The problem being solved
If anyone can add to a shared ledger, anyone can add a false entry, and cheap identities mean
an attacker can simply outnumber everyone. Both mechanisms answer this the same way: attach a
cost to the right to write, so that flooding the network with fake participants stops being
free.
Where they diverge is what the cost consists of. Proof of work spends something external: hardware and electricity. Proof of stake commits something internal: units of the asset, which can be destroyed for misbehaviour. Every downstream difference follows from that.
Proof of work
- Cost is electricity and hardware
- Anchored outside the system
- Attack cost is ongoing, not sunk once
- Concentration follows cheap power
Proof of stake
- Cost is capital committed as stake
- Anchored in the asset itself
- Misbehaviour destroys stake directly
- Concentration follows staking services
What each one actually protects
In proof of work, rewriting history means redoing the work behind every block since, faster
than everyone else is extending the chain. That is a continuous expenditure, which is why
security is often described in terms of the cost per hour of attacking rather than a one-time
figure.
In proof of stake, an attacker must control a large share of staked capital, and provable
misbehaviour destroys it. The defence is not that the attack is hard to mount but that it is
self-defeating: acquiring enough of an asset to attack its network destroys the value of the
holding used to do it.
Where each concentrates
Proof of work concentrates around cheap electricity and hardware access, which is why mining
clusters geographically and why specialised chips reshaped it. Attempts to design
hardware-resistant algorithms have mostly failed; specialised equipment arrives wherever the
reward justifies building it.
Proof of stake concentrates around convenience. Running a validator requires uptime and
operational care, so most holders delegate to services, and a small number of large operators
end up controlling a large share of the stake. The distribution of ownership stays wide while
the distribution of control does not.
The energy argument, precisely
Proof of work consumes a great deal of electricity, and that consumption is the security
model rather than an inefficiency to be optimised away. Proof of stake uses a negligible
amount and buys its security differently. Both statements are true; what neither settles is
whether an internal or an external anchor is the better foundation, which is the actual
disagreement.
What changes for an ordinary holder
Very little day to day, and one thing structurally. Under proof of stake, holding the asset
can earn a yield by participating in validation, directly or through a service. That turns a
holding into something closer to a position with a return, and it changes the arithmetic of
simply holding.
It also introduces a decision that did not previously exist: who to delegate to. Delegating
concentrates control, and a holder choosing the largest operator for convenience is making a
small contribution to exactly the concentration the design is meant to avoid.
What each one does when an attack fails
The comparison is usually framed as cost of attack, which is only half of it. The other half is
what remains afterwards, and the two systems answer differently in a way that is easy to state.
An attacker on proof of work spends electricity and holds hardware. If the attempt fails, the
electricity is gone and the hardware is not, so it can be pointed at the same target again or
sold. The cost is real and it is largely external to the network being attacked.
An attacker on proof of stake commits units of the asset itself, and the protocol can destroy
them for provable misbehaviour. A failed attempt therefore removes the means of attempting
again, which is a stronger deterrent per attempt. It also means the penalty depends on the
network correctly identifying misbehaviour, which is a software judgement rather than a
physical fact, and software judgements have bugs. Neither property is strictly better. One
buys deterrence through unrecoverable loss and accepts a dependency on correct detection; the
other buys it through ongoing expenditure and accepts that the attacker keeps their equipment.
Neither is finished
Proof of work has an unresolved long-run question: what happens when issuance stops and
security depends entirely on transaction fees. Proof of stake has an unresolved question of
its own: whether stake concentrating in a handful of large operators eventually produces the
thing it was designed to prevent.
Both questions are decades out or already slowly arriving, depending on who is describing
them. Anyone presenting either mechanism as a settled answer is selling something, and
usually the asset that uses it.
Be the first to comment