Andy Grove gave a keynote at hot chips, I think around 1999. Back then, almost everyone had a fab. Arm wasn't a big thing, the iphone didn't exist, smartphone generally didn't exist, and many companies had fabs. ,<p>He basically outlined how each process shrink would allow more transistors, but also get more expensive. His conclusion was that fewer fabs would use the leading process in each generation and that costs would almost double for each generation.<p>It wasn't a particular opinion at the time. Generally it was thought if you were serious about making CPUs, that you would have your own fab.<p>Impressive how true his predictions were.
Can someone explain how this works?<p>We've got the following chip companies: Intel, AMD, Samsung.<p>And we've got a bunch of "fab" companies: Intel, TSMC, Global Foundries, Samsung.<p>What's the difference between the two sets of companies? Why can't Intel just buy a "7nm" processor from TSMC?<p>And what exactly is "7nm"? The distance between the closest transistors in a chip? Just guessing.<p>[edit]<p>From browsing Wikipedia, I gather that:<p><pre><code> - A chip company designs a processor. The design is in the form of a circuit diagram(?)
- A fab company turns that diagram into a physical product.
</code></pre>
So for example, ARM designs a chip, which then gets built by Samsung. ARM is strictly responsible for design, and Samsung provides a factory.
I'm reminded of the chart in this article showing the number of leading edge foundries dropping with every generation:<p><a href="https://www.extremetech.com/extreme/240901-tsmc-announces-plans-new-16-billion-foundry-push-envelope-5nm-3nm-process-nodes" rel="nofollow">https://www.extremetech.com/extreme/240901-tsmc-announces-pl...</a><p>and now one more step down.<p>(As explained in the article, GF is not in the righmost box because they were not yet on the leading process when the chart was made).
>Along with the cancellation of the 7LP, GlobalFoundries essentially canned all pathfinding and research operations for 5 nm and 3 nm nodes.<p>Is it just me or does this seem like a drastically bad move for the long term?
This is the type of news that will be relevant in 20 years.<p>There are only 3 foundries left: Intel, TSMC and Samsung.<p>If (when) Intel gives up, none of them will have strong roots in the USA. One of the biggest shifts in technological expertise from West to East in history.
What unusual thing was happening as of recently, is that legacy processes up to 90nm began "coming back back from the dead," because the bleeding edge became so "congested" by very few behemoth consumers with exclusive deals with fabs: GPUs, top tier phone SoCs, and high end network switch chips.<p>On my memory, 65nm was the last process on which a "cookie cutter SoC" was still a good business. But with more opportunities coming up in "niche microcontroller" market today, thanks to boom in "smart things," a generic cheap low power process might too become a viable business again.<p>GloFo might have just noticed that, and are trying to capitalise by being first in the new niche: tier 1 fab service on cost optimised legacy process.<p>They were already the biggest fab for companies to whom always getting the best process is not raison d'etre, who can't afford gigantic MOQs of last gen processes.<p>So they were picking up whomever TSMC was losing due to MOQs, and lack of first class treatment. TSMC was too greedy putting so much focus on work with tier 1 superplayers.<p>Question thought, what will this mean for their no. 1 customer...? Though they announced them moving to TSMC, their 7nm might still not materialise for quite some time, and they still have to make their low end chip tapeouts somewhere.
AFAICT this means TSMC has no competition for 7nm for their fabless customers, which means they have no incentive to invest in getting to the next process node — their customers' BATNA is now "spend several billion dollars building your own <7nm fab." So I think this is the end of Moore's law for everyone except possibly Intel and Samsung.
This explains the sudden correction AMD saw mid-rally today.<p><a href="https://www.google.com/search?q=NASDAQ:AMD" rel="nofollow">https://www.google.com/search?q=NASDAQ:AMD</a><p>The impact of the TSMC monopoly on 7nm might not have been fully priced in quite yet as I doubt everyone caught this bit of news.(Disclosure: I'm long AMD)<p>Edit: Actually, this news broke embargo at 4pm eastern, no? The downswing took place mid-day...
See also <a href="https://news.ycombinator.com/item?id=17851278" rel="nofollow">https://news.ycombinator.com/item?id=17851278</a>
What are the political implications of the entire world becoming so reliant on TSMC, with nearly all of its manufacturing concentrated in Taiwan?<p>It's no secret China is eyeing semiconductor manufacturing as sort of a last frontier they need to cross before they become a vertically integrated powerhouse. If China took over Taiwan, and assumed influence over TSMC, wouldn't this be a major achievement?<p>1. Why aren't TSMC scattering their fabs across different continents - not only for political, but also to protect against natural disasters etc?<p>2. How much of the USA protection of Taiwan takes into account semiconductor manufacturing?