BNB Chain on Wednesday unveiled plans for a purpose-built layer-1 blockchain designed specifically for AI agents that trade, pay, and operate onchain without human intervention, arguing that existing blockchain infrastructure was not built for the speed and privacy requirements autonomous systems demand.
The new network, detailed in BNB Chain's H2 2026 technical roadmap , will run alongside the existing BNB Smart Chain rather than replace it. A public testnet is planned for late 2026, with mainnet launch targeted for early 2027.
"Six months ago, BNB Chain set three priorities for BSC: speed, throughput, and protocol stability," the developers wrote. "This roadmap opens with the receipts and closes with what comes next — a second half focused on doubling performance again, and an architecture designed for the decade ahead."
The core problem
BNB Chain's argument is straightforward: AI agents may be ready to trade crypto, but the infrastructure they would trade on is not. Blockchains with public mempools — where pending transactions sit visible before confirmation — create latency and expose agents to front-running. Traditional financial markets move in microseconds; most blockchains confirm transactions in hundreds of milliseconds or longer.
The response is TxStream, a component of the new layer-1 that eliminates the public mempool entirely. Instead of broadcasting transactions to a shared pool, TxStream routes them directly to block leaders, cutting latency and removing the window where MEV (maximal extractable value) bots can observe and front-run agent activity.
The technical targets are ambitious. BNB Chain said the new network is designed to eventually handle more than 100,000 transactions per second — roughly 19 times current Bitcoin network throughput — with transaction confirmation targeted at under 50 milliseconds and block finality under one second. By comparison, BNB Smart Chain currently confirms blocks at 450 milliseconds, down from 750 milliseconds at the start of 2026.
Building on a foundation already in motion
The new layer-1 does not start from zero. BNB Chain's H1 2026 work already laid groundwork: block intervals were reduced from 750 milliseconds to 450 milliseconds; in-memory finality dropped from 1,125 milliseconds to 650 milliseconds; benchmark throughput nearly doubled from roughly 2,800 TPS to 5,200 TPS following the Osaka/Mendel hard fork.
Four technical upgrades drove those gains: Block-Level Access List (BAL), which pre-declares state access patterns to enable faster execution and future parallel processing; Incremental Snapshot for faster node synchronization; EVM SuperInstruction, which fuses common opcode sequences to reduce interpreter overhead; and Extended Voting Rules, which tightened the fast finality mechanism under adverse network conditions.
Quantum and the longer horizon
Beyond speed, the roadmap flags quantum-resistant security as a forward-looking priority. BNB Chain said it is researching post-quantum cryptography to protect against a future where sufficiently advanced quantum computers could decrypt data collected today. The work is at the research phase, but the approach uses account abstraction to allow users to adopt quantum-safe security without changing wallet addresses.
"There's no finish line here," the developers wrote. "Quantum computing will keep evolving, and so will our testing and research. The point is that when it matures, BNB Chain's infrastructure is already prepared."
The competitive context
BNB Chain is not alone in building for AI agents. The infrastructure layer for autonomous on-chain agents has become one of the more active construction zones in crypto.
In March, Tempo — backed by Stripe — launched a payments-focused layer-1 alongside the Machine Payments Protocol, an open standard for AI-to-AI and AI-to-service transactions. MoonPay that same month released the Open Wallet Standard, a framework developed with contributions from PayPal, Ethereum Foundation, Solana Foundation, and Ripple, aimed at letting AI agents manage funds and execute transactions across chains.
In May, Amazon Web Services partnered with Coinbase and Stripe to launch Amazon Bedrock AgentCore Payments, which allows AI agents to use USDC stablecoins to pay for APIs, data feeds, and online services. Coinbase followed in June with Coinbase for Agents, a product enabling AI agents to trade crypto, make payments, and manage portfolios within user-defined parameters.
BNB Chain's approach is differentiated by its focus on dedicated chain infrastructure versus payment rails or wallet standards — but it enters a field where major platforms are already building.
A structural bet on agentic AI
The announcement reflects a broader wager in the crypto infrastructure space: that AI agents will become a primary driver of on-chain transaction volume, and that the chains which can accommodate their requirements — speed, privacy, predictable fee structures — will capture that demand.
BNB Chain is also explicit about its commercial motive. The H2 roadmap lists three commitments beyond raw performance: advanced resource isolation so one application's traffic spikes do not degrade another's performance; refined gas fee structures to reduce entry costs for both Web2 and Web3 enterprises; and new token standards making it easier to issue and move stablecoins.
Resource isolation and precision gas pricing are, in part, an acknowledgment of a persistent DeFi problem: popular applications can saturate network capacity, slowing everything else. For AI agents running strategies across multiple protocols simultaneously, congestion on any single application could disrupt execution in ways that are costly or difficult to recover from.
The new layer-1 will face a testnet before it faces a verdict. Whether the architecture ships as specified, and whether developers building AI trading agents choose it over alternatives already in production, will determine whether the bet pays off.
