Over the past 90 days, I watched a protocol lose 40% of its LPs in a single week. Not because of a hack, not because of a governance exploit—but because the execution layer upgrades they had been waiting for finally landed, and their entire liquidity provision strategy became instantly obsolete. The irony isn't lost on me: we built faster chains, but we forgot to build the bridges that let users actually walk across them.
This isn't an isolated incident. It's a symptom of a deeper structural problem that mirrors the 'time misalignment' Big Tech is currently grappling with in AI. The same article that made headlines at Crypto Briefing—warning that tech giants may need to rethink AI spending because technology is evolving faster than enterprise adoption—is a story that crypto has been living for years. We just haven't named it yet.
Context: The Infrastructure-Application Gap
The parallel is striking. In AI, model capabilities are doubling every 6-12 months, but enterprise customers take 12-24 months to integrate even basic copilots. In crypto, L1 throughput has jumped from 15 TPS to well over 100,000 TPS (when you count rollups and sharding), but daily active users on the top ten chains haven't grown proportionally. According to data from Dune Analytics, the number of unique active addresses across Ethereum, Solana, and Avalanche peaked in late 2021 and has only recently recovered to about 70% of that level—despite a 10x increase in on-chain capacity.
We built a superhighway, but the cars are still stuck in the driveway.
This gap isn't just about user numbers. It's about capital deployment. The money flowing into infrastructure—new L1s, modular frameworks, zkEVM rollups—is being driven by a narrative of 'future demand,' not current usage. The original article's insight about 'time misalignment' applies perfectly: the investment cycle is accelerating faster than the adoption cycle. And when that happens, you get capital misallocation at scale.
Core: The Three Dimensions of Crypto's Time Misalignment
Let me break this down with the same analytical rigor I used during DeFi Summer when I audited 150 Uniswap V2 pools. I saw the same pattern then: liquidity was being deployed to chase yield, not to build sustainable markets. Today, the pattern is infrastructure.
Dimension 1: Infrastructure oversupply
We now have over 60 active L1s and 80 L2s, according to L2Beat. Most of them offer similar capabilities: high throughput, low fees, EVM compatibility. But the differentiation is marginal. The original article's analysis of AI's 'technology roadmap uncertainty' maps directly here: when the technical paradigm is still shifting (e.g., from optimistic to zk-rollups, from monolithic to modular), pouring billions into a specific architecture is a bet that could be obsolete in 18 months. Liquidity isn't just about money; it's about trust—and trust in any single chain's longevity is eroding as the field becomes more crowded.
Dimension 2: Application layer stagnation
Despite the infrastructure boom, the application layer remains dominated by DEXs, lending protocols, and NFT marketplaces. We haven't seen a breakout use case that pulls in non-crypto-native users at scale. The original AI article noted that only 30% of enterprise AI pilots reach production. In crypto, the percentage of dApps with more than 1,000 daily active users is probably under 5%. Mining for truth in the noise of NFT mania reveals that most of the 'innovation' is still speculative churn, not real-world utility.
Dimension 3: User adoption bottleneck
The bottleneck isn't technical—it's experiential. Setting up a wallet, funding it, bridging assets, and understanding gas fees is still a multi-step process that terrifies normal people. The AI industry has the same problem: Copilot is powerful, but most office workers don't know how to prompt it effectively. The technology is ready, but the human layer isn't.
During my time at the Berlin Hackathon in 2017, I watched a team build a decentralized identity protocol in 48 hours. It was elegant. But getting a single user to actually use it required a months-long onboarding process. That hasn't changed. We didn't build a future; we built a mirror—reflecting our own technical obsessions rather than the needs of the people we claim to serve.
Contrarian: The Slowdown Is a Feature, Not a Bug
Here's the counter-intuitive angle the original article barely touched: investment slowdowns are often the best medicine for ecosystems. When the AI hype cycle cools, it forces companies to focus on actual product-market fit. Same in crypto.
A pause in infrastructure spending could be exactly what we need. It would force capital away from 'build another chain' and toward 'build something that people actually use.' The original article's analysis of 'competitive dynamics' suggested that capital-rich players like Microsoft can afford to wait, but capital-constrained ones may be forced to pivot. In crypto, that means the next wave of innovation won't come from a new L1 with a 100,000 TPS claim—it will come from an application that makes onboarding take 10 seconds instead of 10 minutes.
But there's a risk. If the slowdown is too sharp, we lose the builders who are working on genuinely novel infrastructure—like zk-proofs for privacy, or decentralized sequencers for censorship resistance. The original article's risk assessment about 'AI winter' has a crypto parallel: if venture capital dries up, the next generation of protocols may never get built. Open source is not a license; it's a state of mind—and that state of mind needs funding to survive.
Takeaway: The Bridge, Not the Road
The next bull run won't be won by the fastest chain. It will be won by the ecosystem that finally bridges the gap between technical promise and user reality. We've spent five years building roads. Now we need to build the cars, the drivers, and the traffic lights.
The original article's question—'Are Big Tech spending too much on AI?'—is the wrong frame. The right question is: 'Are they spending on the right things?' In crypto, the answer is clear. We've overbuilt the infrastructure and underbuilt the adoption layer. The time misalignment is real, but it's not a death sentence. It's a signal to pivot.