If your Ethereum-versus-Polygon research still mentions the Mumbai testnet, Goerli faucets, "Ethereum 2.0," or Polygon zkEVM as something to look forward to, none of it applies anymore. All four references describe a stack that no longer exists. Two of those networks have been switched off, one product line has been retired by its own creator, and the "Ethereum 2.0" branding was abandoned by the Ethereum Foundation years ago in favor of separate execution-layer and consensus-layer terminology. Here is what the two ecosystems actually offer a developer shipping a dApp today.
The three changes that invalidate older comparisons
First, Polygon zkEVM is finished. Polygon Labs shut the network down rather than deprecating it in favor of a successor chain. Alchemy's Polygon zkEVM deprecation notice confirms that support for both polygonzkevm-mainnet and the polygonzkevm-cardona testnet ended on July 1, 2026, and states plainly that there is "no migration path for Polygon zkEVM, as the network itself is being shut down." The sequencer stopped producing blocks in early July 2026. Any guide presenting zkEVM as Polygon's flagship scaling answer is describing a dead product.
Second, the testnet map was redrawn twice. Ethereum's networks documentation lists Goerli, Rinkeby, Ropsten, Kovan, Morden, and Holesky as deprecated and unmaintained; Holesky was retired in September 2025 and replaced by Hoodi. Only Sepolia, Hoodi, and Ephemery remain active.
Third, Ethereum's scaling work shipped. The Fusaka upgrade went live on December 3, 2025, and its headline feature, PeerDAS, is a direct subsidy to every rollup and sidechain that settles to Ethereum.
Testnets: Sepolia and Amoy, with one deadline you must diary
Sepolia is Ethereum's recommended default testnet for application development, running on a permissioned validator set controlled by client and testing teams. Chain ID is 11155111. Hoodi is not a dApp testnet — it exists for protocol developers testing network upgrades and for stakers rehearsing validator operations, and it carries a larger state and heavier sync requirements as a result. Ephemery is the oddity worth knowing about: it resets to fresh state every 28 days and needs under 5GB, which makes it genuinely useful for throwaway prototypes and CI bootstrapping.
One caution that older articles cannot have flagged: Ethereum's documentation gives Sepolia an estimated end-of-life of September 30, 2026. If you are writing deployment scripts or CI pipelines this quarter, do not hardcode Sepolia as a permanent fixture. Hoodi's own estimated end-of-life sits far out at September 30, 2028.
On the Polygon side, Amoy is the PoS testnet, with chain ID 80002. Polygon's Amoy announcement explains the dependency chain that killed its predecessor: Mumbai used Ethereum's Goerli as its root chain, so when Goerli was deprecated, Mumbai broke along with it. Amoy is anchored to Sepolia instead. That anchoring is worth noting given Sepolia's own EOL date — Polygon's testnet inherits Ethereum's testnet lifecycle, and this is the second time that inheritance has forced a migration.
What Ethereum L1 gives you now
Fusaka's PeerDAS implements data availability sampling so that, per Ethereum's Fusaka roadmap page, each full node holds only one-eighth of blob data while cryptographic reconstruction preserves availability guarantees — theoretically enabling an eightfold increase in blob space without proportionally heavier node hardware. Blob capacity was then raised through scheduled BPO forks rather than waiting for the next full upgrade. For anyone building a rollup or an app-chain, this is the single biggest cost variable that moved in the last year.
Fusaka also added a secp256r1 precompile, which enables passkey-based signing. That is a quietly significant unlock for consumer onboarding, since it lets wallets use the same hardware-backed credentials phones and laptops already ship with. Next up is Glamsterdam, targeting activation around the end of August 2026, with EIP-7732 (enshrined proposer-builder separation) and EIP-7928 (block-level access lists) as its headline proposals.
What Polygon PoS gives you now
Polygon PoS spent 2025 and 2026 being rebuilt around payments throughput. Per Polygon's own performance announcement, the block gas limit progressed from 30M to 45M with the Bhilai upgrade in July 2025, to 110M in February 2026, and to 160M in June 2026 — a capacity the company frames as roughly 5,000 payments per second.
The consensus changes matter more than the gas numbers for most builders:
- Heimdall v2 (July 2025) rebuilt the consensus layer on CometBFT and Cosmos SDK v0.50, cutting finality from one-to-two minutes down to about five seconds.
- Bhilai (July 2025) brought gasless transaction support via EIP-7702 account abstraction.
- Rio (October 2025) introduced a validator-elected block producer model and stateless validation, eliminating reorgs so confirmed blocks are final, and lowering node storage requirements.
- Giugliano (April 2026) shaved roughly two seconds off confirmations and published fee parameters onchain.
Five-second finality changes what you can build. Checkout flows, game state updates, and point-of-sale interactions that were impractical against minute-scale finality are now reasonable.
The toolchain is largely shared
The good news for anyone choosing between these networks is that the decision barely affects your tooling. Both are EVM chains, so Hardhat, Foundry, viem, wagmi, ethers, and OpenZeppelin's contract libraries work across both.
Hardhat 3 is the notable shift here. It ships Foundry-compatible Solidity tests suitable for unit, fuzz, and invariant testing while keeping TypeScript integration tests, and it parses foundry.toml so the two toolchains share artifacts. The old "pick Hardhat or pick Foundry" framing has collapsed — many teams now run Foundry for fast unit tests and Hardhat for deployment orchestration and integration work. Hardhat 3 also added multi-chain simulation with first-class OP Stack chain types, which matters if your roadmap includes an L2 deployment.
How to choose
Build on Ethereum L1 when settlement assurances dominate: high-value DeFi, protocol infrastructure, tokenized assets where the security budget is the product. Post-Fusaka, L1 also makes more sense as a data availability layer for your own rollup than it did a year ago.
Build on Polygon PoS when transaction cost and confirmation speed dominate: payments, stablecoin rails, consumer apps, gaming, loyalty systems. Polygon's strategic direction after retiring zkEVM points squarely at stablecoin payments, real-world assets, and the Agglayer as a cross-chain settlement layer — build there if that direction matches yours.
The honest answer for most teams is both, with the same repository. Write contracts once, test with Foundry-compatible Solidity tests, deploy to Sepolia and Amoy in the same CI run, and let cost and finality requirements decide which mainnet each component lands on. Just keep a calendar reminder for Sepolia's September 2026 end-of-life — the last two testnet retirements caught plenty of teams with broken pipelines.