The Reckoning of Blockchain Infrastructure: Who Will Be the First to Cut Capex?

In the last quarter of 2025, I sat through three separate investor calls for L2 sequencer providers, each touting their latest capital raise for hardware and decentralization. The numbers were staggering — one project had committed $180 million to its own data center, promising sub-second finality for a user base that barely exceeded 5,000 daily active wallets. The CFO calmly explained that the capex was ‘necessary for future scale,’ but the room felt a familiar tension: the same tension I heard when I audited EtherTrust’s reentrancy bug in 2017. Code and cash were decoupled. The market was celebrating infrastructure that hadn't yet proven its revenue. This is the moment blockchain infrastructure faces its own Alphabet-style reckoning. Over the past three years, the narrative has shifted from ‘we need more base layers’ to ‘we need the infrastructure that makes those layers usable.’ Rollups, validiums, and zkEVM sequencers have raised billions in venture and token funding to build custom hardware stacks, sequencer networks, and data availability layers. The promise: a vertical stack that gives developers sovereign execution and low latency, all while inheriting Ethereum’s security. The reality: most of these projects are burning through cash faster than they generate fees, and the early returns on infrastructure investment are vanishingly small. I’ve seen this pattern before — in 2020, during DeFi Summer, I wrote about how Compound’s governance was more sound than its treasury strategy. The same blind spot haunts infrastructure today. Teams are treating capex as a signaling mechanism: ‘Look how much we’re spending, therefore we’re serious.’ But the market is beginning to ask the uncomfortable question: what is the unit economics of a sequencer node? If each block produces $0.12 in MEV and $0.03 in transaction fees, and a single node costs $2,000 per month in cloud compute, how many blocks per day do you need to break even? The math doesn’t work for anyone below a million daily transactions — and only a handful of L2s have that volume. Here’s where the contrarian angle cuts deepest. The conventional wisdom says infrastructure investment creates a competitive moat: the team that owns the lowest-latency sequencer and the most decentralized validator set will win the next wave of users. I used to believe that, too. But after studying 20 failed infrastructure projects from the 2022 bear market, I’ve come to a different conclusion: infrastructure is a commodity that only differentiates at the application layer. The real moat is developer mindshare and user experience — not the complexity of your hardware stack. The projects that survive will be the ones that redirect capital away from building proprietary data centers and toward building protocols that abstract away infrastructure entirely. Take the case of one prominent zkEVM. They raised $45 million to build a custom sequencer network across five continents. Six months post-launch, their TVL is $8 million, and average transaction fees are still $0.50 — far higher than their centralized competitors. They are trapped by their own capex: they can't lower fees without subsidizing them, and they can't attract enough volume to lower costs. This is the infrastructure trap. The market is waking up to the fact that most of these projects are building supply before there is demand. The turning point will come when the first major infrastructure provider announces a capital expenditure reduction. It might be a rollup that closes its data center and migrates to a shared DA layer, or a sequencer that gives up on geographic decentralization to focus on a single region. When that happens, the market will interpret it not as a retreat, but as a sign that the era of ‘infrastructure for its own sake’ is over. The projects that survive will be the ones that treat infrastructure as a cost center, not a moat — and focus on the applications that make users’ lives better. Trust is earned, not mined. And right now, the industry is spending billions mining trust it hasn't yet earned.