Buying 51% is not a Costco run. Hashrate is hardware + cheap power + logistics. Exchanges have cash, but wafers and PSUs have lead times and siting takes months. Pooling other people's hashrate is easier, but then you are herding independent miners who can leave in a day.

What 51% can and cannot do:

Cannot change the rules or steal coins. Invalid blocks get orphaned by nodes.

- Can censor (leave out some tx), and with a true majority they can do it indefinitely while they keep that majority.
- Can double-spend targets by privately mining and reorging, mainly to defraud counterparties like exchanges. Mitigation is more confs and risk controls.
- Economics bite hard: visible censorship or reorgs nuke revenue and reputation; miners and users route around the attacker.

The real risk today -- Power over block templates lives at pools. A few large pools under the same policy pressure could soft-censor. That does not require 51%; it just slows some transactions and annoys everyone.

What helps in practice

Miners: prefer pools that commit to neutral templates and adopt Stratum v2 with job negotiation so miners, not pools, pick transactions. Be ready to switch pools.
Users/merchants: self-custody, run your own node, and choose sensible confirmation policies for incoming funds. Big payments get more confs.
Devs/operators: diversify funding, monitor pool shares, and keep pushing open standards that reduce single chokepoints.

Is this already unfolding? Exchanges dabble in mining and some run pools. That is a long way from coordinating a profitable majority attack. The more realistic failure mode is policy-driven filtering by a handful of pools, which market pressure has reversed before.