On Proof of Work

Proof of work is a polarizing topic in crypto. Of the top 10 largest “coins” in market cap, only 3 use proof of work: Bitcoin, Ethereum, and Dogecoin¹. Bitcoin intends to keep using it. Ethereum is actively trying to get away from it. Dogecoin is a “memecoin”, so no one cares.

People have written a lot about proof of work’s energy use, and how proof of stake is “better” because it uses less energy. Whenever I see this as the crux of someone’s argument, it’s clear to me they’re not painting the full picture. Maybe that’s by choice, or maybe it’s because they genuinely don’t understand why Bitcoin—the oldest and largest crypto-asset—chose to use proof of work in the first place.

When you understand the “why” behind proof of work—decentralization—it becomes the benchmark to judge everything else. So I’ll attempt to explain Bitcoin’s proof of work, Ethereum’s move to proof of stake, and the trade-offs with each. Only then should we judge what’s actually “better”.

Proof of work

Proof of work is a consensus mechanism in Bitcoin. A consensus mechanism is something that can get a bunch of different people to agree on the same thing. The reason Bitcoin needs a consensus mechanism is because it’s a distributed ledger. A distributed ledger is a record of transactions that more than one person or entity (each called a “node”) is responsible for. The opposite would be a centralized ledger. A good example of a centralized ledger is the bank statement to your checking account. Only your bank is responsible for keeping a record of the transactions.

Bitcoin’s ledger gets updated as people transact on the network. When I send bitcoins to Bob, for example, that transaction is broadcasted to every node and validated against a set of rules that were coded into Bitcoin’s software. Assuming my transaction follows the rules, it is then recorded in Bitcoin’s ledger, better known as the blockchain.

The process to validate and record a transaction is known as mining. Miners bundle my transaction with hundreds of other transactions and attempt to create a “block”, a bundle of data that has been compressed through a cryptographic hash function. The output is a string of letters and numbers that are supposed to look super random but actually follow a set of rules from the hash function. The block, which has been compressed to a block hash, is “chained” to the block before it and used as an input for the subsequent block’s cryptographic hash function.

There are two other inputs to the cryptographic hash function, and that is the “coinbase” transaction and the “nonce”. The coinbase transaction rewards the miner with bitcoin. The nonce is an arbitrary number that miners are forced to guess. If they guess right, the hash function produces a string with a leading number of zeroes, predetermined by the rules of the network. The frequency at which today’s miners are guessing is in the trillions of times per second.

Now, replace the word “guessing” with “computing”. If you understand computing as a form of “working”—which miners are forced to undertake in order to hash a block—you now understand the basis for proof of work as a consensus mechanism.

Since miners are able to reward themselves with bitcoin for hashing new blocks, there’s a lot of competition on the network, which inevitably leads to competing blockchains. This is where another consensus mechanism kicks in: the longest chain rule. We can safely assume the blockchain with the most “work” is also the longest blockchain; therefore, the longest chain wins. The longest chain rule enables nodes to come and go on the network as they please, having high confidence that the longest blockchain is the “golden copy” of the ledger.

Proof of work, together with the longest chain rule, are known as Nakamoto consensus. Before Nakamoto consensus, there was no way to scale a distributed ledger without trusting your fellow nodes. Nakamoto consensus enables anyone, anywhere in the world, to agree on the same thing without relying on trust. But there is a downside to Nakamoto consensus, best understood through Bitcoin’s “difficulty adjustment”.

The purpose of Bitcoin’s difficulty adjustment is to maintain an average frequency of hashing one block every 10 minutes. It’s a piece of code in Bitcoin’s software that adjusts the number of zeroes a block hash must start with approximately every two weeks². The frequency is arbitrary – what’s important are the implications. The difficulty adjustment makes it harder and harder to mine bitcoin as total computing grows on the network. If you don’t grow with the network, your share of rewards gets smaller and smaller.

Mining is like a basketball game where the rim’s height is adjusted every 10 minutes based on how quickly you score 20 points. If you score 20 points in 8 minutes, someone raises the rim’s height to make it harder to score. If you score 20 points in 13 minutes, they lower the height. Let’s pretend that you can score more quickly by being taller, and you can make yourself taller by simply drinking more milk.

As long as you keep drinking more milk and getting taller, scoring 20 points quicker and quicker, the rim’s height is going to get higher and higher, and before you know it, the world is screaming at you for how much milk you’re drinking. Substitute “milk” with “electricity”, and you have the state of bitcoin mining in a nutshell.

The positive correlation between the difficulty adjustment and mining’s energy use has long been a concern, so other crypto-networks have tried finding an alternative consensus mechanism. The most popular is proof of stake. Ethereum, which currently uses proof of work, plans to transition to proof of stake as part of its Merge upgrade. Other networks have already implemented their own versions of proof of stake³.

Proof of stake

Proof of stake in Ethereum requires a “validator” (the equivalent of a miner in Bitcoin) to deposit 32 ETH into a smart contract. Smart contracts are digital vaults that receive and send money based on a programmed set of instructions. Nobody controls a smart contract; it’s like a vending machine that just does what it’s been programmed to do.

When I send 32 ETH into a smart contract to become a validator, my money gets “locked up” so long as I’m still a validator. This is important because my money is now collateral to the network. As soon as the network sees that I have money “at stake”, I can take part in validating.

Validators are randomly selected to hash transactions into blocks that other validators vote on. If a validator doesn’t hash when selected, or tries something against the rules, the network punishes the validator by “slashing” her stake in the smart contract. If a validator successfully hashes a block, she is rewarded with additional ETH from transaction fees; the reward is proportional to her stake in the network.

The willingness to accept the risk of slashing for the reward of transaction fees is the basis for proof of stake as a consensus mechanism in Ethereum. Whereas miners take on the implicit risk of paying more for hardware and electricity than what they earn from mining, validators take on the explicit risk of getting their money taken when they fail to validate.

There’s no difficulty adjustment in Ethereum because there doesn’t need to be – validators are randomly selected rather than competing for new blocks. Ethereum will also use a hash function that won’t require the same frequency of guessing that Bitcoin’s hash function requires. As a result, Ethereum gets consensus on its distributed ledger using much less energy.

Less is more, until it isn’t

Research on Ethereum’s transition to proof of stake report that Ethereum will use 99.95% less energy than it uses with proof of work. However, proof of stake also comes with trade-offs. One is the complexity of getting distributed consensus without relying on Nakamoto consensus. Lane Rettig, a former Ethereum Core developer, talked about this at length with Peter McCormack on the What Bitcoin Did podcast.

“When you make this change that we just talked about going from Proof of Work to Proof of Stake you break a lot of the really cool stuff…I don’t think any of us appreciated the complexity and how much fixing we’d have to do after breaking Nakamoto consensus.” — Lane Rettig

Ethereum’s transition to proof of stake was originally slated for January 2020 and recently got pushed back again with no firm date. Nakamoto consensus, widely considered an elegant solution to a complicated problem, has been working for Bitcoin since Day 1.

Another trade-off with proof of stake is a network becoming more prone to centralization. There are three prongs to this: governance, economies of scale, and regulation.

Governance: A validators’ stake is directly proportional to their voting power and future rewards. If I stake 5x more coins as a validator than the next guy, I have 5x the votes. I’m also rewarded with more coins. This type of governance is strikingly similar to companies that issue equity, i.e. centralized entities.
Economies of scale: There’s a “shorter runway” to economies of scale in proof of stake. It’s well understood in mining that becoming a large-scale miner relies on access to three things: capital, machines, and energy. It is very hard to consistently align the three. Validators don’t have the same concern with machines and energy, they just need a ton of stake.
Regulation: As more people pool their coins with centralized validators (e.g. exchanges), governments will regulate these validators more stringently, creating a regulatory moat around incumbent entities and hindering new competitors from accessing capital. This is similar to the regulatory moat enjoyed by big banks in traditional finance.

Decentralization matters. Any network blockchain that trends towards centralization is not a distributed ledger; it’s just another database. There is certainly a cost to decentralization, but we can’t lose sight of the bigger picture.

What’s really “at stake”

Mining’s estimated usage in January was 11.16 terawatt-hours, approximately 1% of the net electricity produced globally that same month⁴. We’re not going to save the planet by targeting 1% of electricity demand. We may even hurt it, considering miners are a tool to decarbonize the grid.

For 1% of the world’s electricity, proof of work enables Bitcoin to be the most decentralized, distributed ledger. Replace “ledger” with “money”, and you start to see what’s really at stake here (no pun intended).

Consider this statistic from blockchain company Chainalysis: in 2021, the top 10 most crypto-active countries in the world, on average, scored below the global average on measures of democracy and government integrity. Many people around the world—especially in developing countries—see crypto as a means to property rights and escaping anti-democratic/oppressive/corrupt political regimes. The more you centralize money, the easier it is to control that money and revert back to norm.

Even for people living in developed countries (myself included), Bitcoin provides a means to hedge our economic well-being. While I imagine most central bankers are well-intentioned, there is always the risk that wonky monetary policies could negatively impact our financial future.

Most people know the importance of saving and investing. Why restrict ourselves on the money we can save and invest with? At the micro-level, it’s about optionality. At the macro-level, it’s about giving power to the people. I have never met someone (rational) that is against more self-sovereignty, self-custody, and self-empowerment. These are concepts that—credit to Andre Cronje—are at the core of “crypto ethos”. Decentralization enables these concepts. Centralization does not.

It’s worth noting that centralization isn’t inherently bad. What centralization does enable is greater coordination. Ethereum needs greater coordination if it wants to be the “world’s operating system”. It faces an existential threat not just from competing platforms (such as Solana, Cardano, and Polkadot), but also from the idea that we even need a not-so-distributed ledger. We already have centralized money, and now with Bitcoin we have decentralized money. Is there appetite in the middle? It remains to be seen.

Footnotes

[1] May 28, 2022. Bitcoin’s market cap was $551B, Ethereum’s was $215B, and Dogecoin’s was $10.82B. Market cap data taken from Messari.

[2] Bitcoin’s difficulty actually adjusts every 2,016 blocks, but due to the 10 minute frequency it comes out to approximately every 2 weeks.

[3] Other examples of more well-known PoS networks include Polkadot, Avalanche, and Cardano.

[4] January 2022. Mining usage taken from the Cambridge Centre for Alternative Finance. Total net electricity produced globally (1,005.35 TWh) taken from the International Energy Agency.

Thank you to Ray Deck for contributing to this article.