Hook
When former Fed governor Kevin Warsh warned last week that AI could “drive prices higher over the next 12 months” and even force the Fed to raise rates again, most crypto traders shrugged. After all, the market has been pricing in rate cuts for months. But as someone who spent the 2022 bear market watching protocols die from liquidity bleeding, I see a different signal: the same kind of demand-side shock that crushed DeFi liquidity pools in 2020 is about to hit the data layer of every blockchain. And if Warsh is right, the next “inflation” won’t be in consumer prices—it’ll be in the cost of validating AI agent transactions on-chain.

Context
Warsh’s argument is simple: AI’s capital expenditure boom—data centers, chips, energy—creates a short-term demand overload that outpaces any long-term efficiency gains. That’s textbook demand-pull inflation. Now apply that logic to blockchain. Every AI agent that trades, votes, or stores metadata on-chain consumes blockspace. As agent volume explodes (think: 10,000 autonomous traders per L2), the demand for data availability (DA) rises. Projects like Celestia, EigenDA, and Avail are positioning themselves as the solution. Yet, based on my audit experience during DeFi Summer, I’ve learned that 99% of rollups generate less than 100KB of data per day—enough to fit into a single Ethereum blob. The enthusiasm for dedicated DA layers is, in my view, overhyped.
Core: The Two-Sided Paradox of AI on Blockchain
Here’s the contrarian stake I want to plant: AI agents will increase on-chain data demand, but they won’t increase it enough to justify the current DA infrastructure arms race. Let me unpack both sides.
Side A – The Inflation Narrative (Warsh’s View, Applied to Crypto)
The parallels are eerie. Just as AI data centers drive up copper and electricity prices, AI agent transactions will drive up blob space and L2 gas fees. Already, top Ethereum rollups like Arbitrum and Optimism average 200-300KB of data per day. If AI agents multiply that by 10x, Ethereum’s blob space (currently ~2MB per slot) could fill up. That would push rollup costs higher, making DeFi less profitable—similar to how higher interest rates slow down lending. During the 2022 bear market, I saw protocols lose 40% of their liquidity providers when yields dropped below staking returns. A data cost spike would do the same.
Side B – The Reality Check (Why the Panic Is Premature)
But here’s the catch: most AI agent operations are simple state updates—balance changes, signature verifications—not fat metadata blobs. In my work on the “Trust” Protocol in 2017, we designed a governance system that fit 1,000 votes into 10 bytes. Modern rollups are even more efficient. Data compression and proof aggregation (zk-rollups) already reduce DA needs by 99%. The real bottleneck isn’t DA—it’s execution. Every L2 node has to run the transaction, and AI agents can generate thousands of parallel operations. That’s where innovation matters: parallel execution (like Solana’s Sealevel) or sharded executors.
Contrarian Angle: Warsh’s Inflation Might Actually Bust the DA Bubble
If the Fed does raise rates because of AI-driven inflation, the cost of capital rises. DA tokens (like TIA, EIGEN) become riskier speculative bets, and protocol treasuries may slash their spending on renting DA layers. During DeFi Summer, we saw governance delegation centralize precisely because users were too lazy to research—they just voted with the biggest KOL. The same apathy could kill DA adoption: developers will choose the cheapest, simplest option (Ethereum blobs) over experimental chains. The “DA layer” narrative is a governance problem, not a technology problem. Code is law, but people are the protocol.

Takeaway
Warsh’s warning is a gift to anyone who questions the current hype cycle. If AI truly pushes the Fed into a hawkish corner, the crypto market will face a liquidity squeeze. But in that squeeze, the protocols that survive won’t be the ones with the most exotic DA—they’ll be the ones with the most resilient human communities. We didn’t just build financial infrastructure; we built a social movement. That movement’s value isn’t in kilobytes per block—it’s in the trust between developers, users, and token holders. Let the AI inflation alarm ring. It might just clear out the noise.
