Tracing the gas trail back to the genesis block.
January 2023. A mid-tier L1 launches its mainnet with a TVL of $12M and a flurry of developer grants. Six months later, that same network is hosting exactly two DeFi protocols, both forks of Uniswap V2, neither with more than $200k in liquidity. The foundation’s blog posts still echo the same mantra: "We are building the electric grid of the future." The price of the native token has already halved twice.
The analogy is seductive. Every crypto conference speaker, every Medium thinkpiece, every allocator with a thesis on "infrastructure-first" rolls out the same narrative: "Thomas Edison didn't just invent the light bulb—he invented the entire electric grid. Networks capture value. Applications are commoditized widgets. Bet on the grid, not the bulb."
I’ve been staring at this analogy for six years now. It first surfaced in my 2018 deep dive into 0x Protocol v2, when I ignored the business logic to spend three months dissecting the Order Manager contract’s assembly code. I found seven critical edge cases in the signature verification process that the official audit had missed. My reward? A nod from the core team and a nagging suspicion that most people in this industry were reading the wrong signals. They were betting on the “grid” narrative without understanding the actual wiring.
Let’s be precise. The electric grid analogy works beautifully—until it doesn’t. It works to explain why L1s like Ethereum, Solana, or Bitcoin have maintained dominant valuations. It fails to explain why 95% of L1s launched since 2020 are trading below their ICO price, why the vast majority of L2s have fewer than ten applications, and why the most profitable entities in crypto today are not networks but applications: Uniswap, Aave, even the underlying infrastructure like MEV bots.
The grid analogy is a half-truth. And half-truths are worse than lies in engineering, because they pass a smell test without surviving a formal verification.
Context: The Genesis of a Dangerous Metaphor
The phrase “build the grid, not the bulb” entered crypto vernacular around 2017, during the ICO bubble. It was a reaction to the flood of application-specific tokens that had no sustainable value proposition. Projects like Basic Attention Token, Golem, and Filecoin promised to disrupt industries but failed to achieve product-market fit. The narrative inversion was inevitable: if applications fail, perhaps it’s because the infrastructure wasn’t ready. Bet on the base layer.
This thinking was crystallized by a handful of prominent investors and researchers. The argument goes like this: 1. The internet’s value accrued to the transport layer (TCP/IP, HTTP) and the platform layer (AWS, Google Cloud), not the applications built on top. 2. Blockchain networks, being decentralized by design, are a superior form of “trust infrastructure.” 3. Therefore, the network itself (the L1/L2) will capture the majority of value, while applications on top remain thin and interchangeable.
But this argument conflates two distinct things: network effects and market structure.
From my experience auditing Uniswap V2 forks in 2020, I learned that the protocol’s value wasn’t in the core DEX logic—anyone could fork it. The value was in the liquidity, the UX, the brand, the network of users. Uniswap was an application that became a network. Today, Uniswap Labs derives more revenue than most L1 ecosystems combined. The application-level protocols aren’t just “bulbs”—they’re self-assembling grids of their own.
Moreover, the electric grid analogy fails to account for the fundamental difference between physical infrastructure and digital infrastructure: multi-tenancy and composability.
In the physical world, an electric grid is a natural monopoly. One set of wires, one regulator, one pricing scheme. Competing grids would lead to inefficiency and safety hazards. In crypto, there are hundreds of L1s and L2s, each with its security model, throughput, and governance. The “grid” is a commodity, not a monopoly. The real scarcity is attention and liquidity, which accumulate not at the base layer but at the application layer.
Core: Deconstructing the Code and Economics of the Grid Analogy
Let’s disassemble the analogy into its core assumptions and test each against real blockchain data.
Assumption 1: Networks Are Harder to Replicate Than Applications
False. Forking a blockchain is easier than forking a successful application. In 20 minutes, I can spin up an EVM-compatible L1 using a few command lines. Docker, genesis.json, done. But forking Uniswap’s liquidity? That requires billions of dollars of capital, a team of developers, and years of community building. The code is easy; the network effect is hard.
In 2022, when I was writing the internal memo on Arbitrum’s fraud proofs, I spent weeks modeling the bond size required to deter a sophisticated attacker. The model showed that the network’s security is inversely proportional to its adoption: if too few validators participate, the bond is insufficient. The code is open-source. The hard part is bootstrapping the decentralized validator set.
The L2 scalability paradox I analyzed in that memo taught me one thing: the network’s value is determined not by its code quality but by participant alignment. Any L2 can copy the OP Stack or ZK Stack. But only a few will achieve the liquidity and developer trust needed to survive. The grid is the application.
Assumption 2: The Base Layer Captures More Value Than Applications
Historical data says otherwise. Let’s look at Ethereum, the most valuable “grid” in crypto.
- Ethereum’s market cap: ~$300B (as of mid-2025).
- Combined market cap of top 10 Ethereum-based applications (UNI, AAVE, MKR, etc.): ~$50B.
- Ratio: 6:1.
But look at the revenue distribution. Ethereum’s L1 fee revenue in 2024 was roughly $2B. Uniswap alone generated $1.5B in fee revenue. The MEV ecosystem on Ethereum extracted over $1.5B in 2023. The application layer is not just a “bulb”; it’s a generator. The grid distributes value, but the application captures it.
During my EigenLayer restaking analysis in 2024, I modeled the economic security thresholds for active restaking. The simulation scripts I published showed that slashing conditions were too loose relative to the economic stake required. A coordinated attack could drain the restaking pool if the attacker controlled 15% of the economic weight. The network itself (EigenLayer) had a gap in its security model, but the applications using it (like liquid staking derivatives) were accruing the real yield. The base layer was a mediator, not a value sink.
Assumption 3: The Grid Analogy Applies to L2s
This is where the argument gets most dangerous. The 2024-2025 market is flooded with L2 rollups, each claiming to be the “grid” for a specific application domain. But the market is already showing signs of commoditization.
- There are over 40 active L2s on Ethereum alone, with more launching every month.
- Most have less than $10M in TVL and fewer than five active developers.
- The design space is converging: OP Stack, ZK Stack, and Polygon CDK dominate. The differentiation is minimal.
The real difference between OP Stack and ZK Stack isn’t technical maturity or security assumptions. I’ve spent enough time reverse-engineering both to know that the core difference is which ecosystem can convince more projects to deploy chains using their stack first. It’s a marketing war, not an engineering race.
My experience building the AI-agent smart contract interface in 2025 reinforced this. I built a prototype where an LLM could autonomously execute DeFi trades via a secure oracle. The bottleneck wasn’t the network’s throughput or finality—it was the cryptographic signing overhead required to prove agent actions on-chain. The L2’s latency was irrelevant compared to the ZK proof generation time. The application’s requirement defined the infrastructure, not the other way around.
Contrarian: The Blind Spots of the Grid Narrative
Blind Spot #1: The Fallacy of Unbundling
The grid analogy implies that the network and the application are separable. In crypto, they are intimately coupled. The application’s security depends on the network’s validators. The network’s value depends on the application’s users. The bundling is dynamic and nonlinear.
Consider the case of Tornado Cash. It was a smart contract (an application) that relied on Ethereum’s privacy. When OFAC sanctioned the application, the entire Ethereum network became tainted. The grid couldn’t isolate the bulb. The US government effectively attacked the grid through the application. This coupling is a feature of composability, but it also means that the application layer can introduce systemic risk to the base layer.
Blind Spot #2: The Regulatory Asymmetry
When you build an electric grid, you get regulated like a utility. In crypto, the networks themselves are increasingly being treated as securities. The SEC’s stance on ETH after the Merge, the CFTC’s classification of ETH as a commodity, the fragmentation of global regulation—all of these create legal uncertainty that the grid narrative conveniently ignores.
If a network is a utility, it should be regulated as one. But crypto networks are not natural monopolies. They are permissionless, global, and often pseudonymous. The “grid” that the analogy refers to doesn’t exist. What exists is a network of voluntary participants, and any of them can exit at any time. The value of a network is not the infrastructure; it’s the social contract of continued participation.
My 2022 deep dive into the L2 scalability paradox revealed that the bond size for fraud proofs was mathematically insufficient not because of poor modeling but because the assumption of rationality was flawed. Attackers don’t just maximize profit; they can also act out of ideology or malice. The network’s security depends on the assumption that rational actors will behave rationally. The electric grid doesn’t have that problem. Copper wires don’t have incentive misalignment.
Blind Spot #3: The Commoditization of Infrastructure
As infrastructure becomes cheaper to deploy, its value as a differentiator declines. Celestia makes data availability a commodity. EigenLayer turns Ethereum security into a commodity. The modular thesis is accelerating this trend: the base layer becomes a set of commodities (security, data availability, execution), and the application layer integrates them.
In this world, the real innovation is not the network but the interface between networks. The smart contract that orchestrates across L2s, the wallet that abstracts away the chain, the AI agent that optimizes routing—these are the new “grids.” The analogy has inverted.
Takeaway: The Next Cycle Won’t Be About Networks
The grid narrative served a purpose during the bear market of 2022-2023. It encouraged long-term thinking and redirected capital from scammy application tokens to sustainable infrastructure. But the market has moved on. We are now entering a phase where applications are the new networks.
Uniswap X, Aave’s cross-chain expansion, the rise of intent-based architectures, the explosion of AI-agent smart contracts—all of these blur the line between application and infrastructure. The next cycle will reward projects that can aggregate, compose, and automate across networks, not those that simply host a few DApps.
Entropy increases, but the invariant holds: value accrues to the scarcest resource. In the current market, the scarcest resource is not block space; it’s user attention and developer talent. Applications that capture attention become the new grids. Networks that fail to attract applications become ghost towns.
Based on my experience auditing the EigenLayer restaking architecture, I can state this with confidence: the economic security of a network is only as strong as the weakest application using it. And the weakest application is often the one with the highest TVL and the smallest developer team. The grid narrative encourages complacency. It makes investors believe that holding the native token is a strategic position, when in reality it’s a bet on the collective actions of thousands of independent developers who have no obligation to the network.
Smart contracts don’t lie, but analogies do. The electric grid analogy is a beautiful simplification that fails at the first stress test. I’ve seen too many projects die because they bought into the infrastructure-first thesis without building a compelling use case. The bulbs matter. The bulbs are the grid.
Next time someone tells you to "build the grid, not the bulb," ask them which grid they’re building, and why the previous 200 grids failed. If they can’t give you a quantitative answer, they’re selling a story, not a solution.