Could Facebook use their 350,000 Nvidia AI GPUs to crack Satoshi Nakamoto's Bitcoin Key?

Could Facebook use their 350,000 Nvidia AI GPUs to crack Satoshi Nakamoto's Bitcoin Key?

Facebook have so far secured 350,000 of the latest Nvidia H100 GPUs to power their AI efforts. (H100s cost roughly $30k each, so no small investment).

They are hoping it will be enough power to create the world's first AGI, but is it enough to crack a Bitcoin key by brute force, I wondered - so I ran the Maths (or 'Math' if you're American).

If you could gain access to Satoshi Nakamoto's dormant Bitcoin account by guessing his key, you would find a treasure trove of ~1m Bitcoin, which at today's prices would make you the world's 25th richest person at roughly $56bn.

Let's look at the figures: We need to know two main things:

  • The Bitcoin Private Key Space: Bitcoin private keys are 256-bit numbers, leading to a possible 2256 different combinations.
  • Hashing Speed of an H100 GPU: the speed at which an NVidia H100 can attempt to guess the key, measured in hashes per second (H/s).

NVidia H100 GPUs, incredible as they are, were not designed to smash out SHA256 hashes as efficiently as a Bitcoin ASIC (Application-Specific Integrated Circuit) but they are still brutally powerful and have been estimated with a hash rate of up to 900 MHs - that's ~900,000,000 attempts to crack the key every second.

That gives:

  • Total combinations to attempt: 2256
  • GPU speed of an H100: 900 MH/s (900,000,000 H/s)

Working at this speed, 350,000 NVIDIA H100 GPUs with a hash rate of 900 MH/s would still take approximately 1.17 × 10 55 years to brute force Satoshi Nakamoto's Bitcoin private key.

This is longer than the age of the universe (13.8 billion years).

Quite considerably longer in fact, by a factor of 44.

It is mind-blowing just how many possibilities there are to rearrange only 256bits of information, making the time required to guess a key unfathomably large. This exercise underscores the practical impossibility of brute-forcing Bitcoin private keys with any current or near-future (non-quantum) technology.