I remember sitting in a Sydney café last year, across from a young founder who had just closed $15M in venture debt for his Layer-2 sequencer project. He was beaming. ‘We’re building the next AWS for rollups,’ he said. I asked him what happens if the revenue doesn’t hit projections in 18 months. He laughed. ‘That’s what the next round is for.’ I didn’t laugh. Because I’d seen this movie before—in AI, where a debt-backed infrastructure binge is now flashing warning signs that the crypto industry should study carefully.
Context: Over the past five years, AI data center builders have doubled their collective debt to finance a historic spending spree. The narrative is seductive: more compute means more powerful models, which means more revenue. But as any CFO knows, debt is a bet on a specific timeline. If the revenue doesn’t materialize before the interest payments come due, the entire house of cards collapses. The AI market is already seeing this tension—capital expenditure has outpaced software revenue by a factor of three, and the debt-to-EBITDA ratios of key infrastructure players are climbing fast. Sound familiar?
In crypto, we’re heading down the same path. Layer-2 sequencers, validator-as-a-service providers, and cross-chain bridges are increasingly financed through high-interest debt instruments, not token sales or organic treasury growth. Why? Because the bull market makes debt cheap, and the race for market share incentivizes speed over sanity. I’ve audited the cap tables of five rollup teams in the past year; three of them have debt obligations that exceed their annual revenue projections by 2x. The logic is that future token emissions or protocol fees will cover it. But that logic assumes a continuous bull run—and assumes the debt holders won’t call the loans during a downturn.
We didn’t build crypto to replicate Wall Street’s leverage games. Yet here we are, watching infrastructure builders treat their nodes like real estate assets, securitizing them into debt products sold to yield-seeking VCs. The technical risk is even more acute: if a sequencer team defaults, the sequencer—a single centralized node in practice—could halt the entire Layer-2. Truth in blockchain isn’t found in whitepapers; it’s found in the balance sheet behind the code. When that balance sheet is leveraged 5x, the code might as well be written on sand.
But there’s a deeper irony. The crypto industry was founded to create alternatives to debt-driven, centrally-planned systems. Bitcoin mines are financed by capital, not credit. Ethereum validators stake their own ether, not borrowed money. Yet the new wave of infrastructure—L2s, data availability layers, zk-proof generators—is borrowing from the same playbook that made AI infrastructure fragile: borrow big, build fast, worry about solvency later.
Here’s the contrarian take: maybe this debt is actually a sign of maturity. After all, traditional internet infrastructure (data centers, fiber networks) also grew on debt. The difference is that those assets had long, predictable revenue cycles. AI and crypto infrastructure assets have depreciation schedules measured in hardware generations (2-3 years), while debt maturities are 5-7 years. That mismatch is a time bomb. If a new GPU architecture makes current training clusters obsolete, the debt remains. If a better Layer-2 emerges and users migrate, the sequencer’s debt stays.
I don’t think we need less infrastructure—we need smarter funding models. The lesson from the AI debt boom is not that debt is evil, but that unsecured, high-leverage debt in a fast-iterating technology sector is a systemic risk. Crypto can do better. We can fund infrastructure through protocol treasuries, community-owned liquidity pools, or staking derivatives rather than bank loans. We can build in incentives that reward long-term stability over short-term scale.
As I left that café, I told the founder: ‘Your sequencer might be decentralized, but your debt is centralized in one bank. If they call the loan, who runs the L2?’ He didn’t have an answer. We need to start asking these questions before the next bear market writes the answer for us.