[Cryptography] Question about Bitcoin's security model

[Renato Schiavinato Lopez]

On Sat, Jul 18, 2026 at 3:54 PM John Levine <johnl at iecc.com> wrote:
>
> The difficulty factor can go down as well as up.  If fewer people are mining, the difficulty
> will go down which means more subsidy per unit of CPU.
>
> I can see a lot of problems for Bitcoin, but this isn't one of them, at least until the subsidy
> hits zero and even then the fees can adjust.
>

One major limitation for a 51% attack today is the cost of hardware.
But, if miners abandon Bitcoin for AI due to energy costs, the world
will be flooded with cheap, used ASICs sitting in warehouses. A state
actor or a well-funded attacker won't need to compete with AI data
centers for top-tier energy contracts; they'll just buy up that
mountain of old hardware for pennies on the dollar.

Once the active network difficulty drops, the barrier to entry for a
51% attack plummets. The attacker can just plug in all that offline
hash power at once.

Bitcoin retargets every 2,016 blocks. While a massive influx of hash
power would make those 2k blocks pass in a few days instead of two
weeks, the difficulty adjustment won't save Bitcoin. A 51% attack
happens in real-time—the double-spends and transaction censorship hit
on day one. By the time the network retroactively adjusts to the new
reality, the damage is already done, not to mention the resulting
market panic.

There are a lot of moving parts required for such a scenario to
materialize, but I'd watch carefully if the network's hash power
consistently drops for long periods.

Renato