Hook
Warren Buffett admitted a mistake. Not investing in Google early. Now, he says Google is "more likely to win." A simple line from a billionaire? No. It’s a signal. A signal about how the market misjudges moats.
I’ve seen this pattern before. In 2020, while stress-testing Uniswap V2 flash loan vectors, I noticed the same blind spot: analysts fixate on speed or cost, ignoring underlying network effects. Buffett’s confession mirrors the crypto market’s collective error. We overestimate new chains and underestimate ecosystems built on composability.
Context
Buffett’s comment came during a Berkshire Hathaway annual meeting. He acknowledged missing Google’s rise because he underestimated its competitive moat. "I thought it was a tough business to judge," he said. Now, he sees Google’s dominance in search and advertising as nearly unassailable. He also stated that Apple remains a core holding, praising its consumer lock-in.
This isn’t about stocks. It’s about platform dynamics. Google’s moat is data-network effects: more users → better ads → more revenue → more AI investment → better services. Apple’s moat is ecosystem stickiness: hardware + software + services = high switching costs. Both are "super platforms."
In crypto, we have our own super platforms. Ethereum, Bitcoin, Solana. But the market often misprices their moats. New L1s offer 50,000 TPS for a fraction of the cost, yet they fail to capture meaningful value. Why? Because moats aren’t built on throughput alone.
Core (code-level analysis + trade-offs)
Let’s dissect Ethereum’s moat through a code-level lens. I’ve audited over 40 DeFi protocols. The recurring pattern is composability—the ability for contracts to interact trustlessly. Composability isn't just a feature; it's an ecosystem. Each contract becomes a building block. Uniswap pools, Aave lending markets, Maker vaults—they form a graph of interdependencies.
Consider a flash loan attack I simulated in 2020. I wrote a Python script that exploited a liquidity imbalance between Curve and Uniswap. The attack required simultaneous calls to multiple contracts. On any non-EVM chain, that composability would break. Solidity’s stack-based design and Ethereum’s state access make complex multi-contract interactions possible. This is not a trivial technical detail. It’s a moat.
Here’s the data: Ethereum’s TVL peaked at $120B in 2021. Even after multiple scaling wars, it still commands 55% of total DeFi TVL (source: DeFi Llama, May 2025). Meanwhile, high-throughput chains like Solana have 20% TVL but 80% of that is in liquid staking tokens or wrapped assets—not native DeFi. Why? Because composability is logarithmic. The more contracts, the more value they generate together.
We don't build for the current market; we build for the next cycle. I learned this during the 2022 bear market when I spent six months comparing STARK and PLONK proving systems. The result? A 50-page analysis showing that zero-knowledge rollups preserve composability while scaling. Starkware’s SHARP prover aggregates thousands of StarkNet blocks into one proof, but it fragments state across separate shards. That breaks atomic composability. Aztec’s approach with Noir + ZK circuits maintains a unified state, but at higher latency.
The trade-off is clear: composability vs. raw performance. The market often favors the latter (new L1s, sidechains) until a crash reveals the former’s true value. I saw this during the 2021 bull run—new chains promised 100k TPS, but during the Terra collapse, they couldn’t handle the volatility. Ethereum’s gas skyrocketed, but the composable layer remained functional. That’s a moat.
Contrarian (security blind spots)
Here’s the blind spot everyone misses: Bitcoin’s moat is not its security. It’s its institutional adoption. Post-ETF approval, Bitcoin has become Wall Street’s toy. Satoshi’s "peer-to-peer electronic cash" is dead. The moat is now regulatory and financial infrastructure. But that moat is fragile—it depends on centralized custodians and ETF providers.
Buffett’s mistake was underestimating Google’s data moat. Our mistake is overestimating Bitcoin’s security moat while ignoring its existential risk: the failure of mining decentralization. I’ve analyzed the Bitcoin mining distribution—three pools control over 50% of hashrate. That’s a single point of failure disguised as a moat.
Similarly, many believe Ethereum’s composability moat is unassailable. But consider the rise of modular blockchains. Celestia, EigenDA, and Danksharding separate execution from data availability. This re-architecture could fragment composability across rollups. If every application launches its own rollup, atomic composability dies. We saw this with Cosmos—IBC connects zones, but you can’t call a contract on another zone in the same transaction. That breaks composability for flash loans, arbitrage, and complex DeFi strategies.
The contrarian view: the market is overvaluing new L1s that offer speed but lack network effects. However, the market may also be undervaluing the risk of composability fragmentation in Ethereum’s own roadmap. If L2s become isolated silos, Ethereum’s moat erodes. We don't build for the current market; we build for the next cycle—but we must also build bridges.
Takeaway (vulnerability forecast)
The most likely outcome? Ethereum retains its dominant moat through unified state and composability. But the vulnerability is real: if L2s prioritize independent scalability over shared settlement, the ecosystem fractures. We’ve seen this with sidechains in 2021—they offered speed but centralization. The same pattern will repeat with rollups.
My forecast: by 2026, 30% of DeFi activity will be on L2s, but the remaining 70% will stay on L1 because composability demands it. The "more likely to win" platform is the one that preserves atomic composability—Ethereum mainnet, despite its gas costs. The contrarian bet is that a cross-rollup composability solution (like Arbitrum’s Orbit chains with shared sequencing) succeeds, creating a new super-platform.
Are we making the same mistake Buffett did? He saw Google’s moat too late. We might see Ethereum’s composability moat too late—or we might see its erosion too late. The question isn’t who has the fastest chain. It’s who has the most interconnected ecosystem.
