Yield Rate

Imagine a bakery that puts a hundred identical loaves into an oven, but due to tiny, unavoidable variations in heat, dough and timing, only seventy come out perfectly sellable while the rest are undersized, misshapen or simply ruined. Yield rate measures exactly this outcome for a chip fab, the percentage of chips on a wafer that actually work and meet specification once manufacturing is complete.

A silicon wafer typically holds hundreds of individual chip dies, and manufacturing at the nanometre scale is genuinely difficult to get perfectly right every single time. Microscopic contamination, tiny misalignments or process variation can render an individual die defective, while its neighbour on the same wafer works perfectly. Yield rate is simply the proportion of dies on a given wafer that pass testing and are good enough to sell.

This single number decides a fab's actual economics far more than the sticker price of its equipment does. Fabs incur essentially the same cost to process a wafer whether the yield turns out to be 95% or 60%, so a low-yield process effectively means paying full manufacturing cost for a much smaller number of sellable chips, directly and severely compressing margins.

Yield typically starts low when a new process node or a new fab first ramps up, and improves steadily over months or years as engineers identify and eliminate sources of defects, a curve known in the industry as yield ramp. Mature, well-understood processes, like the 28nm node India's first fab targets, generally yield well above 90%, precisely because the manufacturing kinks have already been worked out over years of industry-wide experience.

Whenever a new fab's early results are discussed cautiously, or a country's semiconductor ambitions are described as a multi-year rather than overnight project, yield ramp, the slow, expensive process of actually getting good at manufacturing, is usually the real reason results take time to show up.