ABCDE: What changes have occurred in the public blockchain infrastructure sector amid the current AI boom?
TechFlow Selected TechFlow Selected
ABCDE: What changes have occurred in the public blockchain infrastructure sector amid the current AI boom?
"It is foreseeable that the AI bubble will peak in the next two years."
Navigating Web3 tides with focused insightsBy Laobai, ABCDE
In the current primary market, the hottest sector is undoubtedly AI, followed by BTC. About 80% of the projects discussed daily are concentrated in these two areas. At my peak, I’ve had conversations with five or six AI projects in a single day.
It’s foreseeable that the AI bubble will reach its peak in the next one or two years. As hundreds of new AI projects launch and the AI sector’s market cap climbs to new heights, eventually the bubble will burst, leaving behind chaos—but also giving rise to true unicorns that have successfully found the intersection of AI and crypto, pushing this sector and the entire industry forward.
Therefore, amid today’s overheated AI environment, it's worth taking a step back to examine what changes have occurred recently—especially over the past few months—in the infrastructure layer, particularly within the public blockchain infra space. Some of these developments are truly noteworthy.
1. The Further Deconstruction of ETH, or Monolithic Chains
When Celestia first introduced the concept of modular architecture and the idea of a Data Availability (DA) layer, the market took considerable time to digest and understand it. Today, this concept is deeply entrenched. Infrastructure for Rollup-as-a-Service (RaaS) has become so saturated that we've reached an absurd stage where the number of infrastructures exceeds the number of applications, which in turn exceeds the number of users.
Over the past few months, each layer—execution, DA, and settlement—has seen distinct technological advancements, spawning new technical approaches. Even the notion of a settlement layer is no longer exclusive to Ethereum. Let’s briefly go over representative technologies from each layer.
2. Execution Layer
The hottest concept in the execution layer is undoubtedly Parallel EVM—represented by Monad, Sei, and MegaETH. Existing projects like FTM and Canto are also planning upgrades toward this direction. However, just as not all ZK projects offer privacy protection, projects labeled as Parallel EVM differ significantly in both technical approach and ultimate goals.
Using a diagram from Sei for illustration, it’s clear that optimistically switching from sequential to parallel transaction processing can yield significant performance improvements.
Parallel EVM itself can be further divided into several distinct technical paths.
Different Approaches to Transaction Parallelization — Nothing New Under the Sun: A Priori vs. A Posteriori
A priori methods, represented by Solana and Sui, require transactions to explicitly declare which parts of the chain state they modify. This allows detection of state conflicts (e.g., access to the same AMM pool) before block packaging, discarding conflicting transactions if detected.
A posteriori, or optimistic parallelization, represented by Aptos BlockSTM, assumes no conflicts initially and includes transactions anyway. After execution, conflicts are detected. Conflicting transactions are marked invalid, results are rolled back, and the process repeats until all transactions in the block are successfully executed. Sei, Monad, MegaETH, and Canto adopt similar solutions.
In the primary market, we’ve also seen solutions aiming to parallelize under state conflict scenarios (such as concurrent access to the same AMM pool), but these appear relatively complex from an engineering standpoint, and their commercial viability remains uncertain and under evaluation.
Degree of Emphasis on Parallel EVM — Two Distinct Schools of Thought
-
One school, represented by Monad and Sei, treats transaction parallelization as the primary scaling strategy—parallelization being the core narrative. For example, Monad not only employs optimistic parallel execution but has also developed a dedicated MonadDB and asynchronous I/O system specifically optimized for parallel processing.
-
The other school, exemplified by Fantom, Solana, and MegaETH, views parallelization as just one of many scaling options—a supporting narrative. Performance gains here rely more heavily on other technical innovations.
For instance, Fantom’s Sonic upgrade focuses on the FVM virtual machine combined with an optimized Lachesis consensus mechanism. Solana’s next phase centers around the Firedancer client’s modular architecture, improved network communication, and signature verification, among other enhancements.
MegaETH aims to achieve a “real-time blockchain.” Building on Paradigm’s high-performance Reth client, it makes comprehensive optimizations across multiple dimensions: full-node state synchronization (syncing only state diffs rather than full data), sequencer hardware design (using large amounts of high-performance RAM with storage capabilities for state access, avoiding slow disk I/O), improvements to Merkle Trie data structures, and more. It integrates software, hardware, data structures, disk I/O, network communication, transaction ordering, and parallel processing to push EVM performance to its theoretical limits, approaching a “real-time blockchain.”
3. Data Availability (DA) Layer
There haven’t been major technical breakthroughs in the DA layer recently, making this segment far less competitive than the execution layer. The key players remain largely unchanged.
Ethereum’s Calldata has been upgraded to Blobs, drastically reducing fees for various L2s. Ethereum is now a “not-so-expensive” DA option.
Celestia’s biggest impact was establishing the DA category after launch. It raised the ceiling of the DA sector from a $200 million FDV to $2 billion, opening up vast new possibilities in terms of scale and imagination. Many new Layer2 appchains now naturally choose Celestia as their DA layer.
Avail has spun off from Polygon. Technically, it resembles an “enhanced Celestia,” adopting Polkadot’s Grandpa+BABE consensus mechanism, which theoretically supports greater decentralization with more nodes compared to Celestia’s Tendermint. It also supports validity proofs, which Celestia does not. However, technical differences matter less than ecosystem development—Avail still has catching up to do in that regard.
EigenDA launched alongside EigenLayer’s mainnet a few days ago. Given EigenLayer’s status as one of this cycle’s strongest narratives and its exceptional business development skills, I expect EigenDA’s adoption rate to be high. In practice, as long as a solution feels secure and is cheap, most projects don’t deeply care whether it uses validity proofs or fraud proofs, or whether DAS is supported.
More interestingly, here are three notable DA solutions:
-
Near DA — Near is a fascinating chain. Originally designed for sharding (which it still pursues), it has also built a DA layer—cheaper than Celestia and enabling fast L2 settlements. Chain abstraction — Near recently launched chain signatures, allowing users to request transaction signatures on any chain via a single NEAR account. AI — Its founder Illia is one of the eight co-authors of the Transformer paper and was famously patted on the shoulder by Jensen Huang at NVIDIA’s event. Near is now planning to hire AI engineers and will announce near.ai next month… A true hexagonal warrior, I’ve included it in the DA landscape.
-
BTC & CKB — Since Bitcoin’s base layer doesn’t support smart contracts and cannot directly settle, dozens of BTC-based EVM L2s are essentially using BTC as a DA layer. The difference lies in whether they post ZK proofs directly onto BTC or just the hash of the proof. It seems almost mandatory for a project to do this to call itself a “BTC Layer2.” Recently, I encountered a new project candidly saying, “I’m not pretending anymore—I’m actually an Ethereum L2, with DA and settlement on ETH, but I serve the BTC ecosystem!” Quite amusing… The only alternative scalability model here is CKB’s RGB++, where CKB acts as a DA-like layer while BTC, through UTXO homomorphic binding magic, effectively becomes the settlement layer for RGB++.
-
New DAs — Two novel DA concepts worth mentioning (project names withheld): One integrates DA with AI, serving not only as a high-performance DA layer but also as a storage layer for large AI models, training data, and training trajectories. The other improves the underlying erasure coding mechanisms of DAs like Celestia, offering greater robustness in dynamic networks where several nodes randomly drop offline each round.
4. Settlement Layer
Previously, this layer was almost entirely dominated by Ethereum. While DA has Celestia as competition and execution has numerous L2s, settlement remained uniquely Ethereum’s domain. Other chains like Solana and Aptos don’t yet have L2s, and BTC’s L2s neither use nor can use BTC for settlement. Until recently, Ethereum was practically the only viable settlement layer.
But this is about to change. Several new projects are moving in the direction mentioned at the beginning of this article, and some established ones are pivoting accordingly—namely, toward a ZK verification/settlement layer that further deconstructs Ethereum (and competes with it).
Why is this concept emerging?
Technically, running ZK proof verification contracts on Ethereum L1 is not an optimal choice.
To verify ZK proofs, developers must write Solidity-based verification contracts tailored to specific ZK projects and their chosen ZK proof systems. These often rely on complex cryptographic algorithms, such as support for different elliptic curves. The EVM-Solidity stack isn’t ideal for implementing such intricate cryptography, making development and verification costly for many ZK projects.
This has somewhat hindered native integration of certain ZK ecosystems into the EVM world. As a result, ZK-friendly languages like Cairo, Noir, Leo, and Lurk currently run only on their own Layer1s. Moreover, upgrading or modifying Ethereum is inherently slow due to its size and complexity.
From a cost perspective, although L2s pay most of their “protection fees” for DA, ZK contract verification still incurs gas costs. Verifying proofs on Ethereum is far from cheap. With ETH gas prices frequently spiking and turning the chain into a “noble chain,” verification costs become highly volatile.
Thus, new projects focused on ZK verification/settlement layers are emerging. Most are still early-stage, with Nebra as a representative example. Established projects are also pivoting—Mina, for instance, and Zen, which recently passed a new proposal.
The general approach across most projects in this space includes:
-
Support for multiple ZK programming languages;
-
Support for ZK aggregation proofs—more efficient and cheaper;
-
Faster finality times;
ZK settlement layers and decentralized proof markets are likely to converge. After all, advanced technology needs computational power. We may see collaborations between settlement layer projects and proof market platforms, or compute-rich settlement layers launching their own proof markets, or tech-savvy proof market operators building integrated settlement layers. The path forward will ultimately be determined by the market.
Other infra domains—such as oracles, OEV in the MEV space, and ZK light clients in interoperability—have already been covered extensively elsewhere, so I won’t repeat them here. When I come across more exciting innovations, I’ll share them with you!
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
ABCDE
