High throughput has to be paid for somewhere
Parallel execution, a high-frequency Proof of History clock, and a design goal of processing a large volume of transactions all require validators capable of keeping up in real time. In practice, this means Solana validators generally need higher-specification hardware than validators on many other networks, meaningfully more RAM, faster storage (NVMe SSDs rather than typical consumer drives), and strong network bandwidth to keep pace with the rate of blocks, votes, and state updates being produced and propagated. This isn't an accident or a flaw, it's the direct cost of the architectural choices covered earlier: a network designed to let validators do more work per second necessarily requires validators with more capacity to do that work.
The honest framing is that this is a real tradeoff, not a free upgrade over other architectures. Higher hardware requirements raise the practical cost of running a validator, which can act as a barrier to participation for smaller or less well-resourced operators compared to networks with lower per-validator hardware demands. This is a genuine centralization pressure worth naming directly: a network's decentralization isn't just a function of how many validators exist on paper, it's shaped by how accessible it actually is to run one competitively, and higher requirements narrow that pool relative to a lower-spec design.