What is the carbon footprint of an NFT?

It wasn’t long after non-fungible tokens started taking the art world by storm that the backlash began.

For many digital artists, NFTs — essentially blockchain-based certificates of authenticity — have recently come to represent a new opportunity to market their art around the world. Even traditional art auction houses Christie's and Sotheby's have started selling NFTs.

But a major criticism of the NFT-based art movement is that blockchain networks — and in particular Ethereum, the home of most NFTs — generate too much carbon dioxide. The argument is that the popularity of NFTs is now amplifying the damage.

While the controversy over the energy use of popular crypto networks is not new, many of the new NFT bandwagoners were apparently at first unaware. 

When the climate-related implications of blockchains began to sink in, many activists in the art community began to push back against the rush to create NFT art. In one high-profile episode, the popular digital art platform ArtStation quickly reversed a decision to launch an NFT proof-of-concept in the face of heavy criticism. One user said ArtStation would be “actively trying to destroy the earth” if it offered NFTs.

But what is the carbon footprint of an NFT, anyway? That question turns out to be as philosophical as it is technical.

Electricity per transaction?

Perhaps the most prominent estimate of Ethereum’s electricity consumption is from Alex de Vries, who was formerly a consultant at PricewaterhouseCoopers and is now a data scientist at the central bank of The Netherlands.

De Vries maintains a website, called Digiconomist, that estimates both the electricity consumption and carbon footprints of Bitcoin and Ethereum. According to Digiconomist, Ethereum uses 50.09 TWh of electricity per year, which is similar to the consumption of Peru. 

On the day of this writing, the site estimated that a single Ethereum transaction used an average of 86.94 kilowatt-hours (kWh) of electricity — the same as an average U.S. household over 2.94 days, according to de Vries. 

That amounts to 41.3 kg CO2, estimates Digiconomist — comparable to the footprint associated with 91,535 Visa card transactions or 6,883 hours of watching YouTube.

In short, de Vries calculates this by determining mining revenue and converting it to U.S. dollars, estimating how much of the mining revenue was spent on electricity, and dividing that number by the average price of a kilowatt-hour to get how much energy was consumed by the whole network. 

To estimate the energy of a single transaction, he divides the energy of the whole network by the total number of transactions. 

“Estimates” is the key word. As de Vries acknowledged in an academic article he authored this year in the journal Joule, although it’s easy to figure out the computational power of a cryptocurrency network like Bitcoin and Ethereum, that number provides limited information about the machines that are being used and thus how much power is being consumed. 

In the case of NFTs, it is also difficult to account for the energy cost of storing the media files associated with the blockchain token. These files are often stored on the decentralized file storage system the InterPlanetary File System (IPFS).  

Besides all that, energy simply isn’t spent per transaction. While it’s possible to come up with a number, the metric doesn’t reflect how blockchains like Ethereum — the home of most NFTs — actually work. 

NFTs are “minted” as part of blockchain transactions. Like all blockchain transactions, they must be added to a block — a set of transactions assembled by a network node that validates them — and then the block is added to the chain of blocks. 

In blockchain systems like Ethereum and Bitcoin, adding a new block is done via a process known as proof-of-work. Computers connected to the network called miners race to solve a complex mathematical problem — technically called calculating a hash. The first to calculate a new block's hash earns the chance to add a new block to the chain and reap a cryptocurrency reward. 

This is why Bitcoin and Ethereum use so much electricity — and also why they are so expensive to manipulate.

But most important for this discussion: the energy cost of mining a block is independent of the block’s contents. In other words, even if the block were empty it would cost the same amount of electricity. 

Guzzling gas

Nonetheless, argues Memo Akten, a self-described “computational artist engineer” who is also an assistant professor in UC San Diego’s visual arts department, it’s still possible to quantify an individual transaction’s carbon footprint. 

The argument that a single transaction has no associated energy cost is “a common fallacy,” Akten wrote in a recent Medium post. “This is based on a gross misunderstanding of what a carbon footprint is."

According to Akten, the argument is similar to saying that since an airplane uses the same amount of fuel regardless of whether you are on the flight, your presence on the flight doesn't make a difference. “It is true that one person deciding to fly (or not) does not have an immediate effect on emissions,” Akten wrote. “However, there is a footprint associated with a seat on a plane.” 

The calculation is complicated and must account for things like market demand for airline tickets. Similarly, in crypto mining, an increase in demand for space in new blocks might inspire miners to plug in older equipment that uses more energy, Akten wrote.

According to Akten, one way to quantify the footprint of an NFT transaction is by looking at how much “gas” it requires. 

Ethereum transactions require a gas fee, which is denominated in Gwei. One Gwei is worth 0.000000001 ETH. Users must pay a fluctuating amount of Gwei depending on the complexity of the transaction and how quickly they want it finalized. The more complex the transaction, the higher the gas fee.

In Akten’s model, the gas cost of executing a transaction is a proxy for its carbon footprint. Last December, he analyzed the gas prices of tens of thousands of transactions on the popular NFT platform SuperRare. At the time, the average gas price for an Ethereum transaction was around 45 Gwei. NFT transactions cost significantly more.

Akten found that the average NFT cost 260 Gwei to mint, 160 Gwei to sell, 95 Gwei to transfer, 75 Gwei to bid, and 22 Gwei to cancel a bid. 

Akten’s calculations formed the basis for an NFT carbon calculator he created, called “cryptoart.wtf.” Users could enter a URL to an NFT listed on an NFT marketplace and receive feedback about the carbon footprint of that NFT. However, he shut the calculator down in March because, he said, individuals had used his calculator “as a tool for abuse and harassment.” 

Now the most prominent crypto carbon calculator is carbon.fyi, which was created by Offsetra, a sustainability consulting organization that also provides carbon offsetting services. Though it uses a gas-based carbon emissions calculator like cryptoart.wtf, it is designed to analyze Ethereum transactions more generally as opposed to NFTs specifically. 

The philosophy of NFT carbon costs

In fact, there are a number of different approaches to estimating the energy cost of a blockchain transaction — and the estimates they produce vary.

Artist and coder Kyle McDonald recently ran thirteen models — including Akten’s “model_gas” method — in an attempt to estimate how much energy is associated with individual Ethereum transactions. McDonald’s models came up with electricity cost estimates that ranged from half to five times what Akten found. 

Each of the models captures “a different story about how responsibility should be allocated,” says McDonald. He adds that the gas-based models that Akten and Offsetra use assume that emissions are fixed 0.29 grams CO2 per unit gas, which is their biggest weakness. “This is not incorrect, but it is an approximation given that emissions vary over time.” 

In fact, it is very challenging to estimate the carbon dioxide emissions associated with these networks because it’s so tough to pin down the energy mix used by the miners at any given time. 

Ultimately, McDonald says, the question of how to best the carbon footprint of an NFT is a philosophical one, as it involves value-based decisions regarding what exactly should be counted and what can be ignored.

McDonald’s own philosophical reasoning focuses on “hashcount,” or the number of hashes that miners have calculated. That’s because miners are using hardware that varies in energy efficiency, and the energy sources they are using also vary. Since hashcount is the least variable proxy for the energy cost, it can be used to allocate “per-transaction responsibility,” he says.  

“Transactions [on the Ethereum network] have been compared to passengers on a train: a fully loaded passenger car is typically more than twice as heavy as the passengers, and the number or kind of passengers is not as important as the weight of the train itself,” says McDonald. “If the ‘passengers’ are negligible or even irrelevant for calculating the hashrate and hashcount, then how should we allocate hashcount responsibility?” 

This question does not have a purely technical answer, argues McDonald, “so we need to come to a social consensus.”

One solution, he says, would be to take cues from the established field of carbon accounting, where responsibility for indirect emissions is typically estimated based on how much money is spent on a given product or service.

“The current discussion seems to be limited to gas, but I would like to see a focus on fiat-equivalent gas fees instead, because this is a known pattern in carbon accounting,” McDonald says. 

In the case of Ethereum, this would entail first finding the total revenue of Ethereum, and then finding the total gas fees associated with a given transaction. Divide the second number by the first to get a percentage that can also represent the transaction’s share of carbon emissions.  

Either way, McDonald argues that the current concept of carbon calculators for Ethereum transactions inappropriately shifts blame away from miners. 

“To hold miners responsible, we will need more action in the political arena to pressure mining activity away from high carbon intensity regions,” he said.