No other way to scale globally than decentralization; hyperscalers keep failing
More than Decentralised Compute | Converge @ EthCC 2024 (Caladan)
“When you remove the need for a centralized scheduler or a centralized control plane… and you go to a fundamental peer-to-peer manner, you can scale unlimited… One of the reasons Akash has been around for 4 years and hasn’t had a single outage except for the planned upgrades is because it’s foundationally [decentralized]… During last one year Amazon went down 100 times — when it goes down you can’t order your Uber or your DoorDash. Is that the foundation that we want to build our AI [on]?… From a technological standpoint, I think there is no other way to scale at a global scale other than decentralization.” — 00:08:35
Context: Second pillar of his argument; he adds he confidently demos live on stage because “this will never fail.”