| The author is arguing that decentralizing the control of compute power and compute-as-a-service is valuable and should be invested in. It is valuable if we ever reach a point where say, AWS has a monopoly on all compute power, and starts making bad/unfair/predatory decisions that hurt customers. Otherwise running decentralized payment networks to exchange compute power is much less efficient and much more costly than relying on purchasing services directly from a few competing (read: non-colluding) service providers using USD, GBP, EUR, credit cards, etc. There's really no need to decentralize compute-as-a-service. This person is getting paid to talk their book because their employer has placed a lot of bets in the cryptocurrency/blockchain world and is trying to brainstorm ways to turn them into actual viable products and exit these investments. Blockchains / decentralized ledgers are almost always solutions looking for a problem. They solve the very niche issue of exchange between untrusted parties, with caveats, and with extra performance/cost added to a centralized solution. There's currently no reason to implement a decentralized ledger solution for exchanging compute-as-a-service. |
At least with applications like storage you can encrypt your data. AFAIK homomorphic encryption isn't general or efficient enough to be very useful yet.