pilot licensing rules, maintenance requirements, air space restrictions, airport security screenings, and so on. People will inevitably disagree over how much risk to accept in exchange for reducing those costs, but let’s not confuse the cost of safety with the cost of fuel.

In summary, it is a logical fallacy to argue that merely supplying the needed energy would give us the world that Klein envisions, in which billions of people commute between continents on a daily basis. Besides the challenges of safety and convenience (and counteracting gravity in the first place), it’s just not clear how many people would voluntarily choose such a lifestyle. Most Americans today could afford to travel more than they actually do, but find life less hectic and more meaningful when they put down roots and spend plenty of time close to home.

Rewriting energy history

Although Klein’s column is ostensibly about the future, his rhetoric relies on comparisons to the past. Such comparisons are all too tempting, because no sane person today wants to go back to the awful living conditions our ancestors had to endure.

But Klein mangles the facts about past energy use.

The most eye-popping howler in Klein’s column is his claim that “Across the 18th, 19th and 20th centuries, the energy humanity could harness grew at about 7 percent annually.” He says this in the context of describing Hall’s book, which contains a somewhat similar claim. Hall’s version is exaggerated badly enough. Klein’s version is far worse.

The words “across” and “harness” are somewhat vague, so there’s no unique way to correct Klein’s claim, but no reasonable interpretation of it is anywhere close to true. If we’re talking about the growth in global primary energy use from 1700 through 2000, then the average annual increase was actually about 1.2%, not 7%. With 300 years of compounding, this means the overall increase was by a factor of about 40, whereas Klein’s 7% would imply a growth factor of (brace yourself!) 650,000,000. (The 1.2% annual increase breaks down into a population growth rate of about 0.8% and an energy-per-capita growth rate of about 0.4%.) Even during the 20th century, when energy use grew more rapidly, the average annual growth rate of global energy use was only 2.3% (again about 2/3 from population growth and 1/3 from energy-per-capita growth). There have been shorter time periods over which world energy use grew somewhat more rapidly, but “7 percent annually” over “centuries” is a ridiculous claim.

Hall, for what it’s worth, arrives at his 7% annual growth figure by looking not at all of humanity but only at “our civilization”—by which he seems to mean the United States since 1800, and perhaps Great Britain during the century or so before that. He plots a graph purporting to show a 2% annual growth rate in U.S. per-capita energy use from 1800 through 1979, although it actually shows an average growth rate of less than 1% over that time period. He then compounds the alleged 2% per-capita energy use growth with an alleged population growth rate of 3% (also exaggerated from the actual U.S. value of about 2% over this time period), to obtain a claimed 5% growth rate in “our civilization’s” total energy use. Finally he adds on a 2% “energy efficiency growth rate” (which he obtains by extrapolating from a much narrower data set), to get a purported 7% growth rate in “usable” energy. (Perhaps Klein intends “could harness” to imply the incorporation of some kind of growing efficiency factor. If so he has not made that clear, nor has he incorporated any such factor into his article’s other energy figures.)

In fact the only way to arrive at anything close to a 7% annual energy growth rate over multiple centuries—for the world or any large portion of it—is to arbitrarily define “energy” to exclude the wood and other biomass energy that accounted for nearly all energy use before the Industrial Revolution. That’s what Hall has actually done, as he confesses on his blog. If humanity’s energy use in 1700 was zero by definition, then it has grown since then by a factor of infinity! Does Klein (who calls Hall’s technical analyses “careful”) realize that he’s propagating this foolishness on Hall’s part? We have no way to tell.

But then Klein does it again, this time with a misleading paraphrase from Charles Mann’s book The Wizard and the Prophet:

Without energy, even material splendor has sharp limits. Mann notes that visitors to the Palace of Versailles in February 1695 marveled at the furs worn to dinners with the king and the ice that collected on the glassware. It was freezing in Versailles, and no amount of wealth could fix it. A hundred years later, Thomas Jefferson had a vast wine collection and library in Monticello and the forced labor of hundreds of slaves, but his ink still froze in his inkwells come winter.

Were King Louis XIV and Thomas Jefferson truly “without energy”? Of course not! The Palace of Versailles has 1200 fireplaces, and surely the king could afford to keep them supplied with wood. The fireplace count at the much smaller Monticello is just 8, but they consumed 10 cords of wood per month, which would have provided about 200 million Btu of energy. The same energy in the form of natural gas would now cost you about $3600. Louis XIV and Jefferson weren’t lacking energy. They were lacking efficient central heating systems that capture most of the energy before it goes up the chimney.

Today’s American homes are not merely better heated and better lit than in Thomas Jefferson’s day. We also have refrigerators, air conditioners, hot running water, automatic washers and dryers, electronic entertainment systems, and a cornucopia of other energy-hungry appliances. And our homes have grown, doubling (at least) in square footage per capita. Yet astoundingly, we use less energy per capita in our homes today than Americans used in 1800. How efficiently we use energy can be more important than how much energy we use.

These historical facts about energy use don’t get discussed much, so perhaps we shouldn’t be shocked that Klein could get them so wrong. Still, I would expect someone who writes for such a large audience to consult a knowledgable expert for some basic fact-checking. It’s unfortunate that the New York Times allows its opinion columnists to spread falsehoods that are so easily refuted.

Why promote energy use as a goal?

When smart people misstate facts, even unintentionally, it’s natural to ask why. So I’d like to end this essay by considering some of the possible incentives, motives, and goals of the “energy abundance” movement. Why are so many writers currently pushing the idea that increased energy use is a moral good—a worthy end in itself?

One incentive for virtually every writer these days is to attract readers, generate clicks, and sell subscriptions or books. To do that it helps to say things that are surprising and provocative, not banal and reasonable. Fuel a culture war whenever possible.

It also helps to say things that readers want to hear. Many Americans are understandably tired of being scolded for using too much energy. They’re eager to believe that the scolds were wrong.

But there are plenty of subjects that can arouse readers’ emotions. Why “energy abundance” in particular?

For any intellectual there is a natural urge to understand broad swaths of the world in terms of a few deep principles: to devise a Grand Unified Theory of How the World Works. The so-called “abundance agenda”, as articulated by Klein’s collaborator Derek Thompson (in an article modestly titled “A Simple Plan to Solve All of America’s Problems”), seems to be a proposal for such a theory. We need (Thompson says) an abundance not just of energy but also of housing, infrastructure, immigrants, COVID tests, and admission slots at elite colleges.

As a physicist I’m all too familiar with the temptation to devise Grand Unified Theories. But I also know that virtually all of these theories turn out to be wrong (or sometimes “not even wrong”, that is, too vague to make testable predictions). The world of human affairs is vastly more complex than that of fundamental physics, so we should be even more skeptical of Grand Unified Theories in the social realm. A theory that works beautifully in one situation can still fail badly in another, so we shouldn’t become too attached to any particular Grand Unified Theory. In the language of Archilochus and Isaiah Berlin, we should try to think like foxes, not hedgehogs.

At least that’s my opinion. The “abundance agenda” folks are obviously trying to be hedgehogs.

Of course there’s also a Grand Unified Theory that’s opposite to the abundance agenda: what we could call the “scarcity agenda” of those who preach about limited resources and living within our means. That way of thinking had its heyday when I was growing up during the 1970s, and is still prevalent among environmental activists and many academics. Like the “abundance agenda”, it’s a correct and useful viewpoint in some circumstances but fails badly in others. So a worthwhile motive for the energy abundance tribe would be to counteract the worst excesses of the energy scarcity tribe. I think that probably is part of their motivation, even if they go beyond it to indulge in their own excesses.

There’s one further motivation that I think underlies some, though not all, of the recent literature on “energy abundance”. The Center for Growth and Opportunity, whose “Energy Superabundance” report Klein cites for its energy use rate goal of 40 kW per capita, was established (at Utah State University in 2017) by a $25 million grant from the Charles Koch Foundation. Although the Center denies that its activities are “directed or influenced in any way” by its donors, Charles Koch’s fortune came from oil and he makes no secret of his libertarian political agenda.

After decades of casting doubt on climate science, fossil fuel interests have learned that that tactic is no longer acceptable in polite society. How, then, can they promote continued growth of fossil fuel use? One effective way might be to spread the general message that we should use more energy.

Klein’s essay, of course, includes the perfunctory caveat that the energy we use in our utopian future should be “clean”. He even alludes to the “daunting” task of shifting our energy economy to “nonpolluting sources”. But the thrust of his essay is that energy has brought us wondrous gifts, and promises us new miracles to come, if only we’ll make every effort to use more of it. Seeing that message in the Opinion section of the New York Times surely brought a smile to the face of Charles Koch.

The Gee-Whiz Energy Graph

In his 1954 book How to Lie with Statistics, Darrell Huff devoted a full chapter to “The Gee-Whiz Graph”—the trick of omitting zero from the vertical axis and stretching what’s left, to make a trend look steeper than it actually is.

We’re currently experiencing an outbreak of people applying this ploy to plots of one particular data set: historical per-capita energy use in the United States. Before this outbreak spreads any further, I’d like to at least document it.

First, here’s a straight plot of the data with a vertical axis that starts at zero (calibrated in two different unit systems for later comparisons):

There’s a lot to notice here! A long plateau at about 4 kilowatts during the 1800s. Dramatic upward spurts between 1897 and 1973, interrupted by temporary plateaus and a deep dip during the Great Depression. A peak at 12 kilowatts in 1979, followed by a gradual downward trend to the present level of about 10 kilowatts. (For comparison, the energy use rate of a typical adult’s food diet is about 0.1 kilowatts.)

If you want to understand this fascinating history, the best place to start is the work of Suits, Matteson, and Moyer, from whom I obtained most of the data.

On the other hand, to quote Huff:

“That is very well if all you want to do is convey information. But suppose you wish to win an argument, shock a reader, move him into action, sell him something. For that, this chart lacks schmaltz. Chop off the bottom.”

Planet Money

In 2013, Jacob Goldstein and Lam Thuy Vo of NPR’s Planet Money did just that, publishing a version of the graph with the bottom chopped off at 50 million Btu/year:

I don’t think these authors were trying to win any arguments, or to sell anything more than an innocent-seeming narrative—that there was a “200-year-long rise in per capita energy use”. But chopping off the bottom of the graph makes the rise appear about twice as dramatic as it actually was.

Besides having a chopped-off bottom, this version of the graph differs in some detailed, substantive ways from my version above. That’s because the Planet Money version uses an older and less complete data set that omits some important energy sources for the years before 1949, as explained here. Including all energy sources raises the 19th century plateau above 110 million Btu per year, and removes the abrupt uptick in 1850, which is an artifact of switching between two different underlying data sources.

To be clear, I don’t blame the Planet Money authors for using an older, incomplete data set, which was the only one conveniently available at the time. But they should have made sure they understood the limitations of their data, and been more suspicious of the 1850 uptick. Without that uptick, essentially all of their “200-year-long rise” occurs over less than 80 years. And by combining incomplete data with a chopped-off bottom, the Planet Money authors made a three-fold increase look superficially like an eight-fold increase. Gee whiz!

Flying Cars

Planet Money’s Gee-Whiz Energy Graph sat dormant, so far as I’m aware, between 2013 and 2020. Meanwhile, in 2018, another plot of the same data appeared. It came from computer scientist and futurist J. Storrs Hall, in his then-self-published book Where Is My Flying Car?. Hall chopped off the graph at 2 kilowatts (60 million Btu/year), again more than doubling the apparent relative increase:

Hall also put a smooth curve on the chart, intended to approximately fit the data up to about 1980. He named this curve after Henry Adams, an American historian who observed in 1907 that the world’s total coal power had been growing exponentially for several decades. Hall’s book says his smooth curve depicts “a 2 percent [annual] growth in actual energy consumed per capita”. To quote Duff again:

“That is impressive, isn’t it? Anyone looking at it can just feel prosperity throbbing in the arteries of the country.”

At first glance, Hall’s “Henry Adams curve” appears to be a pure exponential curve, representing a steady percentage growth rate. But this appearance is an illusion, caused by the graph’s chopped-off bottom. In fact, what Hall has plotted is a 2-percent exponential growth curve shifted upward by an additive constant of 2.5 kilowatts. The book is silent about this sleight of hand, though Hall confesses to it on his blog, where he tries to excuse the trick by arguing that fuel wood—America’s largest energy source until the 1880s—shouldn’t count as a source of energy. When we count all energy sources, the average growth rate from 1897 through 1973 comes to just 1.4 percent—not 2 percent. Over any time period longer than that, the average percentage growth rate was less.

For reference, here is a comparison of the data to some actual exponential growth curves:

No exponential curve fits the data very well, because this is a complex social system—not bacteria growing in a laboratory.

Unlike the Planet Money authors, Hall is definitely trying to sell us something. He says our failure to grow our energy use by 2 percent per year after the 1970s “had real and continuing consequences. There has been a marked drop-off in the technological advances that make a big difference in people’s lives—measured in productivity, health, and, yes, speed and ease of getting around.” Later in the book he writes, “To really reclaim our birthright and an optimistic future, we must get back on the Henry Adams Curve.”

Hall’s book got a publicity boost in November 2020, when Jason Crawford posted an enthusiastic review of it on his Roots of Progress blog. Crawford included Hall’s “Henry Adams curve” graph in his review, and posted the graph again in a follow-up article in February 2021. Crawford is explicitly trying to sell his readers on Hall’s narrative: “We should not seek to merely sustain current per-capita energy usage—we should get back on the Henry Adams Curve and increase it.” Whether we’re talking about energy, population, or technological progress, “Our baseline expectation should be no less than exponential growth.”

Exponential Growth

One quantity that has grown exponentially in recent years is the rate at which the Gee-Whiz Energy Graph keeps popping up. In December 2020, blogger Noah Smith revived the Planet Money version of the chopped-off graph in a Substack post decrying a “stagnation in energy technology” that began in the 1970s (and predicting that cheap wind and solar energy will end this stagnation). Smith doesn’t mention Hall or Crawford or exponential functions, though he does briefly mention flying cars—so the chain of influence isn’t completely clear.

In July 2021, economics journalist Ryan Avent used the same graph in a Substack post. He also mentions flying cars, and his message is similar to that of Hall and Crawford: “Increased energy use is essential to progress.”

Then pundit Matt Yglesias copied the Planet Money graph from Avent into an October 2021 Substack article titled “The case for more energy”. Yglesias doesn’t mention flying cars (or exponential functions) at all, but he does argue that geopolitical and pollution concerns “put us on an energy diet” starting in the 1970s, often preventing us from implementing “cool” inventions such as indoor farming, desalination, and direct removal of carbon dioxide from the air. He expresses the hope that we’ll be able to do all these things in the future, opening up “incredible new vistas”, by deploying zero-carbon energy sources on a massive scale.

In March 2022, Yglesias’s former Vox colleague Cleo Abram released a video based on his “more energy” article. Abram redrew the Planet Money graph, updating it with data through 2020, but continued to use incomplete data from the earlier years and again chose to chop off the bottom:

This graph appears in Abram’s fast-moving video for only a few seconds at a time, so the chance that a typical viewer would notice the chopped-off bottom seems slim. The video’s message is that as long as we can obtain the energy without releasing carbon dioxide, “everyone needs more energy”. As of this writing, the video has over 200,000 views on YouTube.

Over the last year the Gee-Whiz Energy Graph has also proliferated on social media, including tweets from Balaji Srinivasan, Last Contrarian, and Nan Ransohoff. Alec Stapp, co-founder of a new Washington, DC think tank called Institute for Progress, used the Planet Money version of the graph to take a dig at energy efficiency:

What finally provoked me to write this article, though, was the October 2022 special issue of the online publication Works in Progress. That issue, titled “Lost in Stagnation”, consists of six articles that all respond to Hall’s Flying Car book (published by Stripe Press in November 2021) in one way or another. Two of them, by Benjamin Reinhardt and Adam Hunt, feature a replotted version of Hall’s Henry Adams chart:

This version doesn’t even display a 2 kilowatt gridline, so you have to do a mental calculation to realize that the bottom has (yet again) been chopped off.

Both of the Works in Progress articles present this graph without any caveats. Reinhardt’s article, “Making energy too cheap to meter”, echoes Hall’s inaccurate claim in saying “If you plot historical energy use per person over time from the invention of the steam engine, it grew at about two percent per year for hundreds of years.” (This passage originally said the annual growth rate was a preposterous seven percent, as indicated in a “correction” at the bottom of the article.)

So that’s where we stand today: Four separately drawn versions of the Gee-Whiz Energy Graph, all with the bottoms chopped off to make the rise that ended in the 1970s seem bigger than it was. All four versions are being shown and spread to sell us on the idea that energy use goes hand in hand with prosperity, and therefore we should increase Americans’ energy use in the decades ahead.

So What?

I don’t want to exaggerate the harm done by deceptive presentations and interpretations of the Gee-Whiz Energy Graph. Even without the bottom chopped off, the graph shows a striking change during the 1970s, when a pattern of rapid (though lurching) increases in per-capita energy use switched to a pattern of gradual decline. Even though the period of rapid increases lasted less than a century, the changes in Americans’ everyday lives during that period were extraordinary.

But I think it’s notable that making these qualitative points hasn’t been good enough for any of the promulgators of the Gee-Whiz Energy Graph. They’ve all tried to add more quantitative juice to their narrative by displaying the graph in a distorted way.

Here’s my challenge to all such promulgators: Try to rewrite your narrative using an honest version of the U.S. per-capita energy use graph. Start your vertical axis at zero, and use a more complete data set. Acknowledge that the rise in per-capita energy use between 1800 and the 1970s was by a mere factor of 3. Acknowledge that no simple mathematical law can express the rate at which that growth progressed from each decade to the next. Acknowledge that the graph shows multiple time periods when Americans’ material lives improved even while per-capita energy use was not increasing. Then, with all these facts established, ask yourself: Does the rest of your narrative still hold up? If so, great! I’d like to read it. If not, maybe consider whether your worldview is too simplistic.

For what it’s worth, I don’t have my own simple narrative to replace the one I’m challenging. I disagree with a lot of that narrative, but not all of it. I’ll try to organize my thoughts enough to say more in future posts about the complex connections between energy, technology, and prosperity. My point at the moment is just that we can’t have fruitful discussions about these connections if we don’t first agree upon the verifiable facts.

Finally, I’d like to sincerely thank all the “promulgators” I’ve named above for raising fascinating questions and keeping this discussion going. Although I dislike their distortions, at least we agree that the questions are important.

[Revised 8 December 2022 to cite the article by Noah Smith, which I somehow forgot to mention in the original version of this article.]