Sui achieves scalability by using object-centric design and parallel transaction execution instead of traditional sharding. Each object—like coins, NFTs, or liquidity positions—tracks its own state, so transactions that touch different objects can be processed in parallel without waiting for global consensus. Shared objects are only serialized when conflicts occur, allowing most operations to execute concurrently. This object-level separation, combined with Sui’s Narwhal and Bullshark consensus for ordering and finality, lets the network handle thousands of transactions per second while maintaining security. You can read more here: Sui Scalability.

// Conceptual parallel execution
for tx in transaction_batch {
    if tx.objects are independent:
        execute in parallel
    else:
        serialize execution
}

I can also make a visual showing object-level parallelism versus traditional shard-based scaling to make this concept clearer.

ar owner, allowing independent transactions to run in parallel.

Parallel Execution: Most transactions don’t conflict, so they don’t need global consensus—this boosts throughput.

Fast Path Finality: Simple transactions (like t

bject-centric model → Instead of global accounts, every asset or smart contract state is an independent object with its own ownership.

Parallel execution → Transactions touching only owned objects can run in parallel without consensus, giving massive throughput.

Consensus only when needed → Only transactions involving shared objects (like AMM pools or marketplaces) go through Narwhal + Bullshark consensus.

Horizontal scalability → Validators can process different

How Sui Scales Without Sharding

Object-Centric Model: Each object has a clear owner, allowing independent transactions to run in parallel.

Parallel Execution: Most transactions don’t conflict, so they don’t need global consensus—this boosts throughput.

Fast Path Finality: Simple transactions (like token transfers) finalize instantly without consensus.

Selective Consensus: Only shared or conflicting transactions require consensus, reducing network load.

Horizontal Scaling: Validators can scale with more machines, like modern cloud systems—no need to shard.

Bottom line: Sui scales efficiently by processing transactions in parallel and minimizing consensus, avoiding the complexity and limitations of traditional sharding.