Ethereum Researcher: Why Monolithic L1 Blockchains Are a "Dead End"?
TechFlow Selected TechFlow Selected
Ethereum Researcher: Why Monolithic L1 Blockchains Are a "Dead End"?
Monolithic L1 blockchain networks are not only unsustainable, but they will never actually deliver the required throughput—falling short by multiple orders of magnitude.
Navigating Web3 tides with focused insightsAuthor: Polynya
Editor: Nanfeng
I’ve never written an article about “Optimistic Rollups vs. ZK-Rollups.” They’re both great. My articles are all about opposing monolithic L1 blockchains that are extremely inefficient, unscalable, insecure, and brittle — we’ve already seen the absurd capital allocation toward these L1s.
My sole purpose is to see blockchains scale globally and ubiquitously in a technically, economically, and socially sustainable way.
Monolithic L1 blockchains are not only unsustainable, but they will actually never deliver the required throughput—falling short by multiple orders of magnitude. But this is no longer the issue — we need modular execution layers & data availability sampling to scale blockchains. This has never been clearer. The sooner we recognize this and shift accordingly, the better off the blockchain industry will be.
Yes, this is just my opinion. As always, please treat this piece as a stream-of-consciousness ramble from an amateur rather than a professional research paper.
Monolithic L1 Blockchains (Still) Extremely Inefficient
I've been writing about Rollups since 2020 and published blog posts around this time last year when the mainstream narrative was “Cardano/other L1 smart contract platforms are coming soon, and users will migrate en masse from Ethereum to other L1s overnight.”
Then came the narrative that “L2s are temporary stopgaps; the only way to scale is through L1s.”
Yet even the mildest testing at the end of 2021 caused these alternative L1 narratives to collapse at an astonishing rate:
1. Binance Smart Chain (BSC) increased its system requirements due to state bloat. As a result, nodes began desynchronizing, causing numerous issues. The problem here is that they initially specified very low system requirements — but failed to inform people that requirements would grow exponentially with state expansion. Now, reckless network changes are being made without adequate audits or testing. Of course, switching to Erigon and operating multiple chains offers incremental benefits, but serious limitations will always remain.
2. Solana avoided such mistakes by being upfront from the start about high system requirements. Now, I won’t criticize Solana’s various failures and problems — largely because it was an early beta product. Bugs and issues stemming from missing features are inevitable in any beta project, whether it’s Rollups, dApps, or monolithic L1s; I simply hope developers fix them. But over time, as Solana matures, undergoes rigorous testing to eliminate bugs, and implements a fee market, it will require delivering incremental throughput at enormous technical and economic costs. Solana is fundamentally unscalable. Optimistic Rollups are a far superior solution and will mature quickly.
3. Polygon PoS is arguably the most adopted chain after Ethereum. Now, Polygon PoS is indeed a "commit chain" rather than a true alternative L1, but it remains a monolithic chain subject to the same inefficiencies as L1s. Polygon PoS has hit its limits, suffered from spam transactions, and raised its minimum gas price. Even then, many projects continued flooding it with spam, pushing gas prices above $0.10. To be clear, this is a much better outcome than on Solana or Cardano, where during congestion 99% of transactions fail, leaving only tiny MEV bots successful. Credit to the Polygon team for openly acknowledging these limitations — unlike others building monolithic blockchains — and focusing all efforts on ZK-Rollups, which will truly enable high scalability. Actions speak louder than words; the Polygon team’s $1 billion investment in zero-knowledge technology deserves praise.
4. Regarding Cardano, it too is an early-stage beta product like Solana and must implement a fee market. Cardano’s system requirements remain low. Recently, I’ve noticed growing interest in Rollups within the Cardano community — great to see! However, until Cardano itself implements data availability sampling, all of this is futile.
5. We’ve seen many other projects fail to live up to the hype. Whenever block space saturates, Avalanche C-Chain fees spike — well, that’s a fundamental characteristic of monolithic blockchains. Subnets lead to fragmented security or decentralization and face equally severe constraints (i.e., fee spikes). As for Avalanche’s “online pruning” (referring to node compression of past transaction history): let’s wait and see, but it appears to implement Geth’s “offline pruning” at higher frequency. It might be a good addition to Geth, but absolutely cannot solve the fundamental limitation of state bloat. We’ve already seen Harmony fail, among others. But I also want to highlight projects building next-generation scalability solutions: data availability sampling from Ethereum, Tezos, Celestia, and Polygon Avail; validity proofs from Mina and Aleo; and numerous Rollup designs — new ones seem to appear every week! Clearly, we’ve entered the era of “modular architecture” — few take seriously building new monolithic L1s anymore. At best, projects like Polkadot and NEAR (“proto-modular”) serve as mid-term solutions, maintaining sustainability and security despite not solving many of the above issues. If you don’t care about sustainability and security, Dfinity/ICP is building interesting things.
6. I should clarify that monolithic L1s do have a path forward: upgrading to next-generation tech. For example, Avalanche could implement a “data availability sampling subnet” at the base layer, validated by the full validator set, and introduce modular execution layers. Unless this becomes a priority on their roadmap, I’ll continue advocating this view until such transitions become widespread across the industry.
7. Not every chain needs to be a Rollup or have some form of modular design. Sovereign L1s still have a place when security isn’t critical or when pursuing novel features hard or impossible to achieve in Rollups. Of course, this is largely a niche market, but it’s real. The Cosmos ecosystem excels here, and Polygon Edge is building compelling solutions — though I’d like to see these chains proven secure and IBC evolve to verify validity proofs. The same applies to cross-L1 bridges. Even perfectly proven-valid L1 bridges, like the “Mina <> Ethereum bridge” being built by =nil; Foundation, still require trusting that the weaker chain (i.e., Mina) won’t be attacked; in contrast, Rollup bridges offer full security guarantees — even if the weaker Rollup chain is compromised, you still inherit the security of the stronger chain.
8. Finally, timing matters. Currently, Optimistic Rollups aren’t ready yet, so using non-beta monolithic L1s still makes sense. Note that both Optimistic Rollups and monolithic L1s vary in maturity/stability, so assessments must be context-specific. I won’t elaborate on that here — my sole interest is how blockchains can scale massively and sustainably in the long run. But Optimistic Rollups are maturing rapidly and have a clear path to becoming sustainable solutions. It’s merely an engineering effort — likely at least one Optimistic Rollup will achieve full decentralization, data compression, highly liquid bridges, and scale within a year. Once Optimistic Rollups are ready for prime time, the game is over for all monolithic L1 networks.
In the past, I’ve taken an extremely long-term view, so I may have discussed ZK-Rollups more favorably than warranted.
I’ve always believed ZK-Rollups will be the endgame, and that all Optimistic Rollups will either transition into ZK-Rollups or adopt validity proofs to replace their fraud-proof systems. But that might be three years away.
The conclusion from Part One above is this: monolithic L1 blockchains are a dead end — the evidence is overwhelming, so I wanted to make that crystal clear. Below is Part Two, focused primarily on Ethereum, since Ethereum is where all Rollup innovation happens and has the most ambitious scalability roadmap.
Optimistic Rollups vs. ZK-Rollups
Now, let’s discuss Optimistic Rollups and ZK-Rollups.
By the way, recently I learned about Obscuro (a TEE Rollup) and Urbit’s Naive Rollup.
The design space for Rollups is a blank canvas — there can be many types of Rollups! But here, I’m specifically discussing secure Rollups based on fraud proofs (Optimistic Rollups) and validity proofs (ZK-Rollups).
Transaction Fees
Let’s start with application-specific Rollups: clearly, ZK-Rollups lead here. Solutions like Loopring and zkSync are welcomed in payments, with ERC-20 token transfers costing around $0.10. At the time of writing, ZigZag, a decentralized exchange built on ZK-Rollup tech, charges a flat fee of $0.28, including gas and trading fees. Meanwhile, dYdX, the decentralized derivatives platform, has no user-facing gas fees, but we can calculate its transaction cost: during high volume, they pay ~$0.08 per trade to Ethereum; on quieter days, it’s around $0.10.
Now, Optimistic Rollups could very well reach this fee range — in fact, Hubble (an Optimistic Rollup L2 used by Worldcoin) currently has ERC-20 transfer fees below $0.10. Let’s move to general-purpose smart contract Rollups — why are fees so high on current Optimism and Arbitrum? At the time of writing, according to Lefees.info, swapping tokens costs $0.85 on Optimism and $1.35 on Arbitrum — that’s really high!
The answer is simple: these Optimistic Rollups aren’t optimized yet, and most early dApps deployed on them are forks of existing Ethereum projects, not designed natively for Optimistic Rollups. Recently, Aave V3 — optimized specifically for Optimistic Rollups — successfully reduced fees by about 10x!
But further optimizations are coming. Signature aggregation/compression will save ~1,000 gas per transaction directly. Basic calldata compression saves ~2.5x, and more advanced schemes will emerge. As Optimistic Rollups and their dApps mature, transaction fees will easily reach the $0.01–$0.10 range.
Wait — this is before The Surge phase of Ethereum’s roadmap (which drastically increases Rollup scalability via sharding)! If the Ethereum Shanghai hard fork in late 2022 (the first upgrade post-merge) goes smoothly, the first step may be introducing “blob-carrying transactions” (a new transaction format proposed by Vitalik), reducing calldata costs to nearly zero — making Optimistic Rollups even cheaper!
Today, ZK-Rollups have high fixed costs. Currently, ZK-Rollups are cheaper because their calldata savings outweigh these fixed costs; but as calldata costs become negligible, transaction fees will be dominated entirely by ZK-Rollups’ fixed costs — which are significantly higher.
Clearly, there are now many different proof systems with varying costs, but typical estimates are $0.01–$0.02. That will be the floor cost for ZK-Rollups. However, Optimistic Rollups can go lower freely — since a mature Optimistic Rollup’s cost is 99% calldata, they can access ~5200 TPS of blobspace at trivial cost.
By the way, Optimistic Rollups will also be cheaper than Validiums (which store data availability off-chain), because this minimum cost applies to Validiums too! I believe zkPorter (a Validium) transaction costs are between $0.01–$0.03, while Optimistic Rollups can go well below that. Personally, I don’t recommend such schemes due to trade-offs in transaction quality, but in this case, Optimistic Rollups would ultimately earn higher “profits,” which could be redistributed to stakeholders or used for public goods funding, development, etc.
Of course, eventually ~5200 TPS will saturate, and if demand exceeds that, calldata costs will rise again. But by then, Danksharding will launch, first expanding this space to 125,000 TPS, then millions of TPS over several years. (By the way, “TPS” numbers are meaningless when discussing Rollups, but useful for illustrative purposes.) Long-term, calldata cost will definitely not be the bottleneck — the bottleneck will be the Rollups themselves. Again, the bottleneck will be the Rollups, not Ethereum.
Long-term, as ZK-Rollups mature, we’ll have ASIC provers, and ZK-Rollup costs will become negligible. Then, ZK-Rollups’ advantages — most notably instant withdrawals (no waiting) — will dominate. For use cases with highly compressed state deltas (like dYdX), they’ll soon be much cheaper than Optimistic Rollups.
Finality
On this point, a common misconception is that Optimistic Rollups have a 7-day lock-in period. In reality, Optimistic Rollups will achieve L1-equivalent finality faster than ZK-Rollups. We’ve already seen Optimistic Rollups submit transaction batches every 5–10 minutes — that’s the latency. The 7-day window ensures finality holds, assuming at least one honest actor maintains it. As Optimistic Rollups scale, batches will be submitted more frequently; once reaching ~20 TPS, they can submit per block, making finality equal to L1’s. Due to ZK-Rollups’ much higher fixed costs, submitting per block requires more activity (>100 TPS) to be economical. However, with Vitalik’s proposed blob transaction EIP, ZK-Rollups could redesign to submit proofs to blobs — potentially resolving this.
So what is the 7-day period for Optimistic Rollups? Since the challenge period is 7 days, one type of transaction does have a 7-day lock-in: withdrawals from Optimistic Rollups back to L1. For certain use cases, like cross-chain NFTs, this could be challenging. However, we’ve already seen many bridges launched on Optimistic Rollups, and as they mature, activity and liquidity will increase, making the 7-day challenge period less problematic. The bigger point is that ZK-Rollups enable many new use cases and cross-Rollup activities impossible on Optimistic Rollups.
For instance, with Danksharding, ZK-Rollups could make synchronous calls to L1. I even speculate that if we implement L1 pre-confirmations via crLists, some degree of composability between different ZK-Rollups might become possible! Optimistic Rollups are excluded from these novel and innovative scenarios. Appendix: While Optimistic Rollups can support private transactions, ZK-Rollups will handle them at much lower cost.
Throughput
Application-specific Rollups are lean and can scale massively. If I recall correctly, StarkEx demonstrated 9,000–18,000 TPS as early as mid-2020. However, general-purpose smart contract Rollups face greater challenges. We’ve seen StarkNet achieve lower throughput than dYdX or Immutable X, making throughput optimization a top priority. Since Optimism and Arbitrum are based on EVM clients, they a) have battle-tested codebases, b) relatively optimized clients. A few days ago, I asked how far EVM could scale before needing parallelism — Alexey Sharp suggested Erigon could scale to 500M gas/sec. Thus, there’s significant room for optimized EVM-based Optimistic Rollups, with even more headroom via multithreaded clients or multiple instances/recursive Rollups.
This is our key contention: because Optimistic Rollups rely on an honest minority assumption, you can’t push system requirements too tightly. To be clear, this honest minority assumption makes Optimistic Rollups safer than any monolithic L1 (which rely on honest majority assumptions). But you still need guardrails — inherently lower than ZK-Rollups, which can easily verify via validity proofs.
But Optimistic Rollups have a solution — stateless clients. I saw this on Optimism’s roadmap, though I forget where. With stateless clients, verifying Optimistic Rollups becomes easy, and verifying transactions becomes trivial with the right tools. Moreover, since Rollup state can be easily reconstructed directly from L1, Rollups can aggressively implement state expiry beyond L1s. Once statelessness and frequent state expiry are implemented, Rollups will have already overcome the severe problem of state bloat! Additionally, new solutions like Fuel V2 enable parallelization via UTXO-like systems.
That said, via recursive validity proofs, ZK-Rollups remain the more elegant solution for maximizing network throughput. Likely, as StarkNet matures, we’ll see application-specific ZK-Rollups built atop it achieving massive throughput, with StarkNet as a whole scaling far beyond any Optimistic Rollup. However, the trade-off is atomic composability — unless ZK-Rollup teams figure this out quickly!
Summary
Ultimately, it all comes down to timelines.
For application-specific chains, ZK-Rollups are already the optimal solution in most cases. This is obvious to anyone who’s used dYdX, zkSync, Loopring, or Immutable X. There are trade-offs, but all will be resolved by year-end (e.g., dYdX V4).
Smart Contracts: Monolithic L1s have about one more year of relevance. Their doom is sealed — unless they urgently pivot to fraud proofs, validity proofs, and DA (data availability) proofs, there’s no escape.
Optimistic Rollups will scale faster than ZK-Rollups. Much of the Optimistic Rollup codebase is already mature; once sequencing and upgradability are decentralized, they’ll be ready for mass adoption. By end of 2022, Optimistic Rollups may see sub-cent gas fees, achieve full decentralization, offer security and finality equivalent to Ethereum L1, and have sufficient liquidity for fast withdrawals. And with statelessness and state expiry implemented in 2023, Optimistic Rollups will keep improving.
ZK-Rollups will continue evolving, maturing, and undergoing rigorous testing — optimizing proof times, shifting to GPU provers, and eventually ASIC provers. Their novel VMs and sequencer nodes will also mature and scale over time. By end of 2023, I expect ZK-Rollups to catch up and surpass Optimistic Rollups. But Optimistic Rollups will remain relevant until 2024/25, by which time I expect most will either become ZK-Rollups or at least replace their fraud-proof systems with validity proofs.
I’ve said many times my timeframe is 5–10 years. But the progress in Rollups and DAS (data availability sampling) is so rapid that I now believe the end is in sight — and will happen within five years.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
