It amazes me that so many people seem to think TSMC is some simple-minded "EVERYONE'S FOUNDRY", only interested in short term profit maximization.
Come on, think again. TSMC didn't get to where is today by being an opportunistic money grab, sans ability to think strategically.
Come on, think again. TSMC didn't get to where is today by being an opportunistic money grab, sans ability to think strategically.
Modern fabs cost upwards of $10 billion apiece to build. ROI for capital equipment alone is many, many years, not to mention astronomical operating costs on an ongoing basis. If you don't keep those fabs full during it's entire lifetime, you'll go bankrupt pretty damn quick.
Let's assume the rumors are true and Intel does intend to outsource manufacturing of it's crown jewels (mainstream CPUs) to TSMC in the near future, at scale. In order to satisfy Intel's volume, TSMC is going to have to invest many, many billions in new capacity.
Now put yourself in the shoes of C.C. Wei and Mark Liu.
Given Pat Gelsinger's very frequent and very public pronouncements that Intel is coming after your lunch with IDM 2.0, are you simply going to grab Intel's business without thinking about the longer term implications?
Given Pat Gelsinger's very frequent and very public pronouncements that Intel is coming after your lunch with IDM 2.0, are you simply going to grab Intel's business without thinking about the longer term implications?
When Intel fixes their process woes, what do you think they'll do? Continue to outsource their crown jewels and pay 50+% GM to help fund your already massive cap-ex and R&D to further cement your lead? Or take it back in-house? Obvious, isn't it?
https://twitter.com/chiakokhua/status/1412359510069370880?s=20
Given Intel's talent pool and wherewithal, plus Pat Gelsinger's credentials, you'd be *very* foolish to think that this won't happen. Yes, Pat is not god, it will take time to turn the big ship around. Perhaps 3 years, perhaps 5, but make no mistake, it's only a matter of time.
On Intel's willingness to outbid others, in particular AMD, note that for every x86 CPU that you make for Intel, you'll probably make one less for AMD. And if AMD is defeated thus, that becomes your permanent loss. Whatever temporary gain you got is offset by a permanent loss.
To top it off, Pat Gelsinger's recent piece in Politico makes clear Intel is not your true partner or ally, and does not intend to play fair. Can you trust them not to use what you will teach them about advanced process nodes against you?
https://twitter.com/chiakokhua/status/1412257223124815873?s=20
That said, Intel is already a substantial customer, just not for bleeding edge stuff. Of course you can't simply reject Intel's business. The question is timing and scale, and whether you should prioritize Intel's business over existing long-term partners', like AMD, Nvidia, etc.
Without a super long term agreement, enough to *guarantee* attractive ROI, you'd be crazy stupid to spend the many billions on cap-ex necessary to take on Intel's mainstream CPU business at scale.
Heck, you hold all the cards, why not get Intel to prepay for capacity and even fund cap-ex?
Also, don't forget to put in tough rules to safeguard your intellectual property.
You're smart. I have no doubt.😃
- END -
Also, don't forget to put in tough rules to safeguard your intellectual property.
You're smart. I have no doubt.😃
- END -
• • •
Missing some Tweet in this thread? You can try to
force a refresh
