The non-fungible tokens (NFTs) market has seen extraordinary growth – valued from $82 million in 2020 to $37 billion within the first 4 months of 2022. Christie’s sale of Beeple’s artwork ‘Everydays: The First 5000 Days’ for $69 million, the most valuable NFT sale to date, captured public attention and catapulted NFTs into the mainstream, whilst spotlighting its underlying technology, blockchain. However, blockchain has been called out for its intensive energy use and negative environmental impact. Though there are more sustainable blockchain solutions with potential economic, social, and environmental benefits, the most popular blockchain platforms, Ethereum and bitcoin, still use mechanisms leading to high carbon footprints. Therefore, any responsible organisation considering using blockchain platforms will need to understand and evaluate the environmental impact of its “on-chain” activities.
Christie’s has a strong commitment to sustainability and in 2021 committed to the SBTi’s science-based emissions reduction targets aligned with the 1.5°C global warming trajectory. With the sale of ‘Everydays: The First 5000 Days’ dramatically signalling the arrival of NFTs in the art world and its presence for the foreseeable future, Christie’s recognised the potential environmental cost to blockchain and commissioned Avieco to assess the impact of its blockchain-related activities. The results are featured in Christie’s Environmental Impact Report 2021.
Blockchain is a technology that can provide a decentralised, transparent, and secured database solutions that can resolve fundamental issues in the ownership and exchange of traditionally intangible assets such as digital artworks and digital currencies. Constructed as an immutable and continuously growing database, managed by a peer-to-peer network of computers, also called “nodes”, a blockchain is set up using self-governing protocol to ensure that all nodes function correctly and to prevent all malicious or fraudulent attempts, maintaining the security of the blockchain.
Blockchain enables the creation and transferring of cryptographic tokens, digital units of value recorded on the blockchain database. The most popular token types to the public are cryptocurrency and NFTs. Cryptocurrency is a digital currency issued by each blockchain to finance its activities and enable exchanges of cryptographic tokens. While cryptocurrency tokens are interchangeable, i.e., one BTC can be exchanged for another BTC, non-fungible tokens (NFTs) have unique identification codes that distinguished one from others. NFTs run on smart contracts, which are programmes that automatically execute once all set conditions are met. Digital wallets, or crypto wallets, are like bank accounts that hold and secure crypto assets using cryptographic keys.
Blockchain works based on the consensus among all member nodes in the validation of new data added to the database. When a user sends out a transaction request, it is broadcasted to all nodes in the network. The nodes will validate the request following the predetermined, self-governing protocol called a consensus mechanism. Consensus mechanism dictates how a blockchain choses nodes to validate a set of transactions and record them in a new block. The new block is verified by other nodes and added permanently to the blockchain once all nodes reach unanimity. The new block can be rejected if any data discrepancy is detected, and the validator node can receive a penalty.
Ethereum and bitcoin use a well-known consensus mechanism called ‘proof-of-work’. It selects a validator, or “miner”, through a process called “mining”. Mining requires miners to compete using intensive computing power to generate guesses at a random number, called “hash”, to earn the right to validate transactions, create the new block, and receive cryptocurrency rewards. Once the new block is created, other nodes verify its legitimacy, any faulty block will be rejected and there is no reward for miner. For the best chance to win, mining hardware is always in operation and consumes a tremendous amount of energy, most of the energy consumption in a proof-of-work blockchain. To corrupt the verifying system, one needs to control the majority computing power of the network, which can equate to billions of dollars investments in mining hardware and electricity. Therefore, it is extremely expensive for miners to get penalised with no reward and the investment to surpass most of the total computing power outweighs any financial gain of a hack. Mining is energy intensive and costly by design to automatically protect blockchain security.
Beside proof-of-work, proof-of-stake is growing in popularity as a low-carbon alternative as Ethereum blockchain announced the move to proof-of-stake for Ethereum 2.0, expected to complete in early 2022. With proof-of-stake, validators lock in cryptocurrency as collateral. Depending on the “staked” amount and other factors, an algorithm is run to decide a validator for the new block. This validator will receive rewards for the validation and can be penalised and lose the stake if any faulty activity is detected. Staking process does not require any competition on computing power, therefore, moving to proof of stake is expected to reduce Ethereum’s energy consumption by 99.95%.
Avieco proposed to conduct a life cycle assessment (LCA) and create a bespoke LCA calculator to quantify the GHG emissions associated with on-blockchain transactions carried out in the sale of ‘Everydays: The First 5000 Days‘, all conducted on Ethereum. To determine Christie’s emissions accountability under the GHG Protocol in terms of carbon reporting, we defined the scope of the assessment by mapping out a 5-step value chain for an NFT sale (fig.1), starting with the NFT minting (creation), to the NFT transfer to the intermediary’s wallet, the cryptocurrency payments from buyer to the intermediary, the NFT transfer from the intermediary to the buyer, and ending with the cryptocurrency payment to the consignor. The 8 transactions in the sale of ‘Everydays: The First 5000 Days‘ cover all 5 steps of this value chain. Each transaction in the assessment starts with a request being posted to the blockchain network and ends when the request is validated, executed, and recorded on the blockchain. The LCA includes all upstream and downstream emissions of the equipment used for this activity. We wanted to compare the carbon intensity of Ethereum with bitcoin, most popular for cryptocurrency transactions, and with Tezos, a rising proof-of-stake blockchain in the NFT market, therefore added 2 hypothetical cases to the assessment, one where all cryptocurrency payments were conducted on bitcoin, and the other where all the minting and transferring of the NFT were conducted on Tezos.
In a public blockchain, collecting any data unregistered on-chain is inherently difficult as the network of nodes can locate anywhere and use electricity generated from fossil fuels or renewable sources. As we cannot trace the data pathway in blockchain, we used a commonly used top-down allocation approach to assess the emissions of each on-blockchain transaction, starting with estimating the GHG emissions of the entire blockchain network, then allocating these emissions to each transaction. Estimating the total emissions of a public blockchain is complex with many data blind spots, therefore relies on multiple assumptions. The emissions apportioning method is also based on the fundamental assumption that financial rewards are the only drive for validators to create new blocks. The method is built on metrics correlated to the financial rewards issued to execute a transaction.
A blockchain’s total emissions covers the emissions from the generation of the electricity consumed by all nodes’ computers, the production of hardware used, and their end-of-life treatments. Whilst calculating the GHG emissions of public blockchains has proven to be complex, we have leveraged a wide range of data sources, including academic, grey literature, and databases to overcome these data challenges and ensure the highest possible data quality.
For bitcoin and Ethereum, which have attracted the most attention and effort to estimate their energy consumptions, we used the study by Digiconomist, founded by Alex de Vries, a recognised blockchain expert. Its approach assumes that all miners will continue as long as mining remains profitable, that a certain percentage of miners’ profits is spent on procuring electricity, and uses an average price per kWh to estimate a range of the total energy consumption. For Tezos, we used the estimated energy consumption modelled bottom-up from the energy demand of popular ‘baking’ (validating) hardware by Tezos Agora.
Inferring the emissions from a blockchain energy consumption requires measuring the carbon intensity of each miner’s electricity supply. Since this level of data accuracy is currently not achievable, country-level locations of miners are estimated using existing research and matched with the corresponding average electricity grid emission factors, sourced from IEA Emission Factors 2020.
While emissions from electricity can account for over 99% a proof-of-work total emissions, hardware production and end-of-life treatments should also be considered, especially given the growth of mining farms, some use thousands of specialised mining units with the average lifespan estimated to be 1.5 years. To measure the emissions of hardware used by validating nodes, we used results from various studies by CoinShares, Kryptex, Tezos Agora, and the emissions conversion factors from ecoinvent 3.8 database and other academic LCA studies.
To allocated the blockchain’s total emissions to transaction-level, we assumed that the emissions share associated with each transaction is divided proportionally to the financial reward it generated. This reward includes transaction fee, paid by the user who requested the transaction, and block reward, issued by a blockchain to incentivise validators. Block reward per transaction is calculated using the ratio of gas use or size of the transaction over the block’s gas use or size, adjusted for the fluctuating values of cryptocurrencies. We adopted this approach from the work of artist and coder Kyle McDonald and tested the rigor of this method with blockchain experts including the co-founders of blockchain platforms Tezos and Palm NFT Studio.
The allocation of a blockchain’s total emission to each transaction is built on the basis that financial rewards are the driver for validators to invest resources in validating and creating new blocks. One type of reward is transaction fee, paid by the user who request a transaction and is a tool intended to prevent spamming. A user can increase the transaction fee over the network average to appeal to validator to prioritise the transaction, especially when the network is congested with transaction requests. Another reward type, block reward, is a set amount of cryptocurrency, newly issued by the blockchain to a validator for each block created. Currently, bitcoin blockchain sets a block reward of 6.5 BTC per block, while Ethereum sets a reward of 2 ETH per block. As both these rewards are paid in cryptocurrency, the fluctuating value of each cryptocurrency against fiat currencies greatly affects the stimulus that will elicit validating decisions.
Each blockchain has a block space limit to maintain a sustainable growth rate, preventing overloading each block and the blockchain with data. Block space limit is regulated using metrics such as block size limit or block gas limit to cap the transactions number in a block. A transaction size is determined by the amount of data recorded. A transaction gas use is the computational effort required to execute all operations of this transaction. A more complex operation costs more gas, hence the more “space” in a block it occupies. Therefore, it is logical to allocate a block’s energy consumption as well as block reward generation to each transaction proportionally to its share of the block’s total size or total gas usage.
Using the LCA calculator, we measured the GHG emissions associated with the sale of ‘Everydays: The First 5000 Days’ and estimated the emissions for the hypothetical scenarios, where cryptocurrency transactions were conducted on bitcoin and where NFT-related activities took place on Tezos. We used data from transactions randomly selected from around the same time of the ‘Everydays: The First 5000 Days’ sale to calculate the impacts in these two scenarios. We recorded the emissions of each transaction separately to compare not only the overall carbon intensity of the same activity across different blockchains, but also the different individual transactional activities on the same blockchain. This is to identify emissions levers for decarbonisation decisions and clearly define the control boundary of each co-participant in the sale to support reporting needs.
The study results (chart. 2) show that NFT activities on Tezos emit almost 100% fewer emissions in comparison to activities of similar functions on Ethereum, a significant reduction from 357 kgCO2e to 0.001 kgCO2e. Cryptocurrency transactions on bitcoin emit between 1.5 to 6 times the emissions of those on Ethereum. Within Ethereum, NFT-minting transaction is the most carbon intensive, following by NFT-transferring transactions, which have the average emissions higher than those of cryptocurrency-transferring transactions.
From these results, given the most comparable contexts, activities on Tezos are the least carbon-intensive. This is explained by the proof-of-stake underlying consensus mechanism. While proof-of-work requires tremendous computing power from nodes in a hashing race to avoid corruptions, proof-of-stake sets up to use cryptocurrency as collateral and eliminate the need for computing races and the entailed tremendous energy use.
The assessment results have informed Christie’s’ decision to support the development of Ethereum 2.0, a transition from proof-of-work to proof-of-stake as the consensus mechanism. With a total of 80,300 NFTs created on Ethereum so far, this transition would reduce the NFT minting emissions from over 28,600 tCO2e (using the proxy emissions from the Beeple’s ‘Everydays: The First 5000 Days’ NFT) to just over 14 tCO2e. This would effectively eliminate the carbon impact problem of one of today’s most popular blockchain platforms.
We also used the LCA tool to calculate and report 2 tCO2e for the emissions associated with Christie’s on-blockchain transactions in 2021 from the sales of 104 NFTs, mainly deriving from cryptocurrency payments. While this total is marginal, reflecting the very low volume of NFT activities, Christie’s now understands the potential impact of scaling its NFT business – for example, if it was to scale to an equivalent volume of physical art sales. Therefore, Christie’s will continue measuring and reporting emissions from its NFT activity annually.
 According to the Digiconomist
 The list and percentage share of the top 10 most used hardware in Bitcoin mining were based on CoinShares’ research. The list and percentage share of the top 10 most used hardware in Ethereum mining were based on data by Kryptex, assuming the higher the hash rate, the higher the share. Using the model set up by Tezos Agora, Tezos baking hardware consists of Raspberry Pi 4B 8GB and Intel NUC with 50-50 share. Cooling systems for Ethereum and Bitcoin were modelled after the “Green room” concept of TeliaSonera and uses the results from the life cycle assessment by Felipe B Oliveira.
more than a word.