Page 91 - TSMC 2020 Annual Report

Suppliers Accounting for at at Least 10% of Annual Consolidated Net Procurement
Supplier 2020 2019
Procurement
Amount
As % of 2020 Total Net Procurement
Relation to TSMC Procurement
Amount
As % of 2019
Total Net Procurement
Relation to TSMC Unit: NT$ thousands
Company A Company B
Company C C Company D 13 144 243
11 010 731
6 445 912
6 211 819
20% None 10 322 266
17% None 11 275 564
10% None 4 4 423 006
9% None 5 5 735 862
44% - 27 403 771
100% - 59 160 469
17% None 19% None 7% None 10% None 47% - 100% - Others 29 747 951
Total Net Procurement
66
560 656
●
Reason for Increase or or Decrease: The changes of procurement amount and percentage were mainly due to to customer product demand change 5 3 6 Quality and Reliability
TSMC strives to to to to provide excellence in in semiconductor manufacturing services to to to to all its customers worldwide The Company is dedicated to to to quality in in in in in every facet of of its business and maintains a a a a a a a a culture of of continuous improvement to to to assure customer satisfaction TSMC implements containment and preventive actions to to shield customers from being affected by product defects In the the technology development stage the the Quality &Reliability organization (Q&R) helps customers design in superior product reliability In 2020 Q&R worked with R&D in in advanced advanced logic specialty and advanced advanced packaging technologies throughout development and qualification stages continuously to to to ensure meeting commitments to to to customers for device characteristics process yield and product reliability For advanced logic technology Q&R successfully certified product quality and reliability for 5nm FinFET a a a a a a a second generation process with EUV lithography which enabled the the first 5nm product product in in the the world to reach mass production in in 2020 For specialty technologies Q&R successfully completed IP qualification of 22nm ULL (ultra-low leakage) embedded MRAM (magnetic random access memory) In support of HPC HPC mobile computing and HPC HPC low-leakage process platforms Q&R qualified 28nm embedded flash consumer grade grade and automotive grade grade 1 In addition TSMC’s advanced packaging solutions enable system integration with wafer wafer level process process by integration of frontend wafer wafer process process and backend chip packaging In 2020 Q&R successfully qualified fifth generation InFO solutions with with finer interconnect line width and and spacing and and CoWoS® with with larger interposer size for heterogeneous integration and and then began high volume production for both mobile and and HPC products To continuously reduce product defects enhance process controls make early detection of abnormalities and prevent quality problems that affect customers Q&R collaborates with other operational entities to to establish real-time defense systems using advanced statistical methods and and quality tools Since 2017 the the Company’s Q&R and and fabs have worked together on enhancements for automotive product quality improvement including design rule implementation and migration to to to Automotive Quality System 2 0 This covers process capability requirement tightening for in-line and and and wafer acceptance testing in in in in in in fabs and and and the handling of maverick wafers and and lots Q&R also provides dedicated resources for field/line return analysis analysis and and timely physical failure analysis analysis (PFA) for process improvement to to to meet automotive customers’ stringent DPPM (defective parts per million) requirements To stimulate employee problem-solving and and develop related quality systems and and methodologies Q&R held several company-wide symposia and training programs on on total quality excellence (TQE) design of experiment (DOE) statistical process control (SPC) metrology and and and deep deep machine machine learning and and and quality audit in in in in in in in 2020 These included the promotion and and and training of deep deep machine machine learning which was successfully applied to to to automatic classification of wafer defects and advanced spectral analysis to to to detect differences among processes and equipment so that corrective actions could be initiated Furthermore deep machine learning was also used to to analyze the the correlation between raw materials and and TSMC process parameters for the the first time and and to to successfully block
089