Every so often I catch myself trying to explain why a Layer 2 token I actually like has done nothing while the chain underneath it prints activity. The chain is busy. Users are paying fees. The rollup is, in a real sense, profitable. And the token just sits there. Once you look at where the money actually goes, the disappointment stops being a mystery and starts being an accounting problem, which is a much more useful thing to have.
Where the fee actually goes
A rollup runs a business with two line items on the cost side and one on the revenue side, and most people only ever look at the revenue. When you submit a transaction on an L2, you pay a fee. That fee splits into two parts. One part is the L2 execution fee, which is what you are paying the sequencer to order your transaction, run it, and produce a block. The other part is the cost of getting your data back down to the base layer so the rollup stays verifiable, which is the data availability cost. On an Ethereum rollup that second part is what the operator pays to post to L1, whether through calldata or blob space.
So the unit economics look roughly like this. The sequencer collects the whole fee at the top. Out of that, it pays the data availability bill and whatever proving costs apply for a zk system. What is left over is the operating margin. Historically that margin has been thin during quiet periods and genuinely fat during congestion, because DA costs do not scale linearly with how much people are willing to pay to get in. When blockspace demand spikes, users bid the execution fee up and the DA cost per transaction can actually fall as batches get denser. That gap is the entire business.
Here is the part that trips people up. That margin, today, on almost every major rollup, flows to the entity operating the sequencer. It does not touch the token. The token holder is not a claim on that cash flow. In most cases the token is a governance right, a gas token, or a mechanism for future decentralization that has not arrived yet. You are holding a claim on a decision, not a claim on the earnings.
Why most L2 tokens are governance claims on nothing
I do not say that to be cynical. It is just structurally true for a lot of them right now. If the sequencer is centralized and run by the core team, and the fee margin lands in a company treasury or an operating wallet, then the token's cash-flow rights are whatever governance later chooses to grant, which today is often nothing. Governance over a treasury it does not control, or over parameters that do not route revenue, is governance over an empty room.
This is the honest reason the sector has lagged ETH for long stretches. ETH is a claim on base-layer settlement and, through the burn mechanism, on a slice of aggregate activity across everything built on it, including the rollups. An L2 token, absent a fee switch, is a claim on the promise that someday the people keeping the margin might share it. Markets are not stupid. They price the promise at a discount to the cash flow, and the discount is large because the switch is genuinely optional and politically loaded.
Modeling what the switch is worth
The useful exercise is to stop guessing and build the number yourself. You do not need perfect data, you need a defensible range. Here is the workflow I use when I want to sanity-check whether a token is cheap, fairly priced, or a story.
- Find the annualized sequencer revenue. Take gross fees paid by users over a representative period and annualize it. Use a stretch that includes both quiet and busy weeks so you are not extrapolating a spike.
- Subtract the data availability and proving costs over that same window. What remains is the operating margin. This is the pool a fee switch could theoretically distribute.
- Decide what fraction a realistic fee switch would route to holders. It is almost never 100 percent. Some goes to sequencer operators, some to a treasury, some to whoever runs proving. A conservative pass-through assumption is more honest than a generous one.
- Divide that holder-facing cash flow by the fully diluted value, not just the circulating market cap. Unlocks are real and they dilute your claim. This gives you an implied yield if the switch flipped tomorrow at current prices.
- Compare that implied yield to what you would demand for holding a volatile, governance-dependent asset. If the switch has to flip and pass through most of the margin just to reach a mediocre yield, the token is expensive. If a modest partial switch already clears your hurdle, it is interesting.
The step people skip is the fully diluted denominator. A token can look like it trades at a sane multiple of margin on circulating supply and be absurd once you account for everything that will unlock. Do the math on FDV or you are lying to yourself.
What would actually change the picture
A few things move a token from story to something with a floor, and they are worth watching in this order. Real sequencer decentralization, where ordering and revenue are no longer controlled by a single operator, is the precondition for the token meaning anything, because you cannot route margin to holders from a sequencer nobody but the team controls. After that, a concrete fee-switch proposal with an actual pass-through percentage, not a vague governance sentiment, is the moment the market gets a number to price. Revenue-sharing experiments, staking that captures a cut of sequencer income, buyback-and-burn funded by real fees rather than treasury sales, all of these are versions of the same thing, which is connecting the margin to the token.
The failure mode to avoid is buying the token because activity on the chain is going up. Activity going up raises sequencer revenue, and if that revenue never reaches you, higher activity just makes the operator richer while your governance claim stays worth the same nothing. I have watched people confuse chain growth with token accrual more times than I can count, and it is always the same mistake. Usage is necessary. It is not sufficient. The bridge between usage and your token is the fee switch, and until that bridge is built with a real number attached, you are pricing a maybe.
So when someone pitches me a Layer 2 token, I do not argue about the tech. I ask two questions. Does the sequencer margin have any path to holders that does not depend on a vote that has not happened, and what is the implied yield on fully diluted value if the most realistic version of that path opened tomorrow. If both answers are soft, it is a bet on narrative, which is fine as long as you know that is the bet you are making. If one of them is starting to firm up, that is the one worth the modeling time.