What to Check Before Interpreting a 90% Semiconductor Yield
Semiconductor yield depends on how good units are counted and at which testing stage. Learn to distinguish wafer, die, and package yields and calculate cost per good unit.
Yield is the proportion of input units that pass specified quality criteria. It is meaningful only when you know the production stage and denominator, and yields for different chips or test conditions should not be compared directly on the numbers alone.
Define What Counts as a Good Unit First
ASML describes yield as the proportion of functioning chips on a wafer. When reading industry materials, also check the power, speed, and functionality requirements that chips must meet. A yield that counts only the highest performance grade differs from one that counts all saleable chips, including lower grades. Small test chips used in process development also face different exposure to defects than actual large products. It is sensible to exclude yield reports of unclear origin from detailed cost calculations until you have specific information about the denominator and product.
Connect the Testing Stages in Sequence
Chips that pass wafer testing can still fail testing after packaging. Advantest distinguishes wafer testing from final testing after packaging. When combining pass rates across stages, check whether each stage uses the preceding stage's passing units as its denominator. Multiplying again by a yield already measured against the original input double-counts defects. If some products are recovered through rework, include the recovery costs and additional time to understand the difference between a simple pass rate and actual production efficiency.
A High Yield Is Not Enough on Its Own
More good units allow the same input cost to be spread over more products. However, if the pass rate rose because testing standards were relaxed, selling prices may also fall. If improved yield required spending more on testing and rework, calculate its economic benefit separately. Analysts should examine the sales mix by grade and total cost per good unit, as well as the direction of yields. For yields that have not been officially disclosed, avoid settling on an arbitrary single figure; use several ranges only to compare profit sensitivity. An improvement from 80% to 90% is an increase of 10 percentage points, while the number of good units from the same input increases 12.5%. Confusing percentages with percentage points misstates the improvement.
Worked example
Hypothetical example: Of 1,000 candidate dies, 900 pass wafer testing. If 95% of those 900 pass final package testing, 855 chips are saleable. Overall yield is 90% × 95% = 85.5%. Allocating total input costs of KRW 8.55 million to these good units gives KRW 10,000 per unit. Adding the two stage yields or applying 95% to the original 1,000 dies produces an incorrect final good-unit count.
What to check
- Check whether the denominator is wafers, dies, or packages
- Check for changes in testing criteria and saleable grades
- Distinguish stage pass rates from cumulative yield
- Include the additional costs incurred to improve yield
Sources and primary materials
This is educational material about research methods. For an actual company decision, verify the filings and primary sources for the relevant period.