Starknet Asks Its Stakers Whether to Research Leaving Ethereum Over Quantum Risk

October 15, 2026 – The Starknet Foundation opened a non-binding poll asking STRK stakers and delegators whether Starknet should research leaving Ethereum to run as an independent post-quantum blockchain. A Yes would only start research, and any decision to make the move would go through a separate governance process, Unchained reported from the proposal text.

Starknet is a layer-2 network that executes transactions itself and settles them on Ethereum, where it publishes its state data as blobs. The Snapshot poll opened at 09:32 UTC and closes at 09:32 UTC on October 22. By 14:40 UTC on the first day, 293 votes with about 179.8 million in voting power had been cast, 99.9% of it in favor of research, according to Unchained. The poll has no quorum requirement.

About 1.54 billion STRK was staked with 137 active validators on October 17, worth about $117 million at that day’s price of $0.076, according to Staking Rewards. STRK had reached an intraday high of $0.1265 on October 11, AltcoinBuzz reported, a price at which the same stake would have been worth just under $200 million. Ethereum had about 43.7 million ETH staked on October 6, according to Crypto Briefing, worth about $103 billion at the $2,363 price Fortune reported on October 15.

The Governance Post and the Poll Options

Henri of the Starknet Foundation, who posts as henrli, set out the reasoning in the forum post that accompanies the poll. The parts of Starknet under its own control can be made quantum-resistant soon, he wrote, while the bridge and data availability depend on Ethereum’s timeline. By becoming an independent L1, Starknet would remove that dependency, he added, though “no timeline for that exists yet.”

Henri wrote that two things are fixed whatever the outcome. STRK remains the token, and there will be no new one. The bridges from Ethereum to Starknet stay open. If the result is No, Starknet continues its post-quantum work as a layer 2. If it is Neutral, the forum discussion stays open until the missing answers are in.

Five of the seven community replies to Henri’s post by October 16 supported research. Cumulo, which runs a Starknet validator, wrote that it had voted Yes and offered to test early releases. A commenter posting as ceri1 questioned whether Starknet’s current staking set could replace the security Ethereum provides. Henri replied that research may help answer such questions if the vote is positive.

StarkWare’s Case for an Independent Chain

StarkWare, the company that builds Starknet, raised the idea on October 8. Co-founder and chief executive Eli Ben-Sasson asked on X whether Starknet becoming an L1 for post-quantum agility was a good or a bad idea, referring to a talk he gave at Token2049. He listed three requirements for a chain to be future-proofed: the right cryptography, crypto-agility and L1 status, according to Unchained.

Starknet’s own account posted the same day that it was considering becoming an L1, which it said would make Starknet the first fully quantum-resistant network, “with 2027 as our target.” Ben-Sasson wrote that as a layer 2, Starknet can be only as quantum-safe as Ethereum, its base layer, Decrypt reported. On October 10, StarkWare posted that “Starknet becoming an independent L1 is our clearest path forward.”

“Quantum-proofing crypto is urgent, so we must move at lightning speed,” Ben-Sasson told Cointelegraph on October 15.

Starknet’s Post-Quantum Roadmap and Ethereum’s Deadline

StarkWare published a three-phase post-quantum roadmap for Starknet on June 30. The first two phases, in which StarkWare replaces the Pedersen hash with BLAKE2 and migrates existing contracts, are within Starknet’s control. The roadmap did not mention becoming an L1.

The third phase covers two dependencies on Ethereum that StarkWare said would be resolved as Ethereum’s own migration progresses. One is the set of Cairo VM syscalls for the secp256k1 and secp256r1 curves that support the Ethereum bridge and L1-to-L2 messaging. The other is the KZG commitment over the state data Starknet publishes in Ethereum blobs.

The Ethereum Foundation’s Protocol cluster said on September 7 that it aims to make Ethereum quantum-resistant across its execution, consensus and data layers by December 2029. The cluster plans for Q-Day as early as 2030, which it called a deliberately aggressive assumption. It will treat the December 2029 deadline as fixed until a January 2027 reassessment with outside experts.

My Analysis

Voters are deciding only whether Starknet should research a move. On the evidence published so far, I think the move itself would be a poor trade. Starknet would give up settlement on Ethereum, where about $103 billion of ETH is staked. It would rely instead on its own validators, whose staked STRK was worth about $117 million on October 17. In return, StarkWare could set the timing of Starknet’s migration without waiting for Ethereum. Starknet users would benefit from that mainly in one scenario: a CRQC that can forge Ethereum’s validator signatures before the Ethereum Foundation and the client teams replace them with post-quantum signatures. Starknet’s users would also still hold elliptic-curve keys in their existing accounts after any move, and those keys do not depend on Ethereum at all.

What the 2027 Target Refers To

In its October 8 post, Starknet gave 2027 as its target for full quantum resistance and said it could reach that target by becoming an L1. Ben-Sasson wrote that Starknet could get there by 2027, two years before the end-2029 date he cited for Ethereum, if it did not have to wait for Ethereum. Neither Starknet nor Ben-Sasson gave a date for becoming an L1.

Some news outlets reported 2027 as the date for an L1 move. AMBCrypto’s summary said Starknet would migrate from Ethereum and become an L1 in 2027, and KuCoin’s news feed ran the story under a headline about a 2027 L1 migration. The date comes from Ben-Sasson’s and Starknet’s posts on X. The governance post that accompanies the poll says no timeline exists. ceri1 asked whether 2027 is a commitment or a contingency.

If Yes wins on October 22, the next steps Henri described are a research effort, more detailed proposals and further polls. The proposal requires a separate governance process for any transition. An independent chain would also need validators that produce, attest and prove blocks, which Starknet’s documentation describes as the end point of its staking rollout. Staking is in its second phase today. StarkWare targets late 2026 for Staking v3, in which validators begin validating blocks.

Starknet would also need a narrower definition to call itself the first fully quantum-resistant network. The Quantum Resistant Ledger has used hash-based XMSS signatures since it launched in 2018, as I noted in my October 5 article on crypto-agility for digital assets.

What Starknet Would Give Up by Leaving Ethereum

A proof-of-stake chain’s economic security is the value staked with its validators, which an attacker would have to acquire, or put at risk of slashing, to rewrite or halt the chain. Today five permissioned sequencers controlled by StarkWare order Starknet’s transactions, according to L2BEAT. StarkWare proves batches of those transactions and publishes the STARK proofs and state data to Ethereum, whose validators order and finalize the Ethereum transactions that contain them. Reversing a Starknet state update after Ethereum finalizes it would require reorganizing Ethereum. Whoever holds Starknet’s upgrade keys has a separate route to the same contracts through a malicious upgrade.

In a post on October 16, ceri1 called an L1 move “a transfer of dependency, not a gain in decentralization,” and argued that the current STRK staking set cannot take over Ethereum’s role yet. Starknet’s staking protocol is in its second phase, according to its documentation, which lists a 20,000 STRK minimum for validators whose duty in this phase is to attest to blocks with a full node. The same documentation describes Starknet as still centralized, with block production, attestation and proving moving to validators in stages.

On the October figures, Starknet’s staked STRK was worth less than 0.2% of the ETH staked on Ethereum even at STRK’s October 11 peak, and about 0.1% at the October 17 price. Starknet validators can also accept tokenized bitcoin as stake. STRK holders voted in 2025 to cap bitcoin’s share at 25% of consensus power. The comparison is one of scale, not of the cost of an attack, which also depends on how concentrated the stake is, how slashing works and how much STRK an attacker could buy.

StarkWare’s design for validators to sign blocks is still a draft. Under SNIP 33, which Ohad Barta of StarkWare posted in July 2025, the 100 largest stakers would form a signing committee. A block would be confirmed once members holding more than two-thirds of the committee’s stake had signed it. Members holding more than a third could therefore stop the normal quorum from forming. The draft also includes an optional fallback in which StarkWare’s sequencer produces blocks without validator signatures for 12 hours to 3 days when too few blocks are confirmed.

Nodes that verify STARK proofs would still reject an invalid state transition on an independent Starknet, the property StarkWare calls soundness in its roadmap. Soundness does not cover finality, transaction inclusion or data availability. On an independent chain, Starknet’s validators would have to provide all three. Today Ethereum provides finality and data availability for Starknet’s state updates, but not inclusion. L2BEAT notes that Starknet has no user-callable forced-inclusion function, so a user whose transactions are censored has to ask a minority of Starknet’s Security Council to include them.

ceri1 asked what economic security an independent Starknet would target at launch. A second commenter asked whether Starknet’s own validator set could replace Ethereum’s security without giving up decentralization. Both questions are still open.

Starknet benefits from Ethereum’s larger stake only while no one can forge Ethereum’s validator signatures. Ben-Sasson is right that this is a dependency: Ethereum’s validators sign with BLS, and a CRQC that could forge those signatures would let an attacker impersonate validators without owning their ETH. That is the strongest case for leaving. The Ethereum Foundation plans to make Ethereum’s consensus layer quantum-resistant by December 2029 and has a contingency milestone, minimum viable post-quantum readiness, designed to keep Ethereum operating through Q-Day with reduced guarantees.

Which Ethereum Dependencies Remain After a Move to an L1

StarkWare’s roadmap names two Ethereum dependencies. The first, the Cairo VM’s secp256k1 and secp256r1 syscalls, supports the Ethereum bridge and L1-to-L2 messaging. StarkWare estimated their removal at one month after Ethereum migrates, and wrote that the bridge requires them until then.

Henri wrote that the bridges from Ethereum will stay open. StarkWare has not described how an independent Starknet’s bridge would differ from today’s design, which needs those syscalls. As an independent chain, Starknet would no longer settle on Ethereum. Assets bridged from Ethereum through Starknet’s canonical bridge would still be locked in contracts there, so their protection would depend on Ethereum’s migration either way. STRK itself was first deployed as an ERC-20 contract on Ethereum. StarkWare and the Foundation have not said how its supply would be held after a move.

The bridge contracts also depend on the keys that can upgrade them. According to L2BEAT, a 9-of-12 Security Council can upgrade Starknet’s core contract and its ETH and STRK bridges without delay, while StarkWare multisigs handle the regular upgrade path and other bridge escrows. L2BEAT also warns that critical contracts can be upgraded by an externally owned account, which signs transactions with a secp256k1 key. A CRQC could forge those signatures. Starknet would keep that exposure as an L1 unless it moves the keys to post-quantum schemes.

The second dependency is the KZG commitment over the state data Starknet publishes in Ethereum blobs. StarkWare wrote that a KZG compromise would affect the transparency of data availability, not the soundness of state transitions, because those are proven with STARKs. In its migration register, StarkWare also notes that Starknet can still publish its data as calldata. A chain whose proofs stay sound can still stall if its data cannot be retrieved. StarkWare should say how quickly the calldata fallback could be switched on, who would decide and what it would cost at Starknet’s current volume.

Of the two dependencies StarkWare lists, Starknet would drop the data-availability one by becoming an L1 and keep the bridge one for as long as the bridges stay open.

Starknet’s Quantum Exposure That Does Not Depend on Ethereum

Starknet’s standard account contracts, including OpenZeppelin’s, verify signatures over the Stark curve. Stark-curve signatures are a form of elliptic-curve cryptography, so a CRQC running Shor’s algorithm could derive an account’s private key from its public key. Starknet’s prover would still produce a valid STARK proof for a transaction signed with a key derived that way. The proof shows only that the transaction followed Starknet’s rules, including its account’s signature check, which accepts any signature made with the right private key, however that key was obtained.

With native account abstraction, a user can move to a post-quantum account without a protocol change. StarkWare’s register records that OpenZeppelin has demonstrated a Falcon-512 account on Starknet mainnet, using an experimental, unaudited implementation to send 0.01 STRK in one transaction. StarkWare’s roadmap notes that S2morrow built its own Falcon-512 account in Cairo. In the register, StarkWare also states the limit: existing accounts do not change automatically. Wallet support, key management, recovery, custody, audits and adoption remain separate work.

In its roadmap, StarkWare also set out to replace the Pedersen hash, which inherits elliptic-curve assumptions, with BLAKE2 in the state tries, contract-address derivation and the OS configuration hash. StarkWare estimated that most of that work would take about two months but did not say when it would begin. When StarkWare last updated the register on August 19, only the OS program and configuration hashes had moved, in v0.14.3. In the register, StarkWare lists the BLAKE2 changes to the contract and storage tries and to address derivation as “published direction,” and none of those changes is on mainnet. StarkWare has published no specification or tooling for migrating existing contracts’ storage. The notes for v0.14.4, which reached mainnet on October 5 according to Crypto Briefing, cover full-block proving and gas changes.

Henri described the core protocol as provably safe from quantum computers. In a September blog post on ECDSA.fail, StarkWare went further, writing that hash-based proofs mean Starknet’s proof system “never has to change.” In the register, StarkWare limits the advantage to the proving layer and writes that its security still depends on the chosen hashes, parameters and implementation.

The remaining hash, storage and account work on Starknet does not depend on Ethereum and would not change on an L1. The tries, the address derivation, the legacy storage tooling and the account migration are within the control of StarkWare and Starknet’s users today.

On Starknet, any user can add a post-quantum account without a protocol change. Retiring the old curve across the whole network is harder, because each account’s validation logic is its own code. As I argued on October 5, citing Bitcoin Core developer Pieter Wuille, while much of a chain’s value is held in accounts that use the old curve, holders who migrated still lose if that curve breaks, because their coins lose value with everyone else’s. A suspected break could cause substantial market damage on its own.

To call itself fully quantum-resistant, Starknet needs an enforceable policy for Stark-curve accounts whose owners never move, including how a late mover would prove ownership once the old key can be forged. Neither StarkWare nor the Starknet Foundation has published one.

Hash, Account and Consensus-Signature Migrations on an Independent Starknet

In my October 5 article, I argued that chains are planning one post-quantum change when they will need several, and gave six reasons, of which quantum computing is only one. Starknet’s proposal is a public test of that argument.

Starknet’s post-quantum migration already includes a hash change. StarkWare is replacing Pedersen because the hash inherits elliptic-curve assumptions, while describing the proof system as one that will never need replacing.

As an L1, Starknet would raise the stakes of a migration it has barely started. StarkWare’s roadmap already calls for post-quantum consensus signatures for Starknet’s own validators and names Falcon-512 as an example. NIST has not yet finalized Falcon as FN-DSA in FIPS 206. As a layer 2, Starknet relies on Ethereum to finalize what those validators sign; as an L1, those signatures would be its finality. The Ethereum Foundation’s Protocol cluster wrote that Ethereum’s consensus layer “cannot rely on cryptographic agility,” because its cryptography and aggregation schemes change only through a hard fork. For that reason, the cluster hardens consensus cryptography before rollout and ships the complete design at once.

An independent Starknet would become responsible for coordinating changes to its validators’ signature scheme, on top of its account and hash migrations. The Ethereum Foundation and the client teams do that work for Ethereum, the layer Starknet settles on. As an L1, Starknet would have to do it itself, starting from a candidate scheme whose standard is not final.

Starknet starts this test better placed than most chains. Its users can change their accounts’ signature schemes without a fork, and its proofs do not rely on elliptic curves. StarkWare also publishes a register that states what is not done. Seven co-authors of the ECDSA.fail paper are affiliated with StarkWare or the Starknet Foundation. The paper reports a point-addition circuit that scores less than half of Google’s March figure, though the two figures are counted under different rules. Ben-Sasson lists crypto-agility as one of his three requirements. As an L1, Starknet would take on work in the one layer where, by the Ethereum Foundation’s account, agility is hardest to build.

I have seen no published evidence that leaving Ethereum is necessary to meet the urgency Ben-Sasson described. On October 8 he wrote that the quantum threat “may be MUCH closer than we think,” and pointed to AI as a second threat. ECDSA.fail’s contributors, whose work I covered in June, improved an arithmetic component of Shor’s algorithm. StarkWare’s own blog post says the circuits were measured, not run, because no hardware exists to run them. The ECDSA.fail paper gives no date for an attack. The Ethereum Foundation plans for Q-Day as early as 2030 and calls that assumption deliberately aggressive.

The account migration should still start now. Moving holders to new account types depends on each owner acting. In the record I compiled on October 5, planned cryptographic changes on public chains took years. Starknet can start that migration without Ethereum.

What Starknet Should Publish Before Any Transition Vote

My recommendation is that Starknet finish the post-quantum work that needs no L1 before it asks holders to approve one. Three pieces of that work are within StarkWare’s control now:

• the BLAKE2 state tries and contract-address derivation, which StarkWare estimated in June at about two months of work;
• tooling to move existing contracts’ storage off Pedersen-based key derivation;
• an enforceable policy for Stark-curve accounts, including how owners who move late would prove ownership.

Before any transition vote, StarkWare and the Foundation should publish:

• the economic security an independent Starknet would target at launch, in staked value and validator count, compared with Ethereum’s;
• which of Starknet’s Ethereum dependencies end on the first day of an L1 and which remain while the bridges stay open;
• a post-quantum plan for the keys that can upgrade Starknet’s core contract and bridges;
• a specification of validators’ new duties, the consensus signature scheme they would use and the process for changing it.

The Foundation should also publish how much of the final tally came from voting power it delegated. When it published results for its Q3 2025 poll, the option for post-quantum signatures had about 35.7 million in voting power with Foundation delegates counted and about 106,000 without them.

Exchanges, custodians and funds that hold assets on Starknet can start account inventories, migration design and controlled testing now. OpenZeppelin has shown on mainnet that a post-quantum account works on Starknet, but its implementation is experimental and unaudited and is not a production custody solution. Those firms can begin that preparation without waiting for Ethereum or for the October 22 result.

The poll closes at 09:32 UTC on October 22. The next fixed date on Ethereum’s side is the Protocol cluster’s January 2027 reassessment of its December 2029 target, while Starknet has no fixed date for an L1.

From the author