You’re out of free articles.
Log in
To continue reading, log in to your account.
Create a Free Account
To unlock more free articles, please create a free account.
Sign In or Create an Account.
By continuing, you agree to the Terms of Service and acknowledge our Privacy Policy
Welcome to Heatmap
Thank you for registering with Heatmap. Climate change is one of the greatest challenges of our lives, a force reshaping our economy, our politics, and our culture. We hope to be your trusted, friendly, and insightful guide to that transformation. Please enjoy your free articles. You can check your profile here .
subscribe to get Unlimited access
Offer for a Heatmap News Unlimited Access subscription; please note that your subscription will renew automatically unless you cancel prior to renewal. Cancellation takes effect at the end of your current billing period. We will let you know in advance of any price changes. Taxes may apply. Offer terms are subject to change.
Subscribe to get unlimited Access
Hey, you are out of free articles but you are only a few clicks away from full access. Subscribe below and take advantage of our introductory offer.
subscribe to get Unlimited access
Offer for a Heatmap News Unlimited Access subscription; please note that your subscription will renew automatically unless you cancel prior to renewal. Cancellation takes effect at the end of your current billing period. We will let you know in advance of any price changes. Taxes may apply. Offer terms are subject to change.
Create Your Account
Please Enter Your Password
Forgot your password?
Please enter the email address you use for your account so we can send you a link to reset your password:
The world’s biggest, most functional city might also be the most pedestrian-friendly. That’s not a coincidence.

For cities that want to reduce the number of cars, bike lanes are a good place to start. They are cheap, usually city-level authorities can introduce them, and they do not require you to raise taxes on people who own cars. What if you want to do something more radical though? What would a city that genuinely wanted to get the car out of its citizens’ lives in a much bigger way do? A city that wanted to make it possible for most people to live decent lives and be able to get around without needing a car, even without needing to get on a bicycle?
There is only one city on Earth I have ever visited that has truly managed this. But it happens to be the biggest city on the planet: Tokyo, the capital of Japan.
In popular imagination, at least in the West, Tokyo is both incredibly futuristic, and also rather foreign and confusing. Before I first visited, in 2017, I imagined it to be an incredibly hectic place, a noisy, bustling megacity. I was on holiday and trying to escape Nairobi, the rather sprawling, low-height, and green city I was living in at the time, and I picked Tokyo largely because I wanted to get as far away from Africa as I could. I needed a break from the traffic jams, the power cuts, the constant negotiation to achieve anything, and the heat. I was looking for an escape somewhere as different as I could think of, and I wanted to ride trains around and look at high-tech skyscrapers and not worry about getting splattered by mud walking in the street. I was expecting to feel bowled over by the height of the buildings, the sheer crush of people, and the noise.
Yet when I emerged from the train station in Shibuya, blinking jetlagged in the morning light after a night flight from Amsterdam, what actually caught me off guard was not the bustle but rather how quiet the city is. When you see cliched images of Tokyo, what invariably is shown are the enormous crowds of pedestrians crossing the roads, or Mount Fuji in the background of the futuristic skyline. I expected something like Los Angeles in Blade Runner, I suppose — futuristic and overwhelming. From photos, Tokyo can look almost unplanned, with neon signs everywhere and a huge variety of forms of architecture. You expect it to feel messy. What I experienced, however, was a city that felt almost like being in a futuristic village. It is utterly calm, in a way that is actually rather strange.
And it took me a little while to realize why. There is simply no traffic noise. No hooting, no engine noise, not even much of the noise of cars accelerating on tarmac. Because there are so few of them. Most of the time you can walk in the middle of the street, so rare is the traffic. There are not even cars parked at the side of the road. That is not true of all of Tokyo, of course. The expressways are often packed. Occasionally, I was told, particularly when it snows, or during holidays when large numbers of people try to drive out to the countryside, jams form that can trap drivers for whole days. But on most residential streets, traffic is almost nonexistent. Even the relatively few cars that you do see are invariably tiny, quiet vehicles.
Among rich cities, Tokyo has the lowest car use in the world. According to Deloitte, a management consultancy, just 12 percent of journeys are completed by private car. It might surprise you to hear that cycling is actually more popular than driving in Tokyo — it accounts for 17 percent of journeys, though the Japanese do not make as much of a big deal out of it as the Dutch do. But walking and public transport dwarf both sorts of vehicles. Tokyo has the most-used public transport system in the world, with 30 million people commuting by train each day. This may sound rather unpleasant. You have probably seen footage of the most crowded routes at rush hour, when staff literally push people onto the carriages to make space, or read about young women being groped in the crush. It happens, but it is not typical. Most of the trains I rode were busy but comfortable, and I was able to get a seat.
And what makes Tokyo remarkable is that the city was almost entirely built after the original city was mostly flattened by American bombers in the Second World War. Elsewhere in the world, cities built after the war are almost invariably car-dependent. Think of Houston, Texas, which has grown from 300,000 people in the 1950s to 10 times that now. Or England’s tiny version, Milton Keynes, which is the fastest-growing city in the country. Or almost any developing world city. Since the advent of the automobile, architects and urban planners worldwide have found it almost impossible to resist building cities around roads and an assumption that most people will drive. Tokyo somehow managed not to. It rebuilt in a much more human-centric way.
It may come as a surprise that Japan is home to the world’s biggest relatively car-free city. After all, Japan is the country that gave the world Mitsubishi, Toyota, and Nissan, and exports vehicles all over the world. And in fairness, a lot of Japanese people do own cars. Overall car ownership in Japan is about 590 vehicles per 1,000 people, which is less than America’s rate of about 800 per 1,000, but comparable to a lot of European countries. On average, there are 1.06 cars per household. But Tokyo is a big exception. In Tokyo, there are only 0.32 cars per household. Most Japanese car owners live in smaller towns and cities than the capital. The highest rate of car ownership, for example, is in Fukui Prefecture, on the western coast of Honshu, one of Japan’s least densely populated areas.
And car ownership in Japan is falling, unlike almost everywhere else on Earth. Part of the reason is just that the country is getting older and the population is falling. But it is also that more and more people live in Tokyo. Annually, Japan is losing about 0.3 percent of its population, or about half a million people a year. Greater Tokyo, however, with its population of 37 million, is shrinking by less than that, or about 0.1 percent a year. And the prefecture of Tokyo proper, with a population of 14 million, is still growing. The reason is that Tokyo generates the best jobs in Japan, and it is also an increasingly pleasant place to live. You may think of Tokyoites as being crammed into tiny apartments, but in fact, the average home in Tokyo has 65.9 square meters of livable floor space (709 square feet). That is still very small—indeed, it is less than the size of the average home in London, where the figure is 80 square meters. But the typical household in London has 2.7 people living in it. In Tokyo, it is 1.95. So per capita, people in Tokyo actually have more space than Londoners.
Overall in fact, people in Tokyo have one of the highest qualities of life in the world. A 2015 survey by Monocle magazine came to the conclusion that Tokyo is the best city on Earth in which to live, “due to its defining paradox of heart-stopping size and concurrent feeling of peace and quiet.” In 2021 The Economist ranked it fourth, after Wellington and Auckland in New Zealand, and another Japanese city, Osaka. Life expectancy overall is 84 years old, one of the highest levels of any city on the planet. A good part of this has to do with the lack of cars. Air pollution is considerably lower than in any other city of equivalent size anywhere in the world. Typical commutes are, admittedly, often fairly long, at 40 minutes each way. But they are not in awful smoggy car traffic.

So how has Tokyo managed it? Andre Sorensen, a professor of urban planning at the University of Toronto, who published a history of urban planning in Japan, told me that Japan’s history has a lot to do with it. Japan’s urbanization happened a little more like some poorer countries — quickly. At the start of the 20th century, just 15 percent of Japanese people lived in cities. Now 91 percent do, one of the highest rates of urbanization in the entire world. That rapid growth meant that Tokyo’s postwar growth was relatively chaotic. Buildings sprawled out into rice paddies, with sewage connections and power often only coming later. Electricity is still often delivered by overhead wires, not underground cables. And yet somehow this haphazard system manages to produce a relatively coherent city, and one that is much easier to get around on foot or by public transport than by car.
Part of the reason, Sorensen explained to me, is just historical chance. Japanese street layouts traditionally were narrow, much like medieval alleys in Europe. Land ownership was often very fragmented, meaning that house builders had to learn to use small plots in a way that almost never happened in Europe or America. And unlike the governments there, the government in postwar Japan was much more concerned with boosting economic growth by creating power plants and industrial yards than it was with creating huge new boulevards through neighborhoods. So the layouts never changed. According to Sorensen’s research, 35 percent of Japanese streets are not actually wide enough for a car to travel down them. More remarkably still, 86 percent are not wide enough for a car to be able to stop without blocking the traffic behind it.
Yet the much bigger reason for Tokyo’s high quality of life is that Japan does not subsidize car ownership in the way other countries do. In fact, owning a car in Tokyo is rather difficult. For one thing, cars are far more enthusiastically inspected than in America or most of Europe. Cars must be checked by officials every two years to ensure that they are still compliant, and have not been modified. That is true in Britain too, but the cost is higher than what a Ministry of Transport test costs. Even a well-maintained car can cost 100,000 yen to inspect (or around $850). On cars that are older than 10 years, the fees escalate dramatically, which helps to explain why so many Japanese sell their cars relatively quickly, and so many of them end up in East Africa or Southeast Asia. On top of that there is an annual automobile tax of up to 50,000 yen, as well as a 5 percent tax on the purchase. And then gasoline is taxed too, meaning it costs around 160 yen per liter, or about $6 a gallon, less than in much of Europe, but more than Americans accept.
And even if you are willing to pay all of the taxes, you cannot simply go and buy a car in the way that you might in most countries. To be allowed to purchase a car, you have to be able to prove that you have somewhere to park it. This approval is issued by the local police, and is known as a shako shomeisho, or “garage certificate.” Without one, you cannot buy a car. This helps to explain why the Japanese buy so many tiny cars, like the so-called Kei cars. It means they can have smaller garages. Even if the law didn’t exist though, owning a car in Japan without having a dedicated parking space for it would be a nightmare. Under a nationwide law passed in 1957, overnight street parking of any sort is completely illegal. So if you were to somehow buy a car with no place to store it, you could not simply park it on the street, because it would get towed the next morning, and you would get fined 200,000 yen (around $1,700). In fact, most street parking of any sort is illegal. There are a few exceptions, but more than 95 percent of Japanese streets have no street parking at all, even during the day.
This, rather than any beautiful architecture, explains why Tokyo’s streets feel so pleasant to walk down, or indeed to look at. There are no cars filling them up. It also means that land is actually valued properly. If you want to own a car, it means that you also have to own (or at least rent) the requisite land to keep it. In rural areas or smaller towns, this is not a huge deal, because land is relatively cheap, and so a permit might only cost 8,000 to 9,000 yen, or about $75 a month. But in Tokyo, the cost will be at least four times that. Garages in American cities can cost that much too, but in Japan there is no cheap street parking option, as in much of New York or Chicago. Most apartment buildings are constructed without any parking at all, because the developers can use the space more efficiently for housing. Only around 42 percent of condominium buildings have parking spaces for residents. Similarly, even if you own a parking space, it is almost never free to park anywhere you might take your car. Parking in Tokyo typically costs 1,000 yen an hour, or around $8.50.
This is a big disincentive to driving. Sorensen told me that when he lived in Tokyo, some wealthy friends of his owned a top-end BMW, which they replaced every few years, because they were car nuts. But because they did not have anywhere to park it near their home, if they wanted to use it, they had to take public transport (or a taxi) to get to it at its garage. As a result, they simply did not use their car very much. In their day-to- day life, they used the trains, the same as everybody else, or took taxis, because that was cheaper than picking up the car. This sort of thing probably helps to explain why the Japanese, despite relatively high levels of car ownership, do not actually drive very far. Car owners in Japan typically drive around 6,000 kilometers per year. That is about half what the average British car owner drives, and less than a third of what the average American does.
Parking rules are not, however, the limit of what keeps cars out of Tokyo. Arguably, an even bigger reason is how infrastructure has been funded in Japan. That is, by the market, rather than directly by taxes. In the 1950s and ’60s, much like Europe and the United States, Japan began building expressways. But unlike in Europe and America, it was starting from a considerably more difficult place. In 1957, Ralph J. Watkins, an American economist who had been invited to advise the Japanese government, reported that “the roads of Japan are incredibly bad. No other industrial nation has so completely neglected its highway system.” Just 23 percent of roads were paved, including just two-thirds of the only highway linking Osaka, Japan’s historical economic hub, to Tokyo.
But unlike America, the idea of making them free never seemed to cross politicians’ minds, probably because Japan in the postwar era was not the world’s richest country. Capital was not freely available. To build the roads, the national government formed corporations such as the Shuto Kōsoku-dōro Kabushiki-gaisha, or Metropolitan Expressway Company, which was formed in greater Tokyo in 1959. These corporations took out vast amounts of debt, which they had to repay, so that the Japanese taxpayer would not be burdened. That meant that tolls were imposed from the very beginning. The tolls had to cover not just the construction cost, but also maintenance and interest on the loans. Today, to drive on the Shuto Expressway costs from 300 to 1,320 yen, or $2.50 to $11 for a “standard-size” automobile. Overall, tolls in Japan are the most expensive in the world — around three times higher than the level charged on the private autoroutes in France, or on average, about 3,000 yen per 100 kilometers ($22 to drive 62 miles).
What that meant was that, from the beginning, roads did not have an unfair advantage in their competition with other forms of transport. And so in Japan, unlike in almost the entire rest of the rich world, the postwar era saw the construction of enormous amounts of rail infrastructure. Indeed, at a time when America and Britain were nationalizing and cutting their railways to cope with falling demand for train travel, in Japan, the national railway company was pouring investment into the system. The world’s first high-speed railway, the Tokaido Shinkansen, was opened in 1964 to coincide with the Tokyo Olympics, with a top speed of 210 kilometers per hour. That was almost double what trains elsewhere mostly managed. From 1964 to 1999, the number of passengers using the Shinkansen grew from 11 million annually to more than 300 million.
Sorensen told me about how in the 1950s and ’60s, the trains were a huge point of national pride for the Japanese government, a bit like car industries were elsewhere. “And justifiably! It was a fantastic invention. To say we can make electric rail go twice as fast. What an achievement.” Thanks to that, the railways ministry became a huge power center in government, rather than a neglected backwater as it often had become elsewhere. In rail, the Japanese “built up expertise in engineering, in bureaucratic resources and capacities, and political clout that just lasted,” he told me. “Whereas the road-building sector was weak.” Elsewhere, building roads became a self-reinforcing process, because as more was poured into constructing them, more people bought cars and demanded more roads. That did not happen in Japan. Instead, the growth in railway infrastructure led to growth in, well, more railway infrastructure.
If you visit Tokyo now, what you will find is that the most hectic, crowded places in the city are all around the train and subway stations. The reason is that Japan’s railway companies (the national firm was privatized in the 1980s) do not only provide railways. They are also big real estate investors. A bit like the firm that built the Metropolitan Railway in the 1930s in Britain, when Japan’s railway firms expanded service, they paid for it by building on the land around the stations. In practice, what that means is that they built lots of apartments, department stores, and supermarkets near (and directly above) railway stations, so that people can get straight off the train and get home quickly. That makes the trains more efficient, because people can get where they need to go without having to walk or travel to and from stations especially far. But it also means that the railways are incredibly profitable, because unlike in the West, they are able to profit from the improvement in land value that they create.
What this adds up to is that Tokyo is one of very few cities on Earth where travel by car is not actively subsidized, and funnily neither is public transport, and yet both work well, when appropriate. However, Tokyo is not completely alone. Several big cities across Asia have managed to avoid the catastrophe (cartastrophe?) that befell much of the western world. Hong Kong manages it nearly as well as Tokyo; there are just 76 cars per 1,000 people in the city state. So too does Singapore, with around 120 per 1,000 people. What those cities have in common, which makes them rather different from Japan, is a shortage of land and a relentless, centralized leadership that recognized early on that cars were a waste of space.
Unfortunately, replicating the Asian model in countries in Europe, America, or Australia from scratch will not be easy. We are starting with so many cars on our roads to begin with, that imposing the sorts of curbs on car ownership that I listed above is almost certainly a political nonstarter. Just look at what happens when politicians in America or Britain try to take away even a modest amount of street parking, or increase the tax on gasoline. People are already invested in cars, sadly. And thanks to that, there is also a chicken-and-egg problem. Because people are invested in cars, they live in places where the sort of public transport that makes life possible for the majority of people in Tokyo is simply not realistic. As it is, constructing rail infrastructure like Japan’s is an extraordinarily difficult task. Look at the difficulties encountered in things like building Britain’s new high-speed train link, or California’s, for example.
And yet it is worth paying attention to Tokyo precisely because it shows that vast numbers of cars are not necessary to daily life. What Tokyo shows is that it is possible for enormous cities to work rather well without being overloaded by traffic congestion. Actually, Tokyo works better than big cities anywhere else. That is why it has managed to grow so large. The trend all over the world for decades now has been toward greater wealth concentrating in the biggest metropolises. The cost of living in somewhere like New York, London, or Paris used to be marginally higher than living in a more modest city. That is no longer the case. And it reflects the fact that the benefits of living in big cities are enormous. The jobs are better, but so too are the restaurants, the cultural activities, the dating opportunities, and almost anything else you can think of. People are willing to pay for it. The high cost of living is a price signal — that is, the fact that people are willing to pay it is an indicator of the value they put on it.
Especially in this post-pandemic era where many jobs can be done from anywhere, lots of New Yorkers could easily decamp to, say, a pretty village upstate, and save a fortune in rent, or cash in on their property values. Actually, hundreds of thousands do every year (well, not only to upstate). But they are replaced by newcomers for the simple reason that New York City is, if you set aside the cost, a pretty great place to live. And yet, if everyone who would like to live in a big city is to be able to, those cities need to be able to grow more. But if they continue to grow with the assumption that the car will be the default way of getting around for a significant proportion of residents, then they will be strangled by congestion long before they ever reach anything like Tokyo’s success. People often say that London or New York are too crowded, but they are wrong. They are only too crowded if you think that it is normal for people to need space not just for them but also for the two tons of metal that they use to get around.
The sheer anger of motorists might mean that banning overnight parking on residential streets proves difficult. But if we want to be bold, some of Tokyo’s other measures are more realistic. We could, for example, do a lot more to build more housing around public transport, and use the money generated to help contribute to the network. According to the Centre for Cities, a British think tank, there are 47,000 hectares of undeveloped land (mostly farmland) within a 10-minute walk of a railway station close to London or another big city. That is enough space to build two million homes, more than half of which would be within a 45-minute commute to or from London. The reason we do not develop the land at the moment is because it is mostly Metropolitan Green Belt, a zoning restriction created in the late 1940s by the Town and Country Planning Act intended to contain cities and stop them sprawling outward. But the problem with it as it works in Britain at the moment is that it does not stop sprawl — it just pushes it further away from cities, into places where there really is no hope of not using a car.
Developing the green belt too would not be popular. People have an affection for fields near their homes, and they do not necessarily want the trains they use to be even more crowded. But there are projects that show it is possible to overcome NIMBYism. In Los Angeles in 2016, voters approved the Transit Oriented Communities Incentive Program, which creates special zoning laws in areas half a mile from a major transit stop (typically, in L.A., a light rail station). This being Los Angeles, it is fairly modest. One of the rules is that the mandatory parking minimums applied are restricted to a maximum of 0.5 car parking spaces per bedroom, and total parking is not meant to exceed more than one space per apartment, which is still rather a lot of parking. But nonetheless, it does allow developers to increase the density of homes near public transport, and it has encouraged developers to build around 20,000 new homes near public transport that probably would not have been constructed otherwise. These are small but real improvements.
Ultimately, no city will be transformed into Tokyo overnight, nor should any be, at least unless a majority of the population decides that they would like it. I am trying to persuade them; for now, not everyone is as enamored with the Japanese capital as I am. But NIMBYism and other political problems can be gradually overturned, if the arguments are made in the right way, even in the most automotive cities.
This article was excerpted from Daniel Knowles’ book Carmageddon: How Cars Make Life Worse and What to Do About It, published by Abrams Press ©2023.
Log in
To continue reading, log in to your account.
Create a Free Account
To unlock more free articles, please create a free account.
A new dashboard from the Sustainable AI Group, founded by artificial intelligence alums, estimates the relative energy intensity of proprietary tools.
The rise of artificial intelligence is driving an historic surge in electricity demand that’s boosting fossil fuel use and threatening climate progress. All this electricity doesn’t power AI in some generalized, always-on way, though. Data centers’ energy consumption is a function of the millions of individual queries users submit to AI programs such as Claude and ChatGPT.
When it comes to how efficiently models process those queries and generate responses, AI models are not interchangeable. Some are more like gas guzzlers, others more like Priuses. When a user engages an AI chatbot or AI agent, however, there’s essentially no way for them to know which kind of vehicle they are stepping into. They may know which company built it, and even the precise model name and number, but no AI company has published information about how much energy one model uses compared to another.
In the absence of corporate disclosure from the big three proprietary AI developers — Anthropic, OpenAI, and Google — researchers with the Sustainable AI Group, a research and advisory company, developed a backdoor method to estimate and compare the amount of energy these developers’ models consume. They published their findings on Tuesday in an interactive dashboard that ranks AI programs by energy intensity.
“We think this is an important next step to get some science-based information out there to help folks start making better decisions,” Boris Gamazaychikov, the CEO of the Sustainable AI Group, told me. “We also hope that if the model providers think that this is really wrong, that they can come out and prove it with some actual data.”
In general, the researchers found that larger, higher-capability models, such as Anthropic’s Opus and OpenAI’s Sol, used nearly four times as much energy on average as smaller, nimbler models from those companies, Haiku and Terra. Newer iterations of each model also weren’t necessarily more efficient than their predecessors.
While the group has yet to evaluate the latest models that hit the market during the research period, so far the researchers found that for the same task, the least efficient models can consume more than 30 times the energy of the most efficient models. They also found a significant difference between “chat” sessions, where a user asks an AI chatbot a question, and “agentic” sessions,” where a user asks the AI to perform a series of tasks. A typical agentic session used 27 times more energy, on average, than a typical chat session conducted using the same AI model.
The Sustainable AI Group was founded by Sasha Luccioni, the former AI and climate lead at the open source AI platform Hugging Face, and Gamazaychikov, who previously led AI sustainability at Salesforce. In their earlier roles, the two collaborated on a project called AI Energy Score, which is similar in spirit to the Environmental Protection Agency’s EnergyStar program for home appliances. They developed a method to directly measure the energy efficiency of “open-weight” AI models, or those that fully disclose their inner workings, and published the results in a public leaderboard.
Luccioni and Gamazaychikov founded the Sustainable AI Group because they wanted to give AI users, particularly large corporate users, the tools to understand the relative emissions impacts of proprietary AI models. Gamazaychikov told me that Salesforce had tried to get energy-use data from its AI providers for years to no avail.
Their first hire was Nidhal Jegham, a graduate student at the University of Rhode Island who published a landmark paper last year called “How Hungry is AI?” Jegham and his co-authors developed a method to estimate the energy, water, and carbon effects of proprietary models at the level of a single prompt or query. The paper was accepted by the journal Communications of the Association for Computing Machinery, and the peer-reviewed version will come out in January.
The approach the Sustainable AI Group developed builds on both Jegham’s paper and the AI Energy Score project. The work began with testing open-weight models to see how they perform in realistic deployment configurations and directly measuring their energy consumption. From there the researchers identified mathematical relationships between various open models’ energy use and other measurable statistics, such as their size.
The next step was to take those statistical relationships from the open-weight models and apply them to similarly-sized proprietary models. The problem is, no one knows how “big” proprietary models are. The size of an AI model usually refers to the number of parameters it contains, i.e. the quantity of numerical representations of what the model has learned that it uses to produce a response.
“When we have a closed model, we don't have the model size. We don't have the deployment conditions. We don't have anything, so we need to find things we can observe from this closed model that can reflect its size,” Jegham explained to me. One key discovery, he said, was that “knowledge retention,” or how well the model can remember factual information, is a strong predictor of model size. A company called Artificial Analysis tests models for knowledge retention, so the researchers compared those results to model size for open models and applied the same statistical relationship to estimate the size of closed models.
This is a simplified explanation — there were many other variables and data points that went into the Sustainable AI Group’s estimates. The researchers also had to develop a separate methodology to evaluate Google’s models, since those mostly run on the company’s proprietary “tensor processing units,” rather than the Nvidia chips the researchers’ initial measurements were based on.
The group’s main findings are based on a per-token estimate of each model’s energy use, i.e. the energy required to process the smallest units of data that an AI deals with. Every time you type a question into a chatbot, the model breaks down the words into smaller bits — i.e. tokens — each just a few characters long, usually. The model also first formulates its response in tokens before translating it to text, an image, or whatever you’re requesting; input tokens are less energy-intensive than the tokens the models spit out. The Sustainable AI Group reports each of its per-token estimates as a range to reflect uncertainty.
For now, the firm is keeping its per-token estimates behind a paywall, but it has already started to use them to advise corporate clients in estimating their AI-related emissions, Gamazaychikov said. For example, he mentioned working with Etsy to help the online retailer develop a “model router,” essentially some software that routes a given query to the most appropriate model for the task, taking into account carbon and cost. It’s also partnering with the corporate emissions accounting platform Watershed to explore how to integrate its model-specific energy numbers into Watershed’s system.
Instead of displaying per-token energy use, the Sustainable AI Group’s public dashboard ranks models’ energy intensity per “typical” session, whether chat or agentic. It defines a typical chat session as “a short back-and-forth” with “a question, an answer, and a follow-up or two to refine or clarify it,” whereas a typical agentic session is “an hour or two of the assistant reading files, making changes and checking its own work across a project.” There are also results for a “heavier” or “lighter” session — generally tasks that take more or less time or require greater or fewer back-and-forths with the AI.
The least efficient AI model for both a typical chat and agentic session, per the dashboard, is Anthropic’s Claude Fable 5. A typical agentic session uses 76 watt-hours, according to the Sustainable AI Group’s estimate, or about the amount of electricity it would take to charge four smartphones, per Department of Energy estimates. The most efficient model for a typical chat session was Claude Haiku 4.5, while the most efficient model for a typical agentic session was Open AI’s GPT-5 nano.
Jegham said the point of the dashboard is not to villainize particular companies or models or to argue that more efficient models are superior. He acknowledged that a more complex task may require a larger model, and a larger model is likely going to be more energy intensive than a smaller one.
The ranking is also flawed in that it assumes every model delivers responses with the same amount of verbosity. In reality, some models may use more words, and therefore more tokens, to answer the same question. Jegham gave the example of Anthropic’s Sonnet and Opus models: Sonnet is less energy intensive per token, but it typically requires more tokens for the same task, so sometimes it’s more energy intensive than Opus. The dashboard doesn’t reflect these differences.
While energy intensity is the core of the dashboard’s function, it also includes estimates of each model’s carbon emissions per session. That calculation opens up many more cans of worms, since actual emissions depend on where in the country the hardware that’s processing the AI session is located and what’s powering it. There’s no easy way to know which data center is processing a given AI request. Instead, the dashboard offers users the option to toggle between different emissions intensities to reflect different scenarios — a data center powered by behind-the-meter natural gas plants, for example, versus one located on a relatively clean grid.
A typical agentic session with Claude Fable 5 powered by a behind-the-meter gas plant emits roughly 52 grams of CO2, it says, while a heavy session emits just over 200 grams — equivalent to driving about half a mile in a gasoline-powered vehicle.
I reached out to OpenAI and Anthropic to ask why they don’t publish energy intensity data, whether there are barriers to doing so, and whether they have plans to do so in the future. A spokesperson from OpenAI told me the company relies “on infrastructure partners to operate the data centers that run our models, so we don’t directly collect the underlying energy data. That’s an important consideration in how we assess and provide this information.” Anthropic declined to comment.
Google, on the other hand, has published an energy use estimate for “the median Gemini Apps text prompt in May 2025,” but has not provided an update for subsequent model versions. In response to my request for comment, the company reiterated statements from Cooper Elsworth, a senior technical manager for AI energy, which Google shared with me for a previous story on Watershed’s efforts to calculate AI-related emissions. He said there is no industry consensus for how to measure and disclose the environmental footprint of frontier AI models. He also echoed OpenAI’s comments, noting that gathering accurate energy use data requires “highly advanced measurement infrastructure,” which not all AI providers have access to.
“We believe there is immense value in aligning the industry on comparable metrics to fairly compare and incentivize action,” he said.
Current conditions: Last weekend’s nor’easter caused up to $13 billion in damages across the Mid-Atlantic and Northeast regions of the United States • Hurricane Nolo shut down a major highway on Hawaii’s Big Island • A heat dome forming over eastern Africa is driving temperatures in Juba, the impoverished capital of South Sudan, past 100 degrees Fahrenheit.
At last, right after hopes dimmed, we have a deal. Senate negotiators reached a bipartisan agreement on a package of federal permitting reforms, locking in what Politico described as “the contours of long-sought legislation to speed up approvals for new energy projects in the U.S.” Democratic negotiators Senators Martin Heinrich of New Mexico and Sheldon Whitehouse of Rhode Island told the news outlet they were withholding endorsements of a final deal as “the last five yards” of the agreement are hammered out. Whitehouse cautioned that he needed “more clarity from the Trump administration” on what their easing of the blockade on wind and solar approvals would mean. Neither Democrats nor Republicans released text of the bill, which both parties said should come out this week.
The Nuclear Regulatory Commission is set to issue only its second construction permit for a novel type of nuclear reactor in decades. At 11 a.m. EDT, the agency is scheduled to give the Tennessee Valley Authority the go ahead to begin building what could be the nation’s first commercial small modular reactor, a 300-megawatt unit at the federally-owned utility’s Clinch River site. The project is one of two the Department of Energy is financing to support deployment of third-generation SMRs, a technology based on existing large-scale reactors but shrunken down to force developers to buy more and help the industry bring down the cost of atomic power through repeatedly building the same design. (The second one is Holtec’s expansion of the Palisades nuclear plant in Michigan.) The permit comes six months after the NRC gave TerraPower, the Bill Gates-backed fourth-generation nuclear developer, the green light to start constructing its liquid sodium-cooled reactor at the site of an old coal plant in Kemmerer, Wyoming. The unit planned at Clinch River is a BWRX-300, a boiling water reactor from GE Vernova Hitachi Nuclear Energy that borrows from the technology behind roughly a third of the American nuclear fleet. Boiling water reactors, pioneered by General Electric in the mid-20th century, traditionally represented a competitor to the more dominant pressurized water reactor invented by Westinghouse. By the time Clinch River comes online, North America may already have its first BWRX-300 in operation in Canada, where Ontario Power Generation is building the first reactor at its Darlington plant. TVA has said it plans to bring its debut BWRX-300 online by the end of 2033 at the latest. Yet, despite the forthcoming permit, no start date for construction has been announced.
The NRC, meanwhile, has sought to advance plans to restart the functional reactor at Constellation Energy’s Christopher Crane Clean Energy Center, the facility formerly known as Three Mile Island. Last week, the agency issued an environmental assessment finding no significant impact from plans to begin generating electricity at the plant again. While America’s attempt at restarting a permanently shuttered reactor for the first time are largely going according to plan, regulators are investigating what the Detroit Free-Press described as a “mishap” in the handling of fuel for Holtec’s Palisades nuclear plant in Michigan, which could come online in a matter of weeks. The company said nuclear fuel rods “tipped” during installation, halting the refueling process and forcing plant operators to return to the NRC for approval to retrieve the assembly from within the reactor vessel.
Arevia Power marketed itself as a renewable energy powerhouse led by solar industry veterans. Now, my colleague Jael Holzman reported yesterday, the company is making data centers and gas turbines central to its business. “Arevia is an energy company that delivers reliable and affordable electricity to the communities and utilities we serve,” Ricardo Graf, the company’s chief development officer, told her via email, acknowledging that “in some cases, that energy may be solar; in others, it may be gas.” He added that “yes, we also develop data center projects, but ones with accompanying power solutions to ensure ratepayers are not impacted by the data center’s energy needs.”
The shift in focus comes right as American solar offers a major new business opportunity. Solar panels are aging, and newer technologies are as much as 70% more efficient than those designed and built two decades ago. “All across the United States, solar panels are withering on the vine. Equipment installed 10 to 15 years ago is still capturing sunlight and pumping out electricity, but significantly less of it than when the cells were new,” my colleague Emily Pontecorvo wrote yesterday about a new report examining the potential to swap out the country’s existing panels for new ones. “This is not a story about decline, however, but about growth. America’s aging solar farms represent an opportunity to expand clean energy capacity without using more land — and potentially without having to wait years for new projects to get through the grid’s interconnection queue.”
Sign up to receive Heatmap AM in your inbox every morning:
The TVA isn’t the only government-owned utility making progress on clean power plants. The New York Power Authority — the state electrical company that then-Governor Franklin Delano Roosevelt established in the 1930s and later used as a model for New Deal investments such as the TVA — said Monday that it will take a 51% stake in a 240-megawatt solar plant in the state’s rural northern reaches, according to the Watertown Daily Times. The Rich Road solar farm in Canton, near the Canadian border, will follow a model promoted by progressive legislators with a bill meant to encourage the state to finance and own renewable projects to speed up decarbonization of the grid. Governor Kathy Hochul, a Democrat, has used that authority to support her plans to build at least 1 gigawatt of new nuclear power through NYPA. (That effort, as I told you yesterday, has drawn some blowback from left-wing Democrats who oppose nuclear energy.) EDF Power Solutions North America, a subsidiary of the French electrical giant, will own the other 49% share of the project, which is set to begin construction next year. Once completed, the facility is expected to provide credits to low-income New Yorkers to lower bills.

When I used to think about the Rhine River, the first thing that came to mind was a song off my favorite album from high school. Written and performed by Beirut, the stage name of an American guy who galavanted around Europe making folksy songs that sounded straight out of an American teenager’s romantic notion of an Old World beer hall, the song was called “Rhineland.” Over mournful horns and a plunky mandolin, the song repeats a refrain: “Life, life was all right on the Rhine,” bringing to mind some kind of bucolic interwar existence in an ill-fated era of European history. Two decades later, I can’t tell which has changed more, me or the place I was imagining. The correct answer is probably “both,” but the clearest answer today is the latter. Levels at a key gauge of the mostly German waterway dropped to 1.2 inches below the threshold ship operators use to determine how much cargo their vessel can safely carry down the river without risking damage or running aground, Bloomberg reported. Despite a slight recovery on Monday, the cost of shipping diesel from Rotterdam to Karlsruhe hit a record €260 per ton (equal to just under $296), after more than doubling this month amid the aftershocks of the summer’s record heat waves and droughts.
The latest trouble comes as the Trump administration weighs the merits of a ban on diesel exports. At Heatmap’s Climate Week event last Wednesday, Secretary of Energy Chris Wright ruled out such a step. But Trump said he was “very seriously” considering the step, despite warnings from Goldman Sachs that doing so would raise prices in Europe.
TotalEnergies may be taking up President Donald Trump on his legally sketchy offer of nearly $1 billion to abandon its offshore wind ambitions in the U.S. But the French energy giant — the second-largest European oil company after Shell — sees the energy shock brought on by the U.S. war against Iran as a boon to that very business. CEO Patrick Pouyanne said “high oil prices” are “accelerating electrification,” according to a snippet shared on X by Bloomberg columnist Javier Blas. “We have seen a huge surge in EV sales,” he added, noting that sales are booming well beyond China, in India, Latin America, and Europe. Increased profits from higher crude prices spurred the company to start buying back roughly $5 billion in shares over the next two quarters.
By neutering the Corporate Average Fuel Economy standards, the Trump administration cements the country’s dependence on oil and liquid fuels.
This is Heatmap Daily, a weekday news digest written by our executive editor.
President Trump’s big fuel efficiency rollback is here. This afternoon, the Department of Transportation significantly weakened the Corporate Average Fuel Economy standards, the federal government’s rules that encourage new cars and trucks to get gradually more fuel-efficient over time. Instead of mandating that new cars and trucks hit a target of more than 50 miles per gallon, as the old Biden-era rules had required, new vehicles sold in the U.S. will now need to average only 34.9 miles per gallon.
That target is below the level that most automakers have already achieved in their vehicle fleet. (For reasons too obscure to recount here, the regulatory standard of 34 miles per gallon aligns to real-world gas mileage in the mid-to-high 20s — something my 15-year-old hatchback manages to achieve without much straining.) The new rules also retroactively rewrite the standard back to 2022, meaning that automakers whose fleets once broke the law may now be in the clear.
These changes, in other words, render the fuel economy law, first enacted in 1975, is now moot. But Republicans in Congress had arguably already achieved this last year, when they zeroed out all of the law’s fines for automakers as part of the president’s tax and spending bill. These two changes, taken together, mean that the Trump administration has successfully neutered the U.S. fuel efficiency rules.
We are digging into the rule-making here at Heatmap, and I hope to have more on the documents in the days to come. But one of the lasting ironies of President Trump’s approach to fuel efficiency will be that his own presidency demonstrates its strategic inadequacy.
The Corporate Average Fuel Economy law, after all, did not originate as an environmental policy — climate change had scarcely emerged as a pressing issue in the mid-1970s — but as a national security and economic sovereignty measure. In the aftermath of the oil embargo, American politicians realized that the U.S. economy was far too dependent on oil for its long-term good. This set off a scramble to find new energy sources, prompting a dash back to coal in the electricity sector and a surge in federal R&D spending on alternative energy. (This funding boost eventually created the modern solar, wind, battery, and fracking industries.)
It also led to a successful push to regulate gas mileage. Crucially, this effort did not limit emissions from any one type of vehicle, as the Environmental Protection Administration’s toxic air pollution rules aim to do. Rather, it targeted the average fuel efficiency of cars and light-duty trucks sold in the United States in each model-year. The point was not to regulate any one type of vehicle out of existence, but to increase the country’s overall fuel efficiency over time.
That decades-long effort was never perfect. It created in American statute, for instance, a lasting distinction between cars and trucks, which has bedeviled regulators as SUVs have taken up a larger portion of the new vehicle fleet. But it has also inarguably succeeded: The United States ekes far more value out of every barrel of oil today than it did half a century ago.
Yet the time is ripe to keep making progress. President Trump’s administration has illustrated the persistence of our oil dependence — and the political and strategic problems that it can still engender. Even though the United States has since become the world’s largest producer of oil, the linked and globalized nature of fuel markets means that a supply shock anywhere leads to price hikes everywhere. When an oil crisis arrives — even a largely self-inflicted one, as in the case of the Iran war — then the price of moving things and people rises, the economy suffers, and the president’s popularity falls. Countries can protect themselves from these shocks on a short-term basis by stockpiling oil (as the United States, in fact, does), but they can avoid them only by switching to a far more efficient and electrified transportation system.
President Trump, in other words, may regret the current oil and refining crisis. But by gutting the fuel economy standards — and waging war on electric vehicle incentives more broadly — he is increasing the likelihood that America will face many more crises like it in future years. Consider it his particular gift to his successors.