Hacker News Reader

Comment by x87678r | Hacker News Reader

The bay area should be renamed Social Media Valley. Hardly anyone there cares about fabs.

companies headquartered in Santa Clara Valley that care about fabs:

Intel, AMD, Nvidia, Lam Research, Applied Materials, KLA, GlobalFoundries, Broadcom, Apple, Xilinx, Maxim Integrated (now part of Analog Devices), Linear Technology (now part of Analog Devices), Marvell, Cisco, Cypress, Google, Cadence, Synopsys, Western Digital, Seagate.

I have to wonder - how much of the fabs going to Taiwan, Shanghai, SK are due to

1) government support - i.e. the national government saying "we want this sector to succeed" and providing funding via grants/loans/subsidies, (or by fiat w/r to PRC)

2) actual labor costs?

3) talent, i.e. American-trained Chinese/Taiwan/Korean talent coming home, then generation 2.0 starting up research at Chineses/Taiwan/korean universities?

Assuming a long-term visionary US government wanted to compete; would we have to fund fab research, development, and capital i.e. similar to the pentagon funding Newport News shipyard to keep strategic know-how alive in the US market?

Probably all of the above. The founder of TSMC was educated at Harvard/MIT/Stanford, and spent the first 25 years of his career working for Texas Instruments in Silicon Valley. He founded TSMC because the Taiwanese government gave him a large grant to found a nonprofit research institute in Taiwan. He was looking at the competitive landscape and realized that the one way Taiwan could capitalize on its biggest advantage (low labor costs of fairly educated labor, at the time) was to do only manufacturing and outsource the design, IP, and marketing to other countries that were better at it.

1) gov't support: TSMC's had plenty of gov't support from the getgo and my understanding is that the govt still has some financial stake in TSMC -- which would be a red flag in the US. Samsung has a big $15B fab in Austin, TX aka, FTZ foreign-trade zone, where Apple's A chips were made until the early 2010's, but after Apple moved to TSMC in Taiwan in 2012-13, it stopped investing in smaller nodes -- now, it's mainly producing memory products (and Tesla's FSD chips). All their latest nodes (< 14nm) are in South Korea.

2) labor cost: probably very insignificant. While they employ high-paying high-tech workers, it's a capital-intensive, not labor-intensive, business. Both TSMC and Samsung spend over $10+B annually on CAPEX.

3) brain drain?: Morris Chang worked for TI for decades before founding TSMC. He pretty much started the outsourcing business in the late 80 single-handedly.

Sure. I don't agree with Trump on a lot of things, but he's at least doing something to bring them back.

Yes, but compare their combined market capitalization (of course, excepting Apple/Google) to any one or two of FAAMNG :-)

If a substantial fraction of the semiconductor industry (IDM + fabless + fab suppliers) is in Silicon Valley, then the name still makes sense even if there is a lot of other money floating around the area.

also, why would this make sense:

> of course, excepting Apple

Apple is probably the single largest customer of TSMC, and their products depend on their semiconductor design team (especially true now that they are moving to in-house CPUs for laptop/desktop). Why throw them out from a discussion of companies that care about fabs?

If one asked the question of which location in the US the semiconductor industry has the highest footprint in, the answer would definitely be the Bay Area (maybe California, to include San Diego, etc).

But if you ask the question of which sectors the Bay Area has the largest investment and footprint in, then semiconductors would (unfortunately) be only a small part of that. That uncomfortable truth (which is exactly what I'm trying to point out in the above comment) is that market valuations today seem to be driven by the things built on top (in particular, the "data economy" and the effectiveness of personalization-driven consumption increases), rather than (for eg.) the semiconductor engg which is crucial to build those platforms.

An analogue of this is when the Ford CEO speculated in 2018 to the effect that it might be more profitable to sell customer data derived by selling cars, than from selling cars in the first place.

TSMC would be third and then there would be a gap again...

Because it isn’t clear if it matters that much at all. You are using devices with larger transistors than what TSMC is shipping right now. If it was a 14nm Intel chip in your iPhone it wouldn’t matter that much, your phone was shipping 14nm chips a few years ago and you still bought one. People are still using those 3 year old phones and by all indications the old phone use trend is accelerating.

However it is utterly inconceivable that anyone would “go back” to using some old social media platform.

>> However it is utterly inconceivable that anyone would “go back” to using some old social media platform.

I haven’t reacted my compuserve account, but I’m fine using email, IRC, phpBB, mailing lists...

E-Mail, IRC, forum software, mailing lists aren't at all comparable to Facebook. They do very different things.

I was really talking about, for example, Vine. You'd rather use TikTok than Vine. There might not actually be any social networks from late 2016 no one uses today that I've heard of, so it's hard to give an example that looks back to when 14nm was the typical node size for phones. Which is sort of my point, it is in some strict sense of the word literally inconceivable to go back. Whereas I remember what an iPhone 7 was like, and it's not that big of a change.

I don't know if you did it in purpose to make a point, but vine is gone, isn't it?

I'm no expert but surely fast chips still matter in stuff like AI? Although, one could argue that modern AI hasn't had a huge impact yet.

Anyway, my point is that having faster and more energy efficient chips would make the training of stuff like GPT-3 cheaper.

Surveillance Capitalism Valley

Reply on news.ycombinator.com