The Learning Curve & YieldNarrow moat

Beyond the machines and the money lies the least visible and most durable part of the technology barrier: the learning curve, the accumulated manufacturing know-how required to make memory at high yield and enormous scale. Buying the same lithography tools as Micron does not make a competitor Micron's equal, because the knowledge of how to wring high yield from those tools — how to tune thousands of process steps, hunt down the defects that ruin chips, and manufacture reliably at volume — is the product of decades of hard-won experience that resides in the company's engineers, processes, and organizational memory. Yield, the percentage of good chips per wafer, is the hidden determinant of memory economics, and it is climbed sequentially over years, not purchased.

This tacit, experience-based capability is what turns the capital and process advantages into a durable moat. A well-funded newcomer with the same equipment still cannot match an incumbent's yields, because it lacks the accumulated learning — and low yield means high cost, which in a commodity is death. Micron's decades in the business, its absorption of earlier makers' knowledge, and its continuous manufacturing at scale give it a learning-curve advantage that a challenger cannot leapfrog with money alone. One weakness: the learning curve, being human, can leak: engineers can be recruited away, and a determined state-backed effort to acquire memory-making expertise — through hiring, partnership, or less legitimate means — can, over time, narrow the gap, which is precisely part of China's strategy. The learning curve is the deepest and most durable part of the technology barrier, but because it lives in people and practices rather than in law, it is the part most exposed to being slowly, deliberately learned by others — the risk export controls exist to slow¹.