Written by Ryan Yi and Michael Atassi.

Disclosures and footnotes: Coinbase Ventures portfolio company backed projects are mentioned in the article including – Optimism, Arbitrum, Celestia, Eigenlayer, Stack, ThirdWeb, Syndicate, Conduit, Alchemy, Socket, Everclear, Reservoir, Starkware, Matter Labs.

L3s (or “Layer 3s”) are an emerging phenomenon among onchain developer deployments especially in the EVM L2 landscape. This document explains the basics of L3s, the value propositions, and implications for the broader ecosystem.

TLDR:

While L2s’ lower gas fees / increased throughput have enabled them to become the hub of onchain activity & account for a sizable portion of the ETH-based economy, they may be constrained by pressure to maintain decentralization & alignment with ETH L1.
Builders that want to [1] experiment and customize their apps, and [2] align with an L2 for distribution, are now opting to build L3s – which are application-chains that settle to an underlying L2.
This document is focused on building a shared understanding of L3s.

Takeaways

What are L3s?

If L2s are the onchain hubs, L3s can be thought of as “onchain servers” that have isolated state environments & fee markets but settle to an underlying L2 & tap into their onramp / distribution mechanisms. This provides apps their own customizable blockspace that they control, while still leveraging the existing liquidity & user bases of L2s.

Cost: Up to 1000x cheaper costs due to a combination of: (1) Lower onboarding costs (onboard directly from CEX to L2), (2) slightly cheaper settlement/execution costs (as transactions settle to an L2 vs. L1), & most importantly (3) alternative data availability (“DA”), which is how chains verify data accuracy (DA is 95%+ of total cost for L2s that use ETH L1 data). Gas fees are also more predictable as L3s have their own fee markets (e.g., on L2, surging activity on one app raises fees for all others)
Customizability: L3s are held to a lower decentralization standard than L2s, which unlocks ability to experiment with new tokenomics (i.e., custom gas tokens), virtual machines (i.e. Solana VM on ETH L2), & alt-DAs (i.e. Celestia instead of ETH L1)

How are L3s different from L2s?

L3s are rollups and therefore share many similarities to how people will understand L2s.

Settlement: Similar to how L2s settle to L1s, L3s settle to an L2
Bridge: Similar to how assets are bridged canonically (or 3rd-party) from L1 to L2, same from L2 to L3
Virtual Machine: The software stack that an L3 utilizes does not necessarily need to run on the same stack as its underlying L2. For example, many of the L3s in production run on Arbitrum Nitro, but settle to Base (which runs on OP Stack). In addition, most L3 stacks are modifications to existing popular L2 stacks. For example – Arbitrum (Nitro) and OP Stack have launched modified stacks tailored to needs of L3 builders
Data Availability: This is the largest difference. L3s will opt to use alternative DA layers (ex. Celestia, EigenDA, Arbitrum AnyTrust) while L2s have to use ETH L1 for alignment / decentralization purposes. As a result, L3s achieve extremely low cost gas environments.

How do you launch L3s?

As L3s are predominantly leveraging permissionless / open-source tech stacks, developers can either (1) run the stack / infra themselves, (2) leverage a Rollup-as-a-Service (“RaaS”) provider (i.e. Conduit, Caldera) that provide managed services to deploy & host your L3, or (3) consult with a white-label service provider (i.e. Syndicate) that “sub-contracts” out to various infrastructure providers (i.e. RaaS, Bridges, DevTools) under the hood.

Will there be L4s?

As L3s offer dedicated blockspace & the ability to natively bridge into L2 “hubs” with liquidity / users, we believe this will cover all material onchain use cases.
L3s are likely the “last” frontier for vertical scaling (i.e., no L4s), even as L2 transaction costs come down
- Integration to an L2 – the core premise for L3s is the ability to leverage the underlying L2 “hubs” with liquidity / users. Building at an “L4” is further in proximity and defeats the purpose.
- No cost improvement – the alternative DA is the reason for the cost delta. Moving up the stack will not meaningfully change the settlement / execution cost component.
If an L3 reaches scaling limits, rather than scaling further vertically (“L4”), they will likely spin up another L3 that settles to the same L2 (linked through native bridge). The end result is that L3s will likely scale horizontally rather than vertically

Ecosystem Implications

L3s will become another preferred direction for onchain builders potentially resulting in a handful of L2 “hubs” with millions of L3 “servers”

L3s represent a potential paradigm shift for onchain developers as they break the subcent barrier and reduce the threshold for building mainstream-scale onchain apps, leading to potentially the “app store” moment w/ millions of L3s.
L3s offer an experimental playground for builders and are perfect for high throughput / low cost applications – which can then tap into the underlying L2 hubs for liquidity / onramps / distribution
There is a potential outcome where there are tens~hundreds of L2 hubs, while there are millions of L3s.

L3s may be the potential “AWS” moment from a cost perspective

One observation is that L2s are becoming their own onchain hubs. The cost profile to operate an L2, due to its L1 proximity, is generally high and can range anywhere from 7~8 figures of USD per year.
L3s on the other hand operate a much lower cost profile, and annual cost of operating an L3 can range anywhere between $25~50K.

L3s developers will drive popularity of frameworks beyond Solidity / Vyper resulting in multi-VM environments

There are projects experimenting with bringing alternative frameworks (e.g. MoveVM, SolanaVM, Arbitrum Stylus) deployed on Ethereum. The goal is to expand the set of developer offerings, while leveraging the existing network effects, liquidity, and onramps of Ethereum.
This will likely manifest first at the L2 level – but we can expect these frameworks to be deployed as L3s that want to be purposeful on leveraging L2 hubs like Base.
The net result is that an L2 can attract the widest TAM of developers at the L3 level while maintaining its own chain on the EVM (vs trying to integrate multi-VM into the L2 directly).

L3s value flows will rely on the application layer

The KPI for individual L3s are users and transactions and token utility, rather than sequencer fees. The average value created by an individual L3 will likely be small – but as the number of L3s multiply this will create network effects.
Growth of L3s will generally benefit on value creation on the software side (ex. Developer Tools, Rollup As A Service) and on the protocol side (Data Availability, Chain Abstraction), but will only scale at large quantities of L3s.
We can expect a single issuer / project potentially launching multiple L3s, resulting in their own L3 ecosystem. For example – an onchain gaming ecosystem might have a single L3 for each game, and spin up additional L3 for another game – and the emergent ecosystem will provide residual value that is accrued and shared with other stakeholders.

L3s will require smoother interoperability and chain abstraction to succeed

If the intended purpose of the L3 is to tap into the L2 user’s UX, and we expect a growing number of L3s per application use-case, then interacting with these L3s needs to become seamless at the user level
Similar to L2s, L3 bridging can be achieved in two ways: natively, if the L3 settles to the L2, or through a 3rd-party provider. 3rd-party providers are better suited for L3s due to the experimental nature of the L3 stack, potentially resulting in a non-uniform and flexible bridging layer (see State of Bridging).
At the same time, L3s might only prioritize interoperability with their canonical L2 settlement chain rather than aiming for full interoperability with all other chains. As a result, they will focus on enhancing the feature set of the bridge, such as reducing latency and providing one-stop liquidity, to improve the overall user experience.
Orthogonally, there is ongoing protocol R&D around how to introduce native concepts at the sequencer level (see Based Rollups)

Future Outlook

In conclusion, the L2 ecosystem should expect to see the growth of L3 builders that will want to create isolated onchain application experiences while tapping into underlying L2 hubs.

Within the L3 landscape, Coinbase Ventures is investing in new use-cases that emerge. If you’re building in these areas, we would love to hear from you – Ryan Yi’s DMs are open!

This material is for informational purposes only, and is not (i) an offer, or solicitation of an offer, to invest in, or to buy or sell, any interests or shares, or to participate in any investment or trading strategy, (ii) intended to provide accounting, legal, or tax advice, or investment recommendations or (iii) an official statement of Coinbase. No representation or warranty is made, expressed or implied with respect to the accuracy of the information contained herein. Coinbase may have financial interests in, or relationships with, some of the entities and/or publications discussed or referenced in the materials. Coinbase does not endorse or approve links or third-party websites that may be provided in the materials.