transaction output (UTXO) set will be at least 4 x 109 x 5 x 50 bytes = 1 TB in size. Unlike the blockchain which can be pruned, nodes require the full UTXO set in order to validate
incoming transactions and blocks. And nodes need to perform fast lookups against the UTXO set, so the UTXO set cannot be written to an optical drive and forgotten about. Instead, the
UTXO set must be maintained in "warm memory," which in practice will mean a combination of RAM and flash. I would argue that maintaining a 1 TB UTXO set is "on the same order of
technological difficulty" as downloading 35 GB blocks every ten minutes. The story that somehow LN will allow bitcoin to scale to global adoption levels on a network of low-cost
raspberry-pis is untrue.
No matter how I look at it, if bitcoin grows to global adoption levels and if most users run non-custodial wallets, then the hardware resources to run a node
will exceed the capabilities of mobile devices and low-cost single-board computers like raspberry-pis. The difference is that in the original design of bitcoin, the users can still be
their own banks, verify their own transactions, and send payments to any other user without the assistance of an intermediary. And this is possible with hardware no more sophisticated
than a cheap feature phone on a SMS plan in a third world country. In the LN future, on the other hand, bitcoin (if you can even call it bitcoin at that point) becomes Banking 2.0."
