TSMC: Customer Contract Economics, Capacity & Yield
Executive Profile Hidden
Summary
Subscribe to access hundreds of interviews and primary research
Interview Transcript
This is a snippet of the transcript.to get full access.
Related to that, what strategies, if any, can customers use to reduce their dependency on TSMC? What can they do?
We're always evaluating new foundries. Even if Intel or others aren't executing well, just considering them brings theatrics to procurement. For example, when TSMC visited Qualcomm, I made sure they saw Samsung and Intel sales teams in the lobby. This way, they could see each other and be checked in one at a time. Such tactics encourage them to talk about what they observed. TSMC is so hyper-aware that it sometimes works against them.
This is a snippet of the transcript.to get full access.
Regarding pricing itself, we discussed this. It's not among the first couple of things that matter, but in the end, especially at the mature nodes, pricing matters more because there's more competition to choose from. Could you explain how TSMC's pricing power has evolved over time and across the nodes? This has obviously evolved a lot.
We also meet quarterly because the key term is "not to exceed," meaning we hope to do better. We generally assume a 2% takedown rate per quarter for planning, but we aim to improve on that. Looking at market pricing now, for older technologies like 28 nanometer and higher, wafer prices are clearly below $3,000. For 12 to 16 nanometer, it's around $4,000, 7 nanometer wafers are around $9,000, and a 5 nanometer wafer is probably around $17,000. As we move down the nodes, wafer pricing doesn't increase linearly.
This is a snippet of the transcript.to get full access.
Regarding pricing itself, we discussed this. It's not among the first couple of things that matter, but in the end, especially at the mature nodes, pricing matters more because there's more competition to choose from. Could you explain how TSMC's pricing power has evolved over time and across the nodes? This has obviously evolved a lot.
In older nodes, the fabs are almost fully depreciated, meaning a higher percentage of the cost is variable. There's more leverage for these older fabs. TSMC, if they want load there, must either get more aggressive on pricing or use the bundle strategy. TSMC's financials show a gross margin around 50% or better. They likely make 70% on 5 nanometer and 7 nanometer, and 30% on older technologies, trying to smooth it out. Knowing TSMC's costs at older technology nodes, I don't think they would go below a 30% margin on any business, except in rare cases, to avoid setting a precedent.
Free Sample of 50+ Interviews
Sign up to test our content quality with a free sample of 50+ interviews.
Related Content

NVIDIA: Beyond the Wafer Constraints, Packaging, HBM & Testing
Former Senior Executive at Qualcomm

TSMC vs. Intel, Samsung & GlobalFoundries: How Fabless Customers Choose Foundries
Former Senior Vice President at Qualcomm

TSMC, Samsung, Intel & Leading Edge Node Competition
Former Vice President at Samsung Semiconductor

Synopsys, Ansys, and Chiplets: Backend EDA and Interface IP Demand
Former Senior Director at Synopsys
© 2024 In Practise. All rights reserved. This material is for informational purposes only and should not be considered as investment advice.