Ethereum layer 2: Starknet weighs leaving

An Ethereum layer 2 is a network of its own that executes transactions outside the
Ethereum blockchain, bundles them
and settles them on Ethereum in batches. It is faster and cheaper than Ethereum itself and hangs on Ethereum for
exactly that reason: the security comes from below. Starknet put that dependency to a vote among its token
holders last week. The record shows this: the vote ran from October 15 to 22, 2026, 93.52 percent of the voting
power cast across 3,236 votes came out in favour of researching an exit from the layer 2 model, and a move is not
thereby decided, because that would need a second, binding vote. Our assessment follows further below: two things
matter for holders over the coming months, the draft whitepaper in November and the question of which network
actually carries their token.

This piece answers three things: what a layer 2 is technically and which designs exist, how much capital the
individual networks actually secure today, and what to check in your own position when a network changes its
foundation. We compiled the capital figures ourselves, which is why they appear here as a table rather than as an
assertion.

Rollup, validium, sidechain: three designs sit behind the term Ethereum layer 2

“Layer 2” is not a protected term but a collective label. Underneath it sit designs with very different
security assumptions, and the difference decides how easily you reach your money and whom you trust along the
way.

A rollup is a network that executes transactions outside Ethereum, bundles them into batches
and publishes the associated data or a cryptographic proof on Ethereum. Because the data sits on Ethereum, anyone
can reconstruct the state of the rollup from it.

An optimistic rollup initially takes the bundled transactions to be correct and leaves a
challenge window open in which any participant can demonstrate an error. Arbitrum One, Base and OP Mainnet work
this way. A ZK rollup instead submits a mathematical validity proof that Ethereum verifies before
the state is adopted; a challenge is then unnecessary. Starknet, Linea, Scroll and ZKsync Era belong to this
group.

A validium publishes the validity proof on Ethereum but keeps the transaction data outside.
That lowers costs and shifts the security assumption: whoever withholds the data can stop you from proving your
claim. A sidechain, finally, is an independent blockchain with its own validator set, attached to
Ethereum only through a bridge. Polygon PoS, whose POL token many investors still remember from the layer 2 debate
of 2021, falls into this category and is explicitly not listed as a rollup by the analytics platform L2Beat
today.

How to recognise the design in practice

Three features are enough to place a network. How long a withdrawal back to Ethereum takes: a week points to an
optimistic rollup, hours to a ZK rollup, minutes to a sidechain. Where the data sits: a rollup writes it to
Ethereum, a validium does not. And who builds the blocks: a rollup with a single sequencer can be halted, while a
sidechain with its own validator set has a different attack surface. The classification of each network is
publicly documented at L2Beat, together with the maturity stage that indicates how far the network has moved away
from central control.

Starknet’s vote permits research into an exit, not the exit itself

Starknet is a ZK rollup and works with so-called STARK proofs, a form of proof the company StarkWare has been
developing since 2019. From October 15 to 22, 2026, a vote ran through the network’s
governance portal on whether
Starknet should research the path to an independent, quantum-safe layer 1. The options were yes, no and
neutral.

The result: 93.52 percent of the voting power across 3,236 votes cast backed the research. StarkWare itself
says it abstained and left the decision to the community. Five points matter for context, and they get lost in
many short reports.

• The vote is non-binding. It commissions a working group from StarkWare and the Starknet
  Foundation with technical research and planning. An actual switch would need a separate, binding vote once the
  results are presented.
• The token stays. No new token is planned; STRK is to remain the network’s native token in
  both scenarios.
• The bridges stay open. The existing connections between Ethereum and Starknet are to remain
  usable unchanged during the research phase, and nothing changes for users at first.
• A draft whitepaper is announced for November and is to be published for review.
• No date for the switch appears in the proposal. The vote commissions research, with no
  migration date and no staged plan.

The market took the news in mixed fashion. When the vote opened, STRK gained by double digits. On Thursday the
token stood at $0.054, 16.1 percent below the previous day, with a market capitalisation of around $410 million.
The wider market gave up 6.6 percent on the same day, Ether 5.5 percent. STRK therefore fell markedly harder than
the market it moves in.

Seven days sit between the withdrawal request and the balance on optimistic rollups

Anyone moving from a layer 2 back to Ethereum runs into the challenge window. At the optimistic rollups Arbitrum
One, Base and OP Mainnet it lasts around seven days. During that time somebody could demonstrate that the published
transaction batch was faulty; only afterwards does the bridge release the balance on Ethereum. The wait is the
price of the design: without a challenge window there would be no mechanism to correct a false state.

Our piece of September 14, 2026 on the
waiting time for layer 2 withdrawals
describes the process in detail, including the cases in which up to ten days add up because the request slips into
the next batch. With ZK rollups the challenge window falls away, but you wait for the next proof the network
submits to Ethereum. That takes hours rather than minutes, depending on the network.

Fast third-party bridges and their price

Third parties offer fast bridges that pay you out immediately and wait themselves. They charge a fee for it and
you carry their default risk: if the provider fails mid-process, your money hangs on its contract rather than on
Ethereum. For larger amounts it is worth taking the slow route over the official bridge and planning for the
wait.

Post-quantum cryptography is the argument: STARK proofs compute without elliptic curves

The drive behind the proposal is not a quarrel with Ethereum but cryptography. Ethereum accounts sign their
transactions today with signatures over an elliptic curve. A sufficiently large quantum computer could derive the
private key from a public key; the algorithm needed for that has been known since the nineties, the matching
machine does not exist. STARK proofs, by contrast, rest on hash functions, and those are regarded as considerably
more resistant to quantum attacks.

Starknet chief Eli Ben-Sasson names 2027 as the target by which his network should be quantum-safe. His
argument: as a layer 2, Starknet stays tied to Ethereum’s timetable, because settlement happens on Ethereum and
therefore also hangs on its signature scheme; as a layer 1 of its own it could take the step earlier. That is his
reasoning, not our judgement.

Against it stands what an exit costs. Ethereum is working on quantum-safe signature schemes itself, and the
security a rollup inherits from Ethereum today would have to be raised by an independent network under its own
steam. Anyone wishing to follow the debate will find the technical arguments in the network’s forum; the announced
whitepaper is more reliable than any short report.

Base secures $15.2 billion, Starknet $395 million: the weight is unevenly spread

The size of a layer 2 shows in the capital that sits in the network and is secured across its bridges, and
barely at all in the token price. We compiled the values at L2Beat on Thursday evening, for all Ethereum rollups
with a meaningful balance. The right-hand column shows the change over seven days.

Base and Arbitrum One together secure $25.4 billion. That is 64 times Starknet, and they are two optimistic
rollups, the design with the long wait. The technically more elegant group of ZK rollups carries less than a
twentieth of that between them. Starknet accounts for just under one percent of the capital L2Beat tracks across
all layer 2 networks, around $40 billion.

That distribution explains why the vote is a small story for the market as a whole and possibly a large one for
your own position: holders of Base or Arbitrum are not affected by Starknet’s path. Holders of STRK, or of a
balance sitting on Starknet, have a network in their portfolio whose foundation is up for discussion.

Four of the five layer 2 projects on our 2024 list are no longer tracked by L2Beat

A look into our own back catalogue shows how quickly such lists age. Our overview of the five largest Ethereum
layer 2 projects, from September 19, 2024, named Polygon, Loopring, OMG Network, Skale and ZKSwap, ordered by
market capitalisation. We checked all five against today’s data.

Loopring, OMG Network, Skale and ZKSwap do not appear in the current L2Beat listing at all: four out of five.
Polygon is still there, but as Polygon PoS and in the “Other” category, so explicitly not as a rollup; POL stood at
$0.0902 on Thursday. The four networks that secure the most capital today, Base, Arbitrum One, OP Mainnet and
Starknet, did not feature on the 2024 list. Base was a few months old at the time.

From that follows a lesson that reaches beyond Starknet: a ranking of layer 2 projects is a state, not a
property. What holds up is the design, the maturity stage and the secured capital, because all three can be
verified. A project name on a two-year-old list says nothing about today.

Starknet’s secured capital fell 39.9 percent in a week, the sector 3.4

The most striking value in the table above is Starknet’s weekly loss. Judging it needs the comparison, because
secured capital is measured in dollars and therefore falls along with the market. Across all layer 2 networks the
figure declined by 3.4 percent over seven days, the Ether price by 2.7 percent over the same period. Most of the
sector decline is simply the market.

Starknet’s 39.9 percent are twelve times that move. Base lost 2.2 percent, Arbitrum One 3.0, Linea 4.0. An order
of magnitude like Starknet’s is not explained by a price fall; that happens when capital leaves the network. The
vote ran in precisely this week. Whether it is the reason cannot be read from these figures; what can be read is
the order of magnitude, and for a network of this size it is considerable.

Add to that the token price. STRK lost 16.1 percent on Thursday while the wider market gave up 6.6 percent. The
same caveat applies here: one day is not a trend. But capital outflow and price weakness point the same way, and
both fall in the week of the vote.

A switch to its own layer 1 shifts the bridge, the settlement and the security assumption

What exactly would change if the rollup became an independent blockchain? Three things, and they hang
together.

Settlement. Today Starknet publishes its proofs on Ethereum, and Ethereum thereby decides which
state applies. A layer 1 of its own decides that itself, through a validator set of its own.

Security. A rollup inherits Ethereum’s security without paying for it. An independent
blockchain has to buy it, usually through rewards to validators, financed from issuance or fees. That burden is
carried in the end by the token.

The bridge. The official connection between Ethereum and Starknet is today part of the rollup
design; your claim on Ethereum follows from the published state. Between two independent blockchains it becomes an
ordinary bridge with its own operators and its own assumptions. It is the same difference that already separates a
rollup from a sidechain.

Our assessment: the decisive date for holders lies in November, with the draft whitepaper, and not with this
week’s vote. Three documented points support that. The vote was non-binding and set no date. The token remains
STRK in both scenarios, and a swap is not up for discussion. And secured capital of $395 million is small enough
that a shift in the security assumption does not move the wider market, yet large enough to count for your own
position. Against it stands the fact that a whitepaper is not yet a decision and that the research phase can
stretch over quarters; anyone betting on a quick result is likely to wait. None of this is a recommendation to buy
or sell, and a total loss is possible at any time with a token of this size.

MiCA decides which exchange may sell you layer 2 tokens in Germany

The smaller layer 2 tokens share a practical problem: they are harder to trade in Germany than Ether. Since the
EU regulation on markets in crypto assets, a provider serving clients in Germany needs authorisation as a
crypto-asset service provider; BaFin supervises this here, and authorisations granted appear in public registers.
What that means for operators we set out in our overview of MiCA obligations for crypto companies.

In practice it means this for you: a token tradable only on exchanges without EU authorisation, or only on
decentralised venues, shifts custody and legal recourse onto you. At an authorised exchange you have a counterpart
under German supervision; at a decentralised exchange you have your private key and nothing else. Which houses
hold an authorisation and which tokens they list is shown in our comparison of
regulated crypto exchanges.

Before buying a small layer 2 token, check two things: whether the provider holds an authorisation in the EU,
and whether it will take the token back. A trading pair with liquidity in only one direction becomes a problem on
the way out. Starknet’s weekly figures above are a good occasion to work through this for your own positions.

One year of holding separates tax-free from taxable, layer 2 tokens included

The same rule applies to layer 2 tokens in Germany as to Ether and Bitcoin. Gains from a sale belong to private
disposal transactions under section 23 of the Income Tax Act. If more than a year lies between purchase and sale,
the gain stays tax-free. Inside the year, the exemption threshold of 1,000 euros per calendar year applies; exceed
it and the entire gain is taxed at your personal income tax rate.

A swap counts as a sale, and the period runs per purchase

Two points are regularly overlooked. First, a swap counts as a sale: swapping STRK for Ether means selling STRK,
even though no euro changes hands. Second, the one-year period runs per purchase, not per token. Five purchases
over five months produce five periods, and the tax office expects a traceable allocation. Buying, swapping and
bridging across several networks produces a history that is hard to reconstruct after the fact. Download the
transaction data from your exchanges regularly, then, rather than waiting until spring.

A note of caution: the tax treatment of individual events in this area is in parts not conclusively settled, and
this section does not replace tax advice. For the basic cases of purchase, sale and swap the legal position is
clear.

Ethereum layer 2: November brings the whitepaper, not the move

The vote at Starknet has triggered a research phase and nothing else. For your own position that means less
haste and more homework: know the design, know the waiting time, check tradability. Three steps make sense now.

1. Sort your holdings. For each layer 2 token in your portfolio, note which network the balance
   sits on and which design that network uses. Whether you can still sell it at a sensible price at all is shown by a
   look at the comparison of crypto
   exchanges, with their trading pairs and fees.
2. Separate custody. A balance you will not need for longer than the challenge window does not
   belong on an exchange. Which devices support layer 2 networks directly and which only through software in front of
   them is set out in our comparison of hardware
   wallets.
3. Document the periods. Record the purchase date, the amount and the value for each purchase so
   the one-year period stays provable. Which tools bring bridges and swaps across several networks together is shown
   in our overview of tax and portfolio
   tools.
4. Mark the date. The draft whitepaper from the working group of StarkWare and the Starknet
   Foundation is announced for November. It contains what the vote did not: a plan with numbers.

The secured-capital data in this piece comes from the public listing at
L2Beat, the prices from
Thursday evening.

(As of October 22, 2026. This article is not investment advice. Prices and fee structures change; check the
terms with the provider before you buy.)