When AI Can Do Everything, the Remaining Moats Are Only These Five

Summary

A widely circulated essay republished on Huxiu and attributed to StartupBoy summarizes a framework from Quiet Capital partner Michael Bloch: as AI makes software and engineering trivial, durable competitive advantages will not come from smarter code but from assets that take real time, capital or political will to replicate. Short version: if your moat depends on scarce intelligence, it’s only a matter of time. If it depends on years, billions, regulation or physical reality, it may survive.

The five durable moats

Bloch identifies five categories that remain hard to compress: compounding proprietary data that is continuously generated; genuine network effects; regulatory licenses and approvals; the ability to deploy very large amounts of capital; and physical infrastructure (factories, power, satellites, data centers). Orchard AI and DoorDash are used as concrete examples: the former’s tree-by-tree, season-by-season sensor data can’t be synthesized overnight; the latter’s logistics density across thousands of cities is not cloneable by a new app. He also argues there are two plausible borderline moats — trust and human attention — but their status as independent categories is uncertain.

Why this matters to China and the geopolitics

Chinese tech giants already sit on several of these assets: Baidu (百度), Alibaba (阿里巴巴) and Tencent (腾讯) control massive live-data flows, user networks and, in some cases, deep infrastructure and capital pools. At the same time, the hardware and capital dimensions are increasingly geopolitical. Export controls, sanctions and industrial policy shape who can build advanced fabs, satellites or nuclear-scale projects. It has been reported that major capital raises and state-backed financing are driving projects that simply cannot be replicated by software alone. Who can muster the money, political approvals and decades of engineering? That question now matters as much as any algorithm.

The takeaway

If you’re building something that rests on “engineering time,” expect competition to arrive fast. If you’re building something that requires years of real-world accumulation — network density, regulatory clearance, bespoke live data, enormous capital or physical installations — you may be building a lasting position. In a world where models get cheaper and faster by orders of magnitude, time itself has become the ultimate moat.