Updated timeline expectations for rollups - polynya

In 2021, my expectation was for rollups to mature around late 2023 / early 2024. I was wrong, and I’ve acknowledged this many times before. I knew the crypto industry - and really, most emerging sectors - have a tendency for planning fallacy, and I was a lot less optimistic than most at the time (remember #L222?). But I was still too optimistic.

I think app-specific rollups, if they had continued development would have achieved it by now - they have been pretty close since 2021, but most development has shifted to smart contract rollups instead. There are exceptions, of course, Fuel V1 and DeGate have achieved Stage 2 decentralization today. Although not covered by L2Beat, which focuses on Ethereum L2s, Tezos’ smart rollups have achieved this today also.

As an aside, most teams focused on rearchitecting for performance rather than decentralization for roughly one year. My expectation was they would decentralize first, and then focus on performance, but rightfully the reverse has happened, and this is why my expectations were off by a year or so. Arbitrum worked on Nitro, Optimism basically started from scratch moving to Bedrock & Cannon, Starknet with Cairo 1, zkSync also rearchitected significantly before Era’s launch etc. The newer players like Polygon zkEVM and Scroll started at the right time, saving themselves the time lost to said rearchitecture.

Arbitrum is getting pretty close too. According to L2Beat.com, they are just three upgrades away from Stage 2 decentralization. I think Arbitrum will get there in 2024, and exit Beta. But most other L2s are farther behind - I think most of them will take till 2025 or 2026.

To be clear, this does not mean that you should not use it - just be aware that it’s a Beta product with risks involved (as illustrated well by L2Beat). Clearly, there’s been strong adoption, and Arbitrum One is consistently the #2 or #3 chain after Ethereum by DeFi volumes and multiple other economic activity metrics.

An important inflection point will be sometime in 2024, potentially earlier rather than later, when data availability becomes trivially cheap and abundant with EIP-4844, EigenDA, Tezos, Celestia, Avail, zkPorter, Solana with Firedancer etc. emerging. However, it’s important to note that data availability sampling will take years to mature, and I don’t expect these DA layers to hit their stride - and Ethereum to be upgraded to full danksharding - till 2025 or 2026. Likewise, zkEVMs will take years to be battle-tested to the point they can be considered reliable - I’m once again looking at 2026 or so. Maybe even 2028 to get to the point that Ethereum L1 zkEVM becomes a thing. Needless to say, the further out we’re looking, the more things can change, so it’s better to focus

So, my updated expectations are as such:

Some rollups like Arbitrum One and app-specific ones like zkSync Lite are already sufficiently usable today with few compromises. But the compromises do exist, as listed on L2Beat. Exceptions exist - you have Fuel V1, DeGate and Tezos’ smart rollups which are fully decentralized today.
I expect Arbitrum-based chains to exit Beta some time in 2024, but most other smart contract rollups will take till 2025 or 2026.
Even after Stage 2, some of the more complex components like zkEVM proof systems will take a bit more time to be sufficiently battle-tested.
Sequencer performance is an ongoing process that’ll keep getting better over time. Some of the latest techniques used by Sui, Solana or Aptos will be adopted by rollups. Proving performance for zk rollups is also up only and costs down only. Indeed, even complex zkEVMs are already cheap enough.
There’s a lot of work happening around multi-chain fractal scaling, but these will also take a couple of years to mature. Application developers for applications that require higher throughput than possible with even the fastest possible chain will get involved and contribute to building their own infra. Of course, for financial stuff like DeFi or NFTs, these are not so important, as a few chains will be able to satisfy all demand.
On the data availability side, it’ll also take several years for data availability sampling to mature, but as it’s a relatively far more abundant resource than execution even without sampling, it’ll no longer be the bottleneck much sooner - potentially early 2024.

The future is bright for a world where scalability is no longer a concern. Yes, I was wrong before about the timelines, and I could be wrong again, but the path is clear and we’re getting close - just need engineering and battle-testing to execute on the vision. Today, most L2s and L1s are barely utilized, and the actual bottleneck has been application demand for the last year or so. As the space matures, the gap between demand and supply will grow even wider. I’ve been calling for a focus on novel applications, UX and outreach for a long time, and I’m still not seeing enough, not much better than this time last year. The focus must shift from infra to applications, UX and outreach now, or most of the infra will remain unused. The hour is late, but there’s still time for the above to be somewhat ready in time for when infra is mature.