Problem
Web2 Dependencies
Most popular crypto projects still depend on a centralized cloud provider and off-chain middleware for order matching, data storage and cross-chain bridging.
The April 2025 AWS outage that froze withdrawals on Binance and KuCoin exposed this weakness, when one data centre failed, billions in user assets were instantly locked. Other Web3 services relying on AWS, including the Rabby crypto wallet and DeBank analytics, also went down during the incident.
These Web2 choke points also remain prime targets for hacks, exploits and MEV manipulation, undermining the core promise of decentralized finance and shaking user confidence every time a service goes dark. Decentralized solutions can't be trustworthy if they rely on centralized dependencies.
Low AI-compatibility
Modern blockchains lack native support for AI models to automate on-chain interactions. Architectural limits, rigid consensus models and limited validator capacity hinder competition with centralized solutions like AWS and Google Cloud, with their remarkable ~100,000 TPS and 25–50 GB/s throughput.
Monolithic blockchains are often limited because their data availability (DA), execution and consensus are developed in one layer. Ethereum with ~25 Mbps node bandwidth excels in decentralized, immutable storage but struggles with scalability and query flexibility, making it hard for AI-driven dApps.
But even blockchains with high throughput and low latency fall short without optimized data storage and management, both of which are critical for efficient AI execution. For example, Solana offers much lower latency than Ethereum with a potential ~1 Gbps throughput and distributed DA. Such infrastructure can be suitable for real-time AI triggers, but lacks native indexing and faces storage bloat.
L2s have poor synchronization and delays of data transfers, slowing real-time AI coordination and isolating dApps across separate L2s. On the other hand, Modular Solutions specializes in optimizing DA interoperability, but operate as a third-party and do not provide in-built execution and storage capacity.
Chromia, as a special case, enhances AI data handling with relational storage but is limited by isolated subchains, reducing cross-dApp synergy and the availability of diverse on-chain data for AI use cases.
Transaction Fees
Transaction fees in blockchains create a significant barrier for users, particularly newcomers, due to their unpredictability and volatility, hindering a seamless Web3 experience. Unlike centralized platforms, where costs are often fixed or subscription-based, blockchain fees fluctuate based on network demand and token prices, limiting high-frequency interactions critical for DeFi, DePIN and Web3 Gaming.
For instance, Ethereum typically charges $0.38–$1.50 for gas fees, with historical spikes reaching $70 per transaction, burdensome for retail users. But network activity and token price surges can significantly escalate costs even on cheap networks, like it was with TRON in April 2025.
The volatility and unpredictable nature of blockchain transaction fees discourages user experience and highlights a critical barrier to a scalable, accessible Web3 ecosystem.
Maximal Extractable Value (MEV)
MEV attacks undermine fairness in DeFi by allowing validators to manipulate transaction ordering through frontrunning, sandwiching or censorship for profit. This erodes trust in Web3, imposes significant losses on traders, and disproportionately harms newcomers unfamiliar with these risks, slowing the overall crypto adoption. For AI-driven on-chain actions, such as real-time DeFi trading or DePIN data processing, MEV disrupts the predictable, low-cost execution essential for automation.
A notable example is a 2025 sandwich attack on Uniswap. On March 12, a crypto trader lost ~98% of a $220,764 USDC transfer to a sandwich attack by an MEV bot on Uniswap v3. The bot front-run the transaction, drained the liquidity from USDC-USDT pool, and swapped the USDC to USDT in 8 sec, earning ~$215,500. The attacker tipped the Ethereum block builder $200,000, netting $8,000 profit.
The conventional transaction execution model makes MEV attacks nearly unavoidable, as miners or validators can exploit transaction ordering for profit.
Last updated