On March 14, 2025, Monad's testnet logged 8,000 TPS for 12 consecutive hours. The crypto Twitter machine erupted. But celebration is cheap when the mainnet is still a ghost. Sei, meanwhile, already runs parallel EVM on mainnet with 1.8 billion in TVL. The market has crowned two kings. One has a crown of code. The other has a crown of users. Tracing the silent bleed from 2017's broken logic, I see a familiar pattern: hype before proof, valuation before delivery.
## Context: The Parallel EVM Thesis Parallel EVM is not new. It's a logical evolution of the sequential EVM bottleneck. Traditional Ethereum processes transactions one by one, like a single cashier at a supermarket. Parallel EVM opens multiple lanes. The promise: higher throughput without sacrificing EVM compatibility. The reality: it's a distributed systems problem wrapped in a smart contract layer. Two projects dominate the narrative: Monad and Sei. Monad bets on physical parallelism — execute everything simultaneously, then reconcile. Sei uses optimistic parallelism — execute first, verify later. Both claim 10,000+ TPS. Both have critical trade-offs.
Sei v2 launched parallel EVM on mainnet in early 2025. Its ecosystem boasts 150+ projects, 2.1 million cross-chain active addresses, and a TVL of $1.8 billion. Monad remains testnet-only, with a mainnet expected sometime in 2025 — no firm date. Yet Monad's OTC valuation sits at $30–50 billion. Sei's FDV is $130 billion with a $25 billion circulating market cap, giving an FDV/TVL ratio of 72x. These numbers scream 'narrative premium'. But narratives collapse when code fails.
## Core: The Technical Dissection Let's strip away the marketing. Monad's architecture: parallel execution engine + MonadDb (custom state storage) + MonadBFT (consensus). The execution engine uses optimistic concurrency control — it assumes no conflicts, executes all transactions in parallel, then checks for conflicts. If a conflict is found, it re-executes the conflicting transactions sequentially. This is similar to speculative execution in CPUs. MonadDb is a custom database optimized for Merkle tree state reads/writes, reducing I/O bottlenecks. MonadBFT is a pipelined consensus based on HotStuff, aiming for 1-second block times.
Sei v2's approach: optimistic parallelism with a twist. It uses a 'parallel EVM' that runs on top of its existing orderbook-focused chain. The key difference: Sei's parallelism is 'optimistic' in the sense that it executes all transactions in parallel without initial conflict detection, then validates results against a deterministic ordering. If a conflict arises, it re-executes. This is functionally similar to Monad, but Sei's implementation is tightly coupled with its native orderbook matching engine. The result: Sei excels at DeFi applications that resemble orderbook trading, but may struggle with general-purpose smart contract interactions that involve heavy shared state.
Based on my audit of 12 ICO contracts in 2017, I learned that optimistic execution without proper conflict detection leads to reentrancy-like failures. The same principle applies here. When two transactions modify the same storage slot, optimistic parallelism must revert one and re-execute. The more conflicts, the less parallelism. In high-congestion scenarios (e.g., a popular NFT mint), both Monad and Sei could degrade to near-sequential performance. The code never lies, only the auditors do. And in this case, the auditors haven't seen mainnet stress tests for Monad.
Monad claims theoretical 10,000 TPS. But theory is cheap. The real bottleneck is state access. Even with MonadDb, if every transaction reads/writes the same hot contract (e.g., a liquid staking pool), parallelism collapses. Sei's advantage is its orderbook-centric design — most orderbook trades hit different trading pairs, naturally reducing conflict. But for general EVM dApps, Sei's optimistic parallelism faces the same re-execution tax.
## Contrarian: What the Bulls Got Right Let's give credit where due. Sei's execution team has shipped. Mainnet parallel EVM is live. 150 projects are building. TVL is real. The narrative that 'Sei is a parallel EVM leader' is not entirely fabricated — they have first-mover advantage in deployment. Monad's team from Jump Crypto brings deep high-performance computing expertise. Their MonadDb design is genuinely innovative, addressing state I/O in ways that most L1s ignore. The bulls are right that parallel EVM is the next scalability frontier. They are right that both projects have strong teams.
But here's the blind spot: complexity is just laziness wearing a tech suit. Monad's architecture has three independent components (execution, storage, consensus), each with its own failure modes. Integration bugs are inevitable. Sei's architecture is simpler but less ambitious — it's an optimized orderbook chain with parallel EVM bolted on. Simplicity reduces risk but limits upside. The market prices Monad as if its complexity will be perfectly engineered. History says otherwise.
Another blind spot: valuation. Sei's FDV/TVL of 72x implies that every dollar of TVL supports $72 of token value. Even with high multiples in crypto, that's extreme. Monad's pre-launch valuation of $30–50 billion is pure speculation. If Monad launches with 1,000 TPS (10% of promise), the valuation will correct hard. If Sei fails to attract general-purpose dApps beyond its orderbook niche, its TVL growth stalls. Both face existential risks.
## Takeaway: The 18-Month Window Parallel EVM will have its 'mainnet season' in the next 12–18 months. Monad must ship. Sei must scale. The winner is not decided by technology alone, but by execution. If Monad delivers 10,000 TPS with low latency and no major bugs, it could dethrone Sei. If Sei continues to onboard users and dApps, it will maintain its lead. But the market's current pricing assumes both succeed. That's a dangerous bet.
Forensics reveal the truth markets try to bury: parallel EVM is not a silver bullet. The real throughput bottleneck is shared state, not execution order. Until a project solves state conflict at scale, 10,000 TPS is a dream. The code never lies, only the auditors do. And the auditors haven't signed off on Monad's mainnet yet. Watch for mainnet launches. Watch for stress test results. Watch for conflict rates. Until then, treat the parallel EVM race as a pre-revenue startup competition with inflated valuations.
Patterns emerge only when emotion is stripped away. Strip away the hype. What remains is a testnet and a live chain. One has users. One has promises. The silent bleed from 2017's broken logic continues.
