Goldman Sachs dropped a bombshell last week that most of the crypto world missed. $600 billion in AI capex this year. That’s 2% of U.S. GDP, 10% of corporate fixed investment, 15% of equipment investment. Yet the net boost to GDP growth by 2026? A measly 0.1 percentage point. The market is overinterpreting the macro impact of all that spending. Sound familiar? Because the same pattern is playing out in crypto infrastructure — except the numbers are even more inflated, and the crowding-out effect hits closer to home.
I’ve been watching this play out since 2017, when I audited over 40 ICO whitepapers and saw capital pour into projects that promised to change the world but delivered only governance holes. Back then, the hype was “blockchain will revolutionize everything.” Today, it’s “rollups will scale everything.” The Goldman report is a gift to anyone who’s been paying attention. It gives us a framework to question the real economic impact of our own infrastructure buildout.
Let’s break down what Goldman actually said. They highlighted three distortions: AI equipment imports inflate domestic output numbers, AI investment crowds out other cloud budgets, and AI debt financing raises costs for everyone else. In crypto, the equivalents are everywhere. Mining rigs imported from China? That’s our import distortion. The billions spent on L2 sequencers and data availability networks? That’s crowding out capital that could go to actual user-facing applications. And the DAO treasuries issuing debt at 8-12%? That’s the financing cost that squeezes out smaller builders.
The core insight: just like AI, crypto infrastructure is a massive story of capital reallocation, not net new value creation.
Based on my experience running OpenLedger Academy, I’ve seen teams raise $50M for a rollup that has fewer daily active users than a 2018 DApp. The technical narrative is seductive: “We need more blob space, more L2s, more DA layers.” But the data doesn’t support it. Post-Dencun, blob gas is already creeping up. My models — built from on-chain data — suggest that within two years, blob space will be saturated, and rollup fees will double again. That’s not scaling. That’s a temporary subsidy that will vanish.
Then there’s the governance trap. Goldman says AI capex is concentrated in a few players — Nvidia, cloud providers, data centers. In crypto, the same concentration exists. The multi-sig wallets that control most L2 upgrade keys? Those are the equivalent of centralized AI budgets. “Code is law” is a beautiful slogan, but every time I audit a DAO contract, I find three admin keys that can override any vote. Democracy isn’t a transaction where every voice holds weight when the admin can call upgradeTo.
The Lightning Network is the perfect case study. Seven years of development, endless routing failures, channel management complexity that keeps it a niche tool. It’s the crypto version of AI’s overhyped capex: lots of investment, very little macroeconomic impact. The market keeps treating it as a scaling solution, but the data shows it’s half-dead. I’ve spoken to node operators who manage 20 channels and still can’t route a payment reliably. That’s not a network; it’s a hobby.
Here’s the contrarian angle: the real value of crypto isn’t in infrastructure at all — it’s in cultural and identity shifts.
When I curated SoulBound Stories in 2021, I saw that NFTs were never about sale volume. They were about identity. People wanted to own a piece of their digital self, not just a speculative asset. The same goes for DAOs: the value isn’t in the multi-sig but in the community that holds the multi-sig accountable. Goldman’s analysis misses this because they only measure GDP and capital flows. They don’t measure trust, autonomy, or resilience.
My latest project, TruthLayer, is a direct response to this. We’re using blockchain timestamps to verify AI-generated content. The irony is that the same infrastructure that Goldman says has minimal macro impact is the only tool we have to fight the AI deepfake crisis. The capital allocation debate is missing the point: we’re not trying to boost GDP by 0.1%. We’re trying to preserve the integrity of human discourse.
The takeaway: stop measuring crypto by how much capital it consumes. Start measuring how much agency it creates.
The next cycle won’t be about which L2 has the highest TVL. It will be about which applications restore trust in a world drowning in AI-generated noise. Democracy isn’t a transaction where every voice holds weight; it’s a continuous process of verification. And that’s exactly what blockchain is built for — not as a GDP booster, but as a truth layer.
So when you read the next headline about $600 billion in AI capex, ask yourself: how much of that is actually creating new value, and how much is just rearranging the deck chairs? The same question applies to every rollup, every DAO, and every shiny new protocol. The numbers look impressive, but the real test is whether they make society more transparent, more equitable, and more resilient. That’s the only metric that matters.
