The number of blockchain nodes and their distribution is a strong indicator of the relative decentralization of the network. Since the cost of storing smart contract data on-chain can be very high, most dApps typically rely on centralized servers and cloud providers (such as AWS, Google Cloud, Microsoft Azure, and Alibaba Cloud) to store data and host their client-server interface.

Ethereum: 6,000 nodes / $73,000,000 per GB

Ethereum has around 6,000 nodes around the world, most of which run on centralized servers. In 2020, Decrypt reported that 70% of Ethereum nodes are running on AWS. The cost of storing 1 GB of data on the Ethereum mainnet fluctuates, but currently costs over $100 million.

Cardano: 3,173 nodes / insufficient data storage

The Cardano Foundation, along with IOHK and Emurgo, initially set up the nodes to run the network. They handed over the block production process to the network’s stake pool operator, which currently stands at 3,173. It's unclear how much is cloud. Cardano is an append-only ledger that records asset ownership and transfers and is impractical for permanently storing files.

Solana: 1,603 nodes / $1,000,000 per GB

Most Solana nodes are hosted on centralized servers like AWS. Its outage raised concerns that Solana would stop validating new transactions if 33% of nodes went offline. Storing on-chain data on Solana costs about $1 million.

Avalanche: 1,243 nodes / $988,000 per GB

Avalanche has documentation for setting up nodes with AWS, indicating that a significant portion of its 1,243 nodes may be hosted on AWS. The cost of storing 1 GB of data on-chain is $988,000.

Algorand: 1,997 nodes / IPFS off-chain storage

Algorand reports around 2,000 nodes, but they are hosted on centralized servers. While exact numbers are not available, Algorand relies on AWS to scale its network. Instead of using on-chain storage, Algorand uses the Interplanetary File System (IPFS) to store data.

Internet Computer: 443 nodes / $5 per GB

Internet Computer has 443 nodes operating in independent data centers around the world. None of them are centralized cloud providers. Node operators have hardware requirements to ensure that the blockchain can respond to HTTP requests and deliver web content without intermediaries. The cost of on-chain smart contract data to store 1 GB of data per year is $5.

Transaction Speed

One of the key metrics that determines the efficiency of a blockchain network depends on how fast it processes transactions. This involves two specific aspects:

Transactions per second (TPS) and block finality or transaction finality. TPS refers to how many transactions a blockchain can successfully process in one second; block finality refers to the duration between the initiation of a block and its final irreversible settlement on the distributed ledger.

Aptos: 160,000 TPS / 1 second deterministic

Aptos blockchain claimed to be able to process 160,000 transactions per second (TPS) but at the moment its TPS is around 23.

Sui: 120,000 TPS / 2-3 second deterministic

Sui is a high-performance smart contract platform under a Monolithic structure. In particular, a Monolithic blockchain is a blockchain that handles all tasks and jobs on a single network. Sui is scalable up to several hundred thousand TPS (number of transactions processed per second) and block times range from 2 – 3s.

Ethereum: 15-20 TPS / 14 minutes deterministic

Ethereum typically processes 15-20 TPS with a block confirmation time of 14 minutes. During periods of extreme network congestion, this rate is much higher, and it can take more than 14 minutes to complete a transaction. Users typically have to pay a “gas” fee for each transaction, depending on traffic, which peaked at $1,000 last year. Ethereum is trying to correct this, with upcoming sharding and layer 2 scaling solutions expected to increase transaction speeds.

Cardano: 2 TPS / 10-60 minutes deterministic

Cardano can purportedly handle 250 TPS, with reported block determination times ranging from 10 minutes to an hour, depending on network congestion. Its Layer 2 scaling solution, Hydra, is said to theoretically achieve up to 1 million TPS and instantaneous block finality. But third-party measurements show that Cardano currently only supports around 2 TPS.

Algorand: 20 TPS / 4-5 seconds deterministic

Algorand theoretically reports processing 1,200 TPS and intends to increase that to 3,000 TPS immediately. According to its metrics page, it actually handled 20 TPS. Algorand has a block confirmation time of 4-5 seconds.

Solana: 2,000-3,000 TPS / 21-46 seconds deterministic

Solana can theoretically handle 65,000 TPS, but in practice, it handles around 2,000-3,000 TPS. When trading peaked at 400,000 TPS in September 2021, it suffered an outage, raising concerns about its instability. Solana's theoretical transaction completion time is 400-500 milliseconds, but in practice "optimistic confirmation" takes 21-46 seconds.

Avalanche: 4,500 TPS / 2-3 seconds deterministic

Avalanche reports that its testnet can handle 4,500 TPS, with a production-ready version eventually targeting 20,000 TPS. But currently, it can only handle about 9 TPS for EVM compatible C-Chain. Avalanche reports that transactions complete in less than a second, but can take around 2-3 seconds.

Internet Computer: 11,500 TPS / 1 second deterministic

Internet Computer is currently configured to handle up to 51,666 TPS, but recent performance tests showed 11,500 TPS with a 1 second dapp finalization rate. This high ratio is due to the blockchain distinguishing update calls (state-enhancing) from query calls (state-neutral). Internet Computer can handle over 250,000 read-only query calls in milliseconds.

Energy Efficiency

Blockchain technology must reduce its carbon footprint to become sustainable. Energy-intensive proof-of-work consensus can have a detrimental effect on the climate, as miners require a lot of energy to power their equipment.

Each Ethereum-based transaction consumes an estimated 238 kWh (kilowatt hour) of energy. (For reference, 100,000 VISA transactions consume 149 kWh of energy.)

Ethereum miners consume about 45 TWh (terawatt hours) of energy per year. Ethereum’s transition to proof-of-stake is expected to drastically reduce its energy consumption.

The internet computer ICP consumes 1.3 kWh of energy per transaction run on-chain, rather than outsourced to a centralized cloud provider.

Existing proof-of-stake blockchains consume a small amount of energy for a single transaction. Cardano consumes only 0.5 kWh per transaction, while transactions on Solana consume 0.00051 kWh of energy. A report by the Crypto Carbon Ratings Institute shows that Avalanche consumes 0.00476 kWh per transaction and Algorand consumes 0.0027 kWper transaction. However, each of these blockchains outsources storage and hosting to centralized servers, suggesting a hidden carbon footprint.