Four networks, judged against their own goals

Ethereum's goal is a general-purpose execution environment in which arbitrary programs can hold and move value. Its documentation describes ether as the asset that pays for execution, and it describes the supply as the net result of validator issuance and the base-fee burn introduced by EIP-1559. The proposal that introduced the burn states the monetary consequence openly: a fixed ether supply can no longer be guaranteed, and the developers lose some control over the long-term quantity.

Judged against its own goal, Ethereum's design is also coherent. A programmable settlement layer needs an asset that can be spent on computation, and the burn ensures that only ether can be used to pay for transactions. It needs a consensus mechanism that can finalise blocks quickly enough for applications, and the transition to proof-of-stake delivered that while cutting issuance sharply. The cost is that the monetary policy is a policy, revisable by the network's governance process, which is a different property from Bitcoin's and not obviously a worse one.

Solana's goal is stated in its whitepaper: a new architecture built around Proof of History, a proof for verifying order and passage of time between events, used to reduce the messaging overhead of a Byzantine fault tolerant replicated state machine and to reach sub-second finality. The engineering documentation lists the components that serve that goal, from Turbine's block propagation tree to Sealevel's parallel execution, which requires each instruction to declare the accounts it will read and write in advance.

Judged against that goal, the design is coherent and the trade is explicit. Throughput is bought with hardware: the parallel runtime uses as many cores as the validator has, and the whitepaper's performance analysis is conducted on a stated network and hardware assumption. The cost is that running a validating node is expensive, which concentrates validation among operators who can afford it. A reader who values capacity over the cheapest possible independent verification will find the trade reasonable; a reader who values the opposite will not.

The Internet Computer's goal is to run applications. Its documentation describes a network of independent blockchains called subnets, each producing a finalised block roughly every second and executing canisters that hold state and serve requests. The distinguishing capability is chain-key cryptography: keys are split into shares across a subnet's nodes, and threshold ECDSA and Schnorr signatures let a canister control addresses and sign transactions on external chains, which the documentation presents as the foundation of Chain Fusion.

Judged against that goal, the design is coherent in a way the other three are not attempting. A canister is closer to a server process than to a script: it has state, it can call other canisters, and it can act autonomously on other chains without a bridge. The cost is that the platform is not a monetary network in the sense Bitcoin is, and its security model is a threshold one that assumes a bound on faulty nodes within a subnet rather than the longest-chain rule.

Judging each network against its own goals is fairer than ranking them on one metric, but it is not neutral either. It assumes that the stated goal is the right goal, and that assumption is doing a lot of work. A network can be perfectly designed for a purpose that turns out not to matter, and a network can be badly designed for its stated purpose and still be useful for something else entirely.

It also cannot settle questions of degree. "Cheap to validate" and "expensive to validate" are not binary, and the honest answer to how decentralised any of these networks is depends on measurements that change month to month and that this site does not publish. The same is true of throughput, of staking concentration and of the distribution of mining capacity. A reader who wants those numbers should go to the networks' own dashboards rather than to a comparison page.

What the comparison can do is make the trade-offs legible. Each of the four networks gives something up to get what it was built for, and in every case the thing given up is visible in the design rather than hidden. Bitcoin gives up throughput for cheap verification. Ethereum gives up a fixed supply for programmability. Solana gives up cheap node operation for capacity. The Internet Computer gives up the longest-chain security model for the ability to host software that signs for itself. None of those is a defect; each is the price of the goal.