For the last few years, the market has treated “Ethereum scaling via rollups” as almost a self-evident bullish thesis. If Ethereum is the settlement layer of the internet, and Layer 2s are how it reaches billions of users, how could you not be long that story? Throughput goes up, fees go down, real users and real enterprises finally get an infrastructure that feels usable.

But from an investor’s perspective, the core issue is not whether the tech works. It clearly does. The core issue is where value actually accumulates. Layer 2s are processing more and more of Ethereum’s activity, while fee levels on L1 have collapsed and sequencer economics have become a real business for a small set of operators. At the same time, most L2 governance tokens have massively underperformed, weighed down by high FDVs, heavy unlocks and weak links to any kind of sustainable cash flow.

This is the tension that runs through the entire L2 story: usage is winning, the architecture is broadly working, yet the token layer often looks structurally fragile. A strong network is not the same thing as a strong token. In this essay, we’ll treat those as separate objects and ask the only question that really matters to capital: are Ethereum L2s becoming the backbone of on-chain execution, while their tokens quietly ossify into low-quality governance scrip?

From an investor’s vantage point, the cleanest way to think about a Layer 2 is as an externalized execution environment that borrows Ethereum’s security and settlement. Users transact on an L2; funds are ultimately custodied and finality is ultimately anchored on Ethereum. Execution risk sits off-chain (or off-L1); consensus and data availability remain on-chain.

Rollups are the canonical mechanism for doing this. Conceptually, they:

The verification model is where the main split appears:

On top of that, you have adjacent designs:

Functionally, all of this targets the same triad of Ethereum pain points:

For a technologist, that’s the story. For an investor, it’s only the starting point. Reducing gas per user is great for adoption, but it is not automatically great for ETH, and it is certainly not automatically great for L2 governance tokens.

Ethereum’s original bottleneck was clear: limited blockspace, rising fees, and a UX that only made sense for high-value transactions. Rollups directly tackled this by:

On paper, this is a pure complement: Ethereum cements itself as the neutral, credibly neutral base; L2s become the high-speed, low-cost execution mesh above it.

In practice, the value picture is more nuanced.

As L2s have grown, L1 fee revenue has collapsed. Post-4844, the cost of posting data to Ethereum has fallen dramatically. That’s excellent for users and L2 operators, but it concentrates the economics upstream. Sequencers pay less to Ethereum, retain more margin on their own books, and L1 fee burn plus staking yield gets squeezed.

At the same time:

So L2s do clearly help Ethereum the protocol: more usage, more relevance, more institutional adoption, a cleaner L1 design. But they also raise the uncomfortable capital question: if execution moves off-chain and L2s optimize away L1 data costs, does ETH remain the best pure play on that growth? Or does value start to skew toward sequencer operators and infra?

More importantly for this essay, L2s helping Ethereum as a network does not automatically mean L2 tokens capture that help. The system can be win–win for Ethereum and users and still be neutral—or even mildly adversarial—for token holders whose only rights are governance and an airdrop memory.

The bullish scenario for L2s is simple, coherent and internally consistent:

In that world, the average user’s “Ethereum experience” is actually an L2 experience. Base, Arbitrum, Optimism Superchain, Polygon CDK chains and a small handful of ZK L2s emerge as the primary homes for on-chain activity.

The bearish scenario doesn’t dispute that the tech will work. It asserts that:

We have already started to see signs of this consolidation dynamic. A small cluster of L2s dominates TVL, usage and revenue; dozens of others are fighting for crumbs, heavily dependent on incentives and grant programs just to sustain baseline activity. Many are one bear market away from irrelevance.

The most realistic path is a middle one: the L2 theme does not die, but it becomes radically more selective. L2s as a category succeed; many individual L2s do not. Capital will have to pivot from “own the theme” to “own the winners,” and that’s where token design and value capture become non-negotiable.

Arbitrum is the flagship optimistic rollup in terms of DeFi depth and raw TVL. It hosts a large share of Ethereum L2 DeFi liquidity, with a broad protocol ecosystem across spot, perp, options, and structured products. Economically, it generates meaningful sequencer fees and protocol revenue, primarily off of that DeFi stack.

For ARB holders, though, the picture is thinner. Those fees accrue to the operator and the DAO treasury; ARB is a governance and incentives token, not a claim on cash flows. Combine that with a high FDV, large airdrop overhang and ongoing unlocks, and you get a classic “strong network, structurally weak token” situation.

Optimism runs its own optimistic rollup, but its deeper bet is OP Stack and the Superchain: a family of chains (Base, Mode, Zora and others) sharing the same stack, with native interoperability as a long-term goal. OP Mainnet’s standalone TVL is decent; the aggregate activity of OP Stack chains is far larger.

Economically, Optimism stands to collect a share of sequencer and infra revenue from OP Stack L2s, plus the usual grant/treasury flows. The OP token, again, is governance- and incentives-first today. The potential upside lies in whether Superchain revenue is ever routed back to token holders or stakers, or whether it remains a public good subsidized from inflation and treasury.

Base is the outlier that matters. It is an OP Stack chain run by Coinbase, with by far the strongest funnel: hundreds of millions of registered users, fiat on- and off-ramps, and direct CEX integration. It has quickly taken a dominant share of L2 transactions and DeFi TVL and, importantly, is already a profitable sequencer business.

From a token perspective, Base is the cleanest illustration of the theme’s dissonance: the most used and most profitable L2 has no native token. Value flows to COIN shareholders (through higher fee and trading revenue, and a stronger strategic position) and partially back into the Optimism ecosystem. If you’re holding L2 governance tokens under the assumption that “activity = token value,” Base is the counterexample you cannot ignore.

Starknet is one of the strongest pure technology bets in the L2 space: STARK-based proofs, Cairo VM and a future-proof vision of ZK scaling. That tech roadmap is real; the economic one is still very early.

TVL and transactional activity lag the optimistic contenders. Prover costs are non-trivial. Sequencer revenue exists but is not yet at meaningful scale. STRK’s initial token distribution, with aggressive unlocks and a high FDV, created heavy sell pressure; even after schedule adjustments, dilution remains a major concern. For now, this is a case where the long-term tech thesis is much stronger than the short- to medium-term token thesis.

zkSync has long been one of the flagship “ZK rollup with EVM-like UX” projects. It has attracted TVL, usage and developer interest, helped by incentives and airdrop speculation. But like Starknet, it still operates in a world where ZK proving costs weigh on unit economics, and where token design has not yet built a clean bridge between protocol revenue and token holder.

ZK launched with high FDV and a long tail of vesting for teams and investors. That creates mechanical unlock pressure for years, unless growth far outpaces dilution. Without a strong value accrual mechanism, the token risks staying a “farm it, dump it, move on” trade whenever the market heats up.

Polygon is a hybrid animal. The legacy PoS chain still hosts significant TVL and usage; Polygon zkEVM is an explicit L2; Polygon CDK is an appchain platform that lets projects spin up custom chains with Ethereum settlement.

POL is meant to be the coordination and security token for this multi-chain universe. Fees from PoS and CDK-based chains can, in principle, accrue to POL via staking, burning or other mechanisms. In practice, the picture is complex: revenues are spread across multiple environments, and value capture depends on how tightly they are wired back into POL. The investment case is less “one clean L2” and more “multi-chain infra bet with Ethereum as a root.”

Mantle is an optimistic, modular L2 that leans heavily on yield. Its TVL is anchored in staked ETH and yield-bearing strategies; its treasury and balance sheet play a central role in network economics.

The key differentiator: Mantle shares yield with MNT stakers. That makes MNT one of the few L2 tokens with an explicit income-like component. Of course, that income is partly funded by treasury assets and external yield sources, which raises questions about sustainability and regulatory classification. But from a pure value capture lens, Mantle sits on the “better than average” side of the L2 token spectrum.

Blast built its narrative on “native yield L2”: users deposit ETH and stablecoins and earn yield via L1 staking and off-chain strategies. The network rapidly attracted billions in TVL, heavily incentive-driven.

The economics are murkier. Much of the “yield” is redistributed from treasury and incentives; it’s not always clear how much is true net protocol revenue vs. dressed-up emissions. As with many incentive-heavy L2s, Blast risks being a high-beta yield farm rather than a durable economic engine unless it can pivot from incentives to organic activity.

Scroll is a ZK rollup with a strong focus on EVM compatibility and developer ergonomics. TVL and activity are mid-sized; the team is respected; the product is serious. But in a crowded ZK field and without a mature token and value capture story, it is still early to treat Scroll as a core investable asset rather than an option on future execution.

Linea is ConsenSys’ zkEVM rollup. Its core advantage is distribution: MetaMask is the default wallet for a large chunk of the ecosystem, and Linea gets front-row access to that flow. Throughput has grown, and TVL continues to climb.

At the same time, there is no public L2 token. Economics accrue to ConsenSys and to the applications built on top of Linea. As with Base, this is a reminder that the best economic leverage to L2 growth may not be a generic governance token but equity and revenue rights at the operator and infra layer.

Mode is an OP Stack L2 focusing on social/consumer apps; Metis is an earlier optimistic L2 with a more grassroots branding. Both have moderate TVL and activity, some traction in niches, and relatively weak economic fundamentals: low protocol revenue, high reliance on incentives, and no strong, differentiated value capture design.

These are the kinds of chains most exposed to consolidation. They can still work as trades during narrative cycles, but the long-term, fundamentals-driven investor must assume many such names will end up as “zombie chains” unless they find a sharp niche and repair their economics.

Immutable is effectively a gaming-centric L2 stack: Immutable zkEVM, marketplace infra, and deep integration with AAA studios like Ubisoft. Here, the business model is clearer: marketplace fees, infra revenue, and game-related economics.

IMX benefits from that structure by tying token value to fee burning and staking. Ubisoft partnerships have already translated into meaningful price moves around announcements. The long-term question is execution: if one or more flagship games onboard millions of players with real on-chain economies, Immutable and IMX could be one of the few L2-adjacent tokens with genuine, scalable cash-flow linkage.

Economically, L2s are pipelines that turn user demand into:

Against this, they pay:

For the leading L2s, this is already a real business. Base is the clearest example: sequencer margins, MEV, and transaction volumes have produced tens of millions in net profit on an annualized basis. Arbitrum and Optimism have meaningful gross revenue; while incentives and grants are still high, the underlying economic engine is there.

The long tail is different. Many L2s have negligible protocol revenue, high incentive budgets, and large operating costs relative to their scale. They live off investor and foundation subsidies. Their TVL and usage are highly reflexive to the presence of incentives. Turn off the faucet and activity collapses.

From an equity analyst’s point of view, the leading L2s are on a path toward real P&Ls; the median L2 is a VC-funded startup subsidizing its own usage. That’s fine in a growth phase, but it means not all governance tokens are created equal even before you ask who gets the cash flows.

And that’s the crux: even when the business is good, the token may not see a cent. Sequencer profits accrue to Coinbase, Offchain Labs, Optimism Collective, foundations and their treasuries. Tokens like ARB and OP give you votes, not dividends. Fee switches and revenue-sharing designs are politically and regulatorily sensitive; they might come, slowly, and in constrained forms — or they might never come at all.

Mantle and Immutable are early attempts at breaking this pattern: Mantle by explicitly sharing yield with stakers, Immutable by burning/redistributing marketplace fees. They are far from perfect, but they highlight what is missing elsewhere: a clear, contractual path from protocol-level economics to token-level value.

The underperformance of L2 tokens in 2025–2026 is not a mystery. It’s what you get when you combine:

Underneath all of this is one structural statement: the bridge between technological success and token economics has been left intentionally weak. As long as that is true, L2 tokens will trade more like high-beta macro instruments than like claims on predictable cash flows.

Summarizing across dimensions:

On the other side:

The upshot is that there is no single “L2 token trade.” There are specific cases where the tech/ecosystem/revenue story is strong but the token is weak; a few where token design has at least attempted to wire itself into the economics; and many where both the network and the token sit on shaky ground.

Institutional and big-brand partnerships are often used as the final justification for an L2 thesis: “Look, X is building on this chain, therefore the token is undervalued.” The reality is more complicated.

The common thread: institutional adoption is absolutely a real signal that L2s are becoming financial and consumer infrastructure. It is not automatically a reason to buy the corresponding token. You have to understand which layer — issuer, operator, Ethereum, L2 token, or application token — is actually capturing the economics.

Each major use case routes value differently.

In all of these segments, value capture is multi-layered: issuers/creators, infra operators, Ethereum, L2s, and applications each take a slice. The naive assumption that “activity on chain = value for the L2 token” breaks once you look at each link in this chain.

Ethereum L2s operate in a crowded field.

L2s win where Ethereum’s existing network, dev base and institution-friendly narrative matter, and where “Ethereum security + cheap execution” is the right bundle. They are less dominant where frictionless UX, unified liquidity or strict regulatory control is the main priority.

The net result is not a winner-take-all outcome but a segmented market:

For L2 token investors, that means you’re not betting on “Ethereum vs. Solana” in a vacuum; you’re betting on specific positions inside this stack, with their own competitive and regulatory constraints.

There are real catalysts on the table for L2 tokens:

Those are all legitimate ingredients for a powerful cyclical rally. If you combine them with currently depressed sentiment and underperformance, there is clear room for mean reversion.

The question is whether that reversion is structural or merely cyclical. If the underlying issues of value capture — weak rights to cash flows, governance-only tokens, slow or non-existent fee switches, persistent dilution — are not addressed, then most L2 tokens will remain trades: you buy them for narrative and liquidity cycles, not because they are durable claims on expanding cash flows.

That doesn’t mean there won’t be huge winners. It does mean that the burden of proof lies with token design, not just with tech and TVL.

A serious L2 thesis must own its risks in plain language:

These are not small footnotes. They are central to whether L2 tokens become high-quality, long-term holdings or semi-permanent high-beta instruments.

We can now answer the three key questions directly.

Ethereum Layer 2s are very likely the execution backbone of the network’s future. That does not obligate their governance tokens to be equally good investments. Until the gap between technology and token economics is deliberately closed — through fee switches, revenue sharing, or fundamentally new token models — the rational stance is to treat “L2 as infrastructure” and “L2 as token trade” as two different conversations, and to be far more selective in the second than the first.