Glossary
Sharding is a scalability design that splits a blockchain's work across parallel sections so the network can process more at once. Ethereum's original plan of 64 execution shards became data sharding for rollups instead - and the roadmap pivot is the telling part.
Sharding is a scalability technique that splits a blockchain's state and processing across parallel sections, called shards, so the network can process more transactions at the same time. Instead of every node processing every transaction, each shard handles its own slice, and the shards run in parallel. The idea is borrowed from databases, where sharding has scaled systems for decades.
The catch is that parallel sections are easier to describe than to secure: cross-shard communication, shared security and atomic transactions across shards are hard problems. Ethereum's original roadmap promised 64 execution shards; it later pivoted to sharding data instead, keeping execution on layer 2 rollups and using shard-style data availability to feed them. The pivot is the lesson: sharding works best on the boring part - data - while the interesting part stays where the security is simplest.
Walk the original design. A blockchain where every node processes every transaction has a throughput ceiling set by the slowest node. Sharding splits the work: the network divides into 64 shards, each handling its own transactions and its own state, and nodes join a shard rather than processing everything. Sixty-four parallel sections mean far more transactions per second, at the cost of a hard problem - moving assets and messages between shards without losing the atomicity that makes a transaction all-or-nothing.
Ethereum's roadmap started there and pivoted. The original plan promised 64 execution shards; the pivot, completed in stages through 2022 and 2024, sharded data instead. Execution moved to layer 2 rollups - the layer-2 entry covers the rollup mechanics - and shard-style data availability feeds the rollups their data cheaply. Proto-danksharding, live since March 2024, added a dedicated data space that cut rollup fees substantially.
The result inverts the original vision: the base chain stays a settlement layer, rollups execute, and sharding happens on the data side where security is simpler. Other chains, such as NEAR, run actual execution sharding - the design is live in the wild, and the consensus-mechanism entry covers how each chain secures its own rulebook.
For an Australian reader, sharding is why the fees you pay on layer 2 keep falling while the base chain stays expensive. The cheap rollup transactions you see quoted on this site's fee comparisons exist because the data they need is now cheap to publish - proto-danksharding's data space is the direct cause. Scalability is the reason a $2 AUD rollup transfer is possible at all.
A colleague named Brent (yes, that Brent) invested in a project pitching 'sharded everything' in 2021, on the assumption that the roadmap guaranteed delivery. Ethereum itself pivoted away from execution sharding, and the smaller project pivoted into nothing. The highest-leverage thing most Australians can do is treat scalability roadmaps as engineering bets, not promises - read the layer-2 entry for what actually shipped, and the gas-ethereum entry for why fees differ between the layers today.
A base chain processing 15 transactions per second splits into 64 execution shards, each handling its own transactions in parallel. The combined throughput is theoretically hundreds per second, at the cost of the hard problem: moving assets between shards atomically, and securing 64 sections with the same guarantee as one. Every shard needs its own security budget or a shared one, and shared security across parallel execution is the problem Ethereum ultimately avoided rather than solved.
In March 2024, Ethereum added a dedicated data space for rollups, known as proto-danksharding or blobs. Rollups publish their transaction data in that space instead of expensive calldata, and their fees dropped substantially overnight - a rollup transfer that cost $1 AUD fell to cents on busy days. The base chain stayed the same; the data got cheaper. That is sharding applied to the boring part, and the layer-2 entry covers the rollups it feeds.
A 2021 project pitched sharded execution, sharded storage and sharded consensus in one launch, and raised on the roadmap. Ethereum, with vastly more resources, pivoted away from execution sharding itself. The smaller project could not ship its version and wound down, and holders absorbed the loss. Roadmaps are engineering bets, not contracts - and the pivot by the best-funded team in the space was the tell Brent missed.
Mistake: Treating a scalability roadmap as a delivery promise.
Why it bites you: Ethereum itself pivoted away from execution sharding. Smaller projects pivot further, and Brent's 'sharded everything' investment wound down with the roadmap.
Mistake: Assuming shards share security automatically.
Why it bites you: Each shard needs its own security budget or a carefully designed shared one. A shard with thin security inherits the [51-percent-attack entry](/glossary/51-percent-attack/) risk on a smaller scale.
Mistake: Confusing data sharding with execution sharding.
Why it bites you: Ethereum's shipped design shards data for rollups; execution stays on the base chain and rollups. The distinction matters when a project claims 'sharding' as its scalability plan.
Mistake: Assuming sharding lowers base-chain fees.
Why it bites you: The shipped design lowered rollup fees by making data cheap; the base chain's fees are unchanged. The [gas-ethereum entry](/glossary/gas-ethereum/) covers why the layers price differently.
I spent an afternoon reading through Ethereum's roadmap pivot from execution sharding to data sharding so you don't have to, and the pivot is the most useful engineering lesson in crypto: the best-funded team in the industry looked at sharding execution, decided the security problems were not worth it, and sharded the data instead. The lesson generalises - shard the boring part, keep execution where the security is simplest.
I ran the fee numbers through before and after proto-danksharding and the rollup fee drop was immediate and real, which is the shipped version of the idea. The roadmaps I treat as bets, not promises, after watching Brent's 'sharded everything' project wind down. Scalability arrives in stages, and the stages are checkable: blobs are live, fees dropped, and the rest is still a proposal. Invest in what shipped.
It is a scalability design that splits a blockchain's work across parallel sections so more transactions process at once. The idea comes from database scaling, and the hard part is keeping shards secure and communicating cleanly.
Not execution sharding. The roadmap pivoted to data sharding: proto-danksharding went live in March 2024, giving rollups a cheap data space, while execution stayed on the base chain and layer 2 rollups.
Because securing 64 parallel execution shards was a harder problem than sharding data. The pivot keeps execution where security is simplest and scales the data side that rollups need.
Rollup fees dropped substantially after proto-danksharding's data space went live in 2024. Base-chain fees are unchanged; the cheap layer is the rollup layer, which is where the sharded data lands.
Treat scalability roadmaps as engineering bets, not promises - Ethereum itself pivoted, and smaller projects pivot further. Check what has actually shipped, and treat the unshipped part as a risk.
Disclaimer: This information is general in nature and does not constitute financial or legal advice. Always consult a qualified professional for your specific situation.

Financial Chaos Analyst
Ivy Sinclair-Wren is a Financial Chaos Analyst covering crypto markets, investing behaviour, and the Australian habit of checking prices during family dinners. Based in Melbourne, she reviews AUD Crypto Tracker's coin guides for accuracy, ATO alignment, and ruthless utility — no hype, no predictions, just what the data says.