by Dave Sutter, Chief Strategy Officer, Marco Polo Network (formerly known as TradeIX) | Developers of Marco Polo
I was recently asked a very interesting and relevant question by Eleanor Wragg, Senior Reporter at the Global Trade Review.
To paraphrase, that question was, “What is the environmental impact of digitization efforts leveraging blockchain technology?”
Eleanor recently wrote an excellent article on this very question and I suggest all of you give it a read. I felt compelled to write a bit more on my thoughts on the subject because it’s a question I’ve wrangled with since I began my career working out of my college dorm room with a small team of entrepreneurs on the ground floor of Bitcoin. The impetus for her question is almost certainly the growing realization that certain blockchain networks, Bitcoin chief among them, are now consuming enormous amounts of energy. Because most of this energy is drawn from sources that significant negative environmental impact, people are beginning to question if the pros outweigh the cons.
Background on Consensus Algorithms and Blockchains
Before exploring the question further, we must first understand some of the fundamental technology concepts that blockchains and distributed ledgers are built on. The technology concept that is most relevant to this particular discussion is something called a consensus algorithm. This is because a great deal of the environmental impact (or lack thereof) for blockchains and distributed ledgers is dependent on which consensus algorithm they use.
Consensus algorithms power replicated state machines. A replicated state machine is a distributed or decentralized network of machines that works together as a coherent group to compute identical copies of the same state. State is a computer science term used to describe a computer system’s condition regarding certain inputs. Examples of state would be the most current copy of a financial ledger based on inputs and outputs (e.g. credits and debits) and/or the output a computer program executing on particular inputs (e.g. a digital signature authorizing ownership transfer of an asset).
Replicated state machines are used to solve a variety of trust and fault tolerance problems in distributed and decentralized systems. A key design feature of replicated state machines is that they are designed to survive failure and/or malicious behavior by one or more of the nodes on the network. The other key feature is that while the network consists of separate, independently owned and operated machines, the end users (clients) experiences one unified system when they interact with it.
Sound familiar? That is because all blockchains and distributed ledgers are replicated state machines that are powered by a wide variety of different consensus algorithms, all designed for a wide variety of different use cases and different network environments.
Some replicated state machines assume the worst of the other nodes on the network, and as such use consensus algorithms, classified as Byzintine Fault Tolerant, that support a network that is censorship resistant, trustless, anonymous, and capable of operating even if large portions of the network fail or act maliciously. But every design choice has a tradeoff; censorship resistance, trustlessness, anonymity, and resilience come at the cost of speed and efficiency; byztine fault tolerant consensus is much slower and can be much more resource intensive.
Others rely on consensus algorithms that assume more trust and that all nodes in the network have a known identity, and thus a means of identifying, punishing, and even removing nodes that are acting faulty or malicious. This allows them to operate much faster and far more efficiently.
Bitcoin’s Proof-of-Work as an Example
Bitcoin’s proof-of-work consensus is the most extreme example of a byzintine fault tolerant replicated state machine that is designed to operate in a totally anonymous, trustless and hostile network environment. Proof-of-work consensus relies on a combination of cryptography and economic game theory to ensure independent, anonymous, and potentially malicious software systems can continuously maintain a single source of truth. In the case of Bitcoin, that truth is a global ledger that maintains accounts denominated in numerical units called bitcoin.
Proof-of-work consensus depends on anonymous network of specialized machines called “miners.” Miners perform several critical functions.
The first function is to validate and record transactions. Miners receive requests from users that describe the ledger update and include proof that the sender is in fact authorized to perform such a ledger update. For example, “Please move 5 units of Bitcoin from Alice’s account to Bob’s account and here is proof that I am authorized to move these units from Alice’s account.”
Miners receive these requests and do two very important things with them. One is to validate that the sender does in fact have the right to move those units, a process that relies on public key cryptography to verify digital signatures. The second is to validate that the user has enough units to complete the transaction. For example, if Alice is proposing to a transaction involving moving 5 units of BTC from Alice’s account to Bob’s account, we need to be sure that Alice actually has 5 units of BTC to move. Miners do this by going back through all of the transactions on the global ledger that involve that account and ensuring that the current balance, calculated by subtracting the number of inputs (credits) from the number outputs (debits), is sufficient to cover the proposed transaction.
The second function is perform proof-of-work. Miners perform proof-of-work by using the transaction data they receive as an input to a complex math problem. Miners race against one another to solve this math problem. The miner who solves the problem first earns the right to update the global ledger by committing a new “block,” which a file containing time-stamped and validated transactions that describe ledger updates which have yet to be reflected in the current state of the global ledger. These races take place within a specified period of time and occur continually, 24/7/365. For Bitcoin, each race lasts about 10 minutes.
To solve the problem, miners generate hashes that include transaction data contained in each block, the hash of the previous block, and a variable changed after each hash called the nonce. A hash is a mathematical function that takes an arbitrarily long input and creates a fixed length output such that it is computationally infeasible to guess the inputs that created such an output. Change one thing in the input, and the output will change wildly and in a completely random way that is impossible to predict. The Bitcoin network sets the difficulty of this math problem by setting something called a difficulty target, which specifies the number of leading zeros a hash needs to have to win the race. The more leading zeros, the harder the problem is to solve. This means that the more computing power a miner uses to solve the problem, the likelier they are to solve the problem and win the race. Bitcoin’s protocol automatically reacts fluctuations in network-wide computing power, known as the hash rate. As the hash rate increases, so to does the difficulty on the math problem miners need to solve. The same is true in reverse.
Since miner’s also include the hash of the previous block, they cryptographically tie each new block to the preceding block and all others before it. If someone changes the data in the preceding block, it will change that blocks hash. Since this hash has been including in the subsequent block, it will also change the hash of that block, and all subsequent blocks. This creates a cryptographically linked “chain” of “blocks” containing timestamped transactions. The effect is that any change in the infinitely long chain of blocks is immediately detectable by all the other nodes on the network.
In a vast but pragmatic oversimplification, miners performing proof-of-work are essentially converting raw energy into brute force computing power that creates a sort of cryptographic “glue” which binds the global ledger together in a way that makes it computationally infeasible to alter or tamper with. This cryptographic glue means anyone hoping to tamper with or alter the global ledger must control a majority of the network’s total computing power. What follows is that the more computing power involved in creating this cryptographic “glue” the more computing power is required to undo it.
Miners do not do all of this for charity. The winning miner earns a reward in the form of newly minted units being credited to their account on the global ledger. These units are denominated in the native unit of account, bitcoin. Today, bitcoin has a real monetary value, as it is convertible into fiat currency at the current market rate. Over time, the units issued by miners increased as rewards have raised in value dramatically. At the time of its inception, one bitcoin was worth mere pennies. At the time of this writing, 1 BTC is valued at approximately $9,000. Considering the current reward for solving the problem is 12.5 bitcoins, and the problem is solved approximately every ten minutes, this means a miner is earning a reward worth over $100,000 approximately every 10 minutes.
This economic incentive has attracted an increasing number of miners to the Bitcoin network. Since more miners means a higher hash rate, this increase has also increased the difficulty of the problem, and thus the amount of computing power and energy required to solve it.
The economic incentives (and disincentives) this process creates is a ingenious form of economic game theory that incentivizes good actors to support, grow, and secure the network and disincentivizes bad actors from trying to do the opposite. Miners must spend lots of time and money on hardware required for mining and energy required to run them. If a miner or miners purposefully collude to compromise the ledger, their reward for doing so will become worthless, as the value of the units they might gain by altering the ledger will plummet because the world no longer trusts the integrity of the ledger, and thus will no longer be willing to exchange goods, services, or other forms of value like US dollars, for that bitcoin. It’s as if by cracking the safe, the thief makes all of its contents worthless.
So what does all this have to do with environmental friendliness of Bitcoin?
At Bitcoin’s inception, most mining was done on normal laptops. Today, Bitcoin mining is a multi-billion dollar industry dominated by industrial mining operations that oversee warehouses full of state of the art, supercharged hardware that can only do one thing; mine Bitcoin.
That is because, as discussed earlier, as this math problem continually becomes harder and harder to solve, miners have required more and more computing power to improve their chances of being the first to solve it and earn the reward. This has meant miners need to spend more and more money on hardware and use more and more energy to earn these rewards. This dynamic has created an exponentially growing arms race in which miners around the world continually deploy ever more powerful hardware that requires ever more energy to run it.
The result has been that modern industrial Bitcoin mining operations can now draw as much power as a small city. One of the reasons that Ethereum is taking the dramatic step of moving from proof-of-work to proof-of-stake consensus is to avoid the obscene hardware arms race and the resulting environmental impacts we’re seeing in the Bitcoin ecosystem.
Many say that this state of affairs is unsustainable and that the environmental impact alone means blockchain networks like Bitcoin have an overall negative impact on society. Others would argue that this energy consumption is a key element of the game theory and security model needed to support a global, censorship resistant, trustless, and truly decentralized network like Bitcoin and provide all of the benefits that such characteristics confer. What follows is that this is energy well spent as the pros outweigh the cons, however that’s a discussion for another time.
A Different Approach for a Different Problem
Fortunately, this is not the case for blockchain initiatives focusing on trade digitization. Many blockchains, not just enterprise blockchains, utilize consensus mechanisms that are not nearly as resource intensive as Bitcoin’s proof-of-work, such as R3’s Corda, Hyperledger Fabric, Ripple, among many others. These blockchains, including the enterprise blockchains underpinning trade digitization initiatives, are no more energy intensive than traditional software systems.
What follows is that, in the worst case, the use of blockchain in digitizing trade should not be any different than the environmental impact of digitization efforts that do not involve blockchain.
This conclusion begs another question, which is what is the environmental impact of trade digitization overall, regardless of whether it involves blockchain or not.
To me, there are a few key considerations when evaluating the environmental impact of trade digitization.
For one, trade digitization is not nearly as resource intensive as other computing activities like big data, data mining, machine learning on large, complex scientific data sets (e.g. CERN’s LHC Data Centers) and video streaming services (e.g. Netflix, Hulu), among many others. If we equate consumption of computing resources to the consumption of fuel, trade digitization is like a compact car and compared to the Saturn-V rocket that is Netflix streaming 4k video to millions of people around the world.
I would estimate that the cloud computing costs to support the trade activities of a bank would be about at least tens of thousands of dollars annually and up to single digit millions of dollars. In comparison, Netflix reportedly spends half a billion dollars a year on cloud computing, a figure that is orders of magnitude lower then it would be for normal users buying at market prices because Netflix is most certainly receiving very deep discounts for massive bulk capacity purchases on long term contracts.
Secondly, the computing required for trade digitization, light as it is, will only become more efficient over time. As time goes on and more computing is required to support an increasingly digital trade ecosystem, the efficiency of the computing supporting it will also continue to improve, as computers will continue to do increasingly more with increasingly less. It’s tough to say how the growth in computing resources required for trade digitization will track against the pace of increase in computing efficiency, but my bet is that computing efficiency will outpace the growth in resource demand by a lot.
Not only is hardware continually getting better and more efficient, but the adoption of cloud computing and modern devops methodology that leverages services like Kubernetes and serverless architecture is allowing companies to become extremely efficient in the way they manage and optimize their consumption of computing resources (and thus energy). Combining more efficient hardware with the use of cloud computing and modern devops methodology enables much better capacity planning, resource mutualization and optimization, and computing resources that are only consumed at run time, all of which leads to a dramatic reduction in waste and equally large increase in efficiency – and most importantly, a corresponding decrease in the amount of energy required to run all of this.
This brings me to my third point. As banks continue to adopt technology and become more digital, it is true that they will require more computing resources, but the net impact is still overwhelmingly positive. That is because technology and digitization allows them to become much, much more efficient and more environmentally friendly. Just think – without technology, not only would the banking industry consume gargantuan amounts of paper, but moving all that paper around the world would create an incalculable amount of carbon emissions from all the cars, trucks, planes, and ships that need to courier all that paper around the world. The armies of people required to process it all would fill entire cities, generating even more unquantifiable, negative environmental impacts. Just consider the colossal amounts of paper that global trade and trade finance consumes today, and that is after decades of sustained, industry-wide digitization efforts.
Lastly, we must also consider the positive economic benefits of trade digitization and how they can help businesses, countries, and even entire regions to become more environmentally friendly. It’s no secret that trade digitization can drive economic growth and prosperity, especially in emerging markets. These same emerging market economies also happen to have poor track records when it comes to environmental friendliness. The economic growth and prosperity supported by trade digitization can lead to a growth in tax bases, grow industrial capacity, and increase the amount of capital available to both private and public-sector players who need it invest in and implement clean, renewable energy infrastructure. In fact, I suspect digital trade finance and transaction banking will be a key part of supporting investments in and development of all types of projects that generate positive environmental impacts, such as clean energy, clean water, public transportation, smart cities, and modern sewage and waste management systems, among many others.
It’s for these reasons I think the environmental impact of trade digitization will be overwhelmingly positive.