Proof of work and proof of stake, and what choosing costs you

Why does a network have to choose what to sacrifice?

Bitcoin keeps security and decentralization and lets scalability go, at seven transactions a second. Ethereum, after the move to proof of stake, looks for a balance across the three and pays a little on each. The fast networks like Solana or BSC keep scalability and security and pay in decentralization, because validating takes powerful machines.

You can have two of them together, and the third is what you pay. Bitcoin keeps security and decentralization and settles for seven transactions a second. For the role it chose, a monetary system built to last, that’s a price that makes sense. Ethereum, after the move to proof of stake, looks for a balance and pays a bit on both sides. The fast networks, from Solana to BSC, buy speed with hardware requirements that narrow who can take part, which means they pay for it in decentralization.

It’s a practical constraint more than a law of physics, but so far nobody has managed to break it, and it’s the lens for reading the promises. When a network announces numbers far above everyone else’s, before you ask how it does it, ask what it stopped doing.

How does proof of work buy security?

That’s where the defense comes from. To rewrite the history of the blockchain you need more than half of all the network’s computing power, and you need it for as long as the attack lasts, not for a moment. On bitcoin that means buying or building hundreds of millions of dollars of hardware and then paying its power bills. And you know that if the attack works, the coin you’re attacking loses value, and with it the point of everything you spent.

The beauty of the mechanism is that it asks nobody for trust and forbids nothing: it makes an attack possible but uneconomic, and lets the arithmetic defend the network. The flaw sits in the same sentence, because that defense costs real energy, every day, forever.

Who actually mines, in the end?

The result is that most of bitcoin’s computing power runs through a handful of pools. That doesn’t mean they can attack whenever they like, because miners can move elsewhere at a moment’s notice, and would. It does mean there are few doors to knock on, for a government or a court.

In 2021 China showed how much geography matters: mining was banned, and the network’s computing power fell by about half in a few weeks. The network kept running, the miners moved elsewhere, and within a year the level was back where it had been. That test showed two things at once, how resilient the network is and how concentrated.

What does it cost in energy, and does that matter?

If bitcoin is an alternative monetary system, the consumption compares against everything it would replace, banks, cash, gold and physical security, and comes out comparable or lower. If it’s a speculative asset, the same energy is spent propping up a bet. Public perception stays negative either way, and that’s a regulatory risk.

If you see it as an alternative monetary system, the right comparison is with the whole apparatus it would replace, from branches to cash, from mining and storing gold to physical security. On that scale the consumption is comparable, maybe lower. If you see it as an asset people speculate on, it’s energy burned to prop up a bet, and the comparison no longer holds.

Two things stay outside this argument, and both are true. A sizeable share of mining runs on renewable energy, or on energy that would be lost anyway, like gas that would otherwise be flared or surplus power from hydroelectric plants far from the grid. And public perception stays negative whatever the details. That’s a real risk, because parliaments write the rules, and they don’t read consumption charts.

How does proof of stake buy security?

To attack here you need the majority of the locked tokens, which is an enormous amount of capital. On top of that you have to buy it on the market, pushing the price up as you accumulate. And there’s one difference that matters more than all the others: if the attack works, the community can change the rules and burn the attacker’s stake.

Under proof of work, after an attack, the attacker still has warehouses full of machines and can try again. Under proof of stake they’re left empty-handed. That’s why, on paper, attacking a network built on locked capital costs more: the ticket is expensive, and above all it isn’t refundable.

Where does power concentrate under proof of stake?

It’s extremely convenient, and sure enough the whole market piled in. Today about 25% of Ethereum’s stake runs through Lido, and three or four operators of this kind, put together, end up coordinating the majority of what was supposed to be spread out. No conspiracy and no attack: it’s simply where convenience leads.

The difference from proof of work is subtler than it looks. There, miners can switch pools in an afternoon. Here, whoever delegated holds a receipt that’s worth something only as long as that operator works, and getting out costs time and a few points on the price, so moving from one operator to another is much slower.

What is MEV, and why does it weigh more here?

The methods are well known. You see a large order coming into a dex, you put one of your own right before it and one right after, and you pocket the price difference you created. Or you see a trade that pays and copy it by jumping ahead of it. Or you’re the first to liquidate a position that has just crossed its threshold. You don’t have to break anything. You just have to get there first, and whoever orders the block always gets there first.

Under proof of work the problem exists but it’s more diluted, because nobody knows in advance who will find the next block. Under proof of stake you know in advance whose turn it is, and whoever validates two minutes from now has time to prepare. That’s where the worst incentive comes from: leaving a transaction out because it pays. That’s censorship, even if nobody calls it that. The fixes under way separate whoever builds the block from whoever proposes it, and they work, but in exchange they add another layer of intermediaries.

How do they differ, point by point?

Cost of security: external and continuous under proof of work, internal and circular under proof of stake. Finality: probabilistic against economic. Barriers to entry: high and physical against low, but with power proportional to capital. Regulatory risk: a ban over consumption against an order to censor, for whoever validates.

The second is finality. Under proof of work it’s a matter of probability: the more blocks go by, the less conceivable a reversal becomes, but in theory you always can. Under proof of stake, past a certain point, going back means destroying capital, so it’s sharper. The third is the barrier to entry. It’s high and physical on one side, with dedicated machines and cheap power. It’s low on the other, where the reward stays proportional to what you put in, so power stays where it already was.

The fourth is the relationship with the rules, which changes in kind. Proof of work gets attacked over consumption, and bans arrive for environmental reasons. Under proof of stake whoever validates has a name and an address, and can be ordered to leave certain transactions out. The first risks being banned, the second risks obeying.

What is there beyond the two?

Proof of authority: validators approved in advance, fast but a club. Delegated proof of stake: validators elected by token holders, tends to settle into a few hands. Proof of history with proof of stake: transactions lined up by a cryptographic clock, needs powerful machines. New mechanisms like Avalanche and Algorand: elegant, with little history behind them.

Delegated proof of stake has token holders vote to elect a fixed number of validators. On paper it’s democracy; in practice the same slates win every time and power settles into a few hands. Solana pairs proof of stake with a cryptographic clock that lines up the transactions before anyone votes, and it works, but it demands machines so powerful that only a small group can validate.

Then there are the ones that changed how agreement itself is reached, like Avalanche and Algorand, with elegant mechanisms and few years behind them so far. And that’s the point that holds for all of them: a consensus mechanism proves its security through the years in which nobody managed to break it, and a blackboard proves nothing.

What should you look at if you have to pick one?

If you’re sizing up a new network instead, go back to the trilemma and ask what it sacrificed. If the answer is “nothing”, you’re reading it wrong. In both worlds, look at where power concentrates, the pools on one side and the liquid staking operators on the other, because that’s where it piles up while nobody is watching.

The last thing is the least technical and the most useful: count the years. Bitcoin has been running for 18 years, Ethereum’s proof of stake for four years, everything else for less. Time doesn’t prove a mechanism is perfect, only that nobody has managed to break it so far, and it’s the one proof that can’t be sped up.