Most people believe that the explosion in AI demand will flatten the brutal boom-bust cycles of blockchain infrastructure. The logic is seductive: more AI agents, more machine-to-machine transactions, more demand for decentralized storage, compute, and bandwidth. Therefore, the next supply glut in Layer2 scaling or DePIN networks will be absorbed without a crash. This is a comforting narrative, and it is dangerously incomplete.
As someone who spent 2022 stress-testing Aave V2's liquidity models under a simulated 30% ETH price drop, I learned that the most dangerous assumption in crypto is that demand elasticity works exactly like traditional markets. The recent analysis from a prominent semiconductor strategist on memory cycles—specifically the claim that AI demand price elasticity of 1.42 will soften the 2028 supply hangover—exposes a similar blindspot in our own infrastructure cycle debate.
The ledger remembers what the bubble forgets.
Context: Why Crypto Infrastructure Faces a 2027-2028 Supply Shock
Before examining the elasticity argument, we need to map the current landscape. Ethereum Layer2 solutions now number over 40, with total value locked (TVL) exceeding $30 billion. DePIN networks like Filecoin, Arweave, and Akash have collectively deployed capacity worth over $10 billion in hardware. The narrative is familiar: these networks are scaling for the next wave of AI and decentralized applications. Venture capital is pouring in. Token prices reflect optimistic TPS and storage capacities.
But the data tells a different story. Active users across the top 10 Layer2s have plateaued at around 3 million daily. That's a fraction of Ethereum mainnet's peak. Filecoin's storage utilization is below 20%. The infrastructure is being built ahead of demand, driven by token incentives rather than organic usage. This creates a structural oversupply that will materialize when token incentives taper—likely in 2027-2028, as vesting schedules end and emissions drop.
This is not unlike the memory industry's 2028 scenario. Storage giants built HBM factories expecting AI demand to fill them. The blockchain analogue is the tens of billions of dollars locked in validator stakes, rollup sequencers, and DePIN physical infrastructure networks. When that capacity comes online without a proportional increase in paying users, the result is a race to the bottom on fees and provider margins.
Core Insight: The Price Elasticity Mirage
The semiconductor report argues that AI demand price elasticity of 1.42 means a 30% drop in HBM prices will drive a 42% increase in quantity demanded, offsetting revenue declines. Applied to crypto infrastructure, the argument would be: if transaction fees on Layer2s drop 30%, user activity will jump 42%, keeping total fees (and thus validator/node revenue) roughly stable.
This ignores three critical structural differences between semiconductor supply chains and blockchain networks.
First: The Multi-Layer Rent Extraction Problem.
In semiconductors, the price signal is relatively direct: DRAM cost decreases → NVIDIA pays less → AI developers get cheaper compute → more usage. In crypto, the value chain is fragmented. A decrease in Layer2 transaction fees might not translate into lower costs for end-users because applications set their own fees. On Arbitrum, for example, the base fee fell 60% in 2024, but average transaction costs on the Uniswap frontend remained around $2 due to interface and MEV surcharges. The elasticity is dampened by each layer taking a cut. Based on my audit experience with Golem's token distribution in 2017, I know that structural inefficiencies in value capture are persistent unless designed out from day one.
Second: Demand Elasticity is Not Uniform Across Use Cases.
AI agents autopaying for compute via smart contracts are indeed price-sensitive. But decentralized finance (DeFi) users—still the majority of on-chain activity—are not. Capital flows in DeFi are driven by yield differentials, not base fee costs. A 30% drop in Layer2 fees might attract a 5% increase in DeFi activity, not 42%. The marginal price reduction goes to arbitrage bots, not real economic activity. The assumption of high elasticity applies only to a narrow segment of crypto demand.
Third: Supply Does Not Respond Quickly Enough.
In the memory industry, supply expansion takes 18-24 months. In crypto, a new Layer2 can launch in weeks. The cost of adding capacity is near zero (just deploy a new contract). This means supply can overshoot far more aggressively. By 2028, we could see dozens more rollups, each promising lower fees. The glut will be worse than any semiconductor overbuild because the marginal cost of a new blockchain is minimal.
Contrarian Angle: The Decoupling Thesis is a Trap
The contrarian view that crypto infrastructure will decouple from traditional cycles—sustained by AI demand—is the same claim the semiconductor strategist made for memory. But there is a hidden assumption: that end-user demand elasticity translates directly to infrastructure provider revenue. In reality, between the user and the infrastructure lies a thicket of middlemen—protocols, oracles, relayers, sequencers—each with pricing power. The price reduction intended to stimulate demand gets absorbed by these intermediaries before it ever reaches the end user.
Consider the Filecoin ecosystem. The cost of retrieval has dropped 80% over three years, but usage growth has been anemic. Why? Because the bottleneck isn't storage price; it's the complexity of integrating decentralized storage into existing AI pipelines. The elasticity is low because the switching costs are high.
Furthermore, the "AI demand" narrative is self-validating. Venture funds invest in AI-crypto projects, boosting token prices and encouraging more capacity. This creates a phantom demand: capacity built for a future that may not arrive. When the VC money dries up (as it always does), the real utilization will be exposed.
Liquidity is not depth, it is just delayed panic.
Takeaway: Positioning for the 2028 Infrastructure Glut
Instead of assuming elasticity will save us, assume that 2027-2028 will bring a margin compression event in crypto infrastructure. The protocols that survive will be those with the lowest marginal cost of delivery and the stickiest demand—not those with the highest headline TVL. Chain-agnostic rollups that can aggregate demand across multiple ecosystems will fare better than siloed chains. DePIN networks with real paying customers (like Helium's IoT subscribers) will outperform speculative compute markets.
The real signal to watch is not user growth but revenue per unit of capacity. When that metric starts falling for two consecutive quarters, the cycle has turned. For now, stay skeptical of the elasticity narrative. The ledger always remembers.