-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
Intel in talks to buy GlobalFoundries for about $30 billion
- Thread starter sidthesexist
- Start date
This isn’t about Intel getting better. It’s about taking away options from Nvidia and RTG while forcing them to compete for wafers. Intel can’t improve but they can produce lots of garbage and sell it at a loss.
Associate
- Joined
- 27 Apr 2007
- Posts
- 966
Obviously they wouldn't be primarily buying buildings.
The cost of creating a large fab building even without the tools for a cutting edge node is still staggeringly high.
I am sure it is much quicker to retool a working fab than to start from scratch or to bring a mothballed building up to speed.
So this is another potential positive, but hardly the main reason.
It's not quite that simple, but certainly a large part of it.
Associate
- Joined
- 29 Jun 2016
- Posts
- 543
Fabs require REALLY special environmental control to be put in. The kind that needs the building to be specifically designed for it.
Associate
- Joined
- 29 Jun 2016
- Posts
- 543
I think people are over-estimating the importance of cutting edge nodes outside of HUGE dies on CPUs, GPUs, FPGAs, etc.
A hell of a lot of Intel's devices will be fabbed on nodes that are 2+ "generations" old, and will stay there for 10+ years. There are various reasons, but it often comes down to power consumption, robustness, safety and validation costs.
For example, their Altera division are still offering Max V CPLD's (~FPGA) that are fabbed on 180nm. These devices are still heavily used throughout industry and is just one example of hundreds they manufacture.
A hell of a lot of Intel's devices will be fabbed on nodes that are 2+ "generations" old, and will stay there for 10+ years. There are various reasons, but it often comes down to power consumption, robustness, safety and validation costs.
For example, their Altera division are still offering Max V CPLD's (~FPGA) that are fabbed on 180nm. These devices are still heavily used throughout industry and is just one example of hundreds they manufacture.
There is also the skilled workforce which I would think costs a lot to get up to speed
Yes, this is often overlooked.
Anandtech ran a TSMC Process Node By Revenue article:
https://www.anandtech.com/show/16621/tsmc-q1-2021-process-node-revenue-more-7nm-no-more-20nm
So leading nodes are usually less than 50% of TSMC's revenue.
Pure foundries keep old nodes as long as they have customers (TSMC will move some stuff so if they have two 7nm lines, one of them might be re-tooled for 6nm as demand for 7nm tapers off).
One of Intel's problem is that aside from Atom and chipsets, they never had any use of the old nodes and re-tool most (almost all?) their factories.
Soldato
- Joined
- 25 Sep 2009
- Posts
- 10,010
- Location
- Billericay, UK
I believe Intel is still running 22nm foundry, there was talk in late 2019 moving it's chipset process to this older process to help free up capacity on it's 14nm line.
You call it a problem for Intel having to re-tool it's existing fabs but equally it allows them to vertically algin all there business operations. Also in a down market they aren't beholdant to previously agreed wafer purchasing and can ramp production up or down when it suits them.
$30b seems really high given the original AMD divestment was around several billion which included AMD's debt at the time.
Good move by Intel, with Global foundries, they'd be able to outsupply AMD by 50 to 1, instead of the current ~20 to 1.
AMD made a silly move to sell their foundries IMO, as it means they'll never be able to compete with Intel, as they simply can't produce enough CPU's. They have to beg TSMC for wafers.
Zen 3 has been a fantastic product for AMD, though hampered by terrible stock for many months. Imagine how many extra sales they'd have had, if they'd had a few million more Ryzen 5000 series in stock, when it was all the hype.
AMD made a silly move to sell their foundries IMO, as it means they'll never be able to compete with Intel, as they simply can't produce enough CPU's. They have to beg TSMC for wafers.
Zen 3 has been a fantastic product for AMD, though hampered by terrible stock for many months. Imagine how many extra sales they'd have had, if they'd had a few million more Ryzen 5000 series in stock, when it was all the hype.
20:1 is a joke. So you're saying AMD have 5% of the market?
And will head to 2%?
You might want to look at marketshare figures rather than plucking figures out of the air.
Besides, AMD's supply is almost totally dependent on how many TSMC 7nm wafers they can get.
Consoles might be using as many as 120,000 wafers which is the equivalent of nearly 80 million Zen 3 CCDs.
TSMC's annual 7nm is currently around 1,000,000 wafers, so Zen3 CCDs output is mainly depend on how much of TSMC's output AMD can buy.
Obviously they'd never get 100% of the output, but there was a race between TSMC and Intel: can Intel produce 600 million Zen 3 CCDs equivalent dies?
And will head to 2%?
You might want to look at marketshare figures rather than plucking figures out of the air.
Besides, AMD's supply is almost totally dependent on how many TSMC 7nm wafers they can get.
Consoles might be using as many as 120,000 wafers which is the equivalent of nearly 80 million Zen 3 CCDs.
TSMC's annual 7nm is currently around 1,000,000 wafers, so Zen3 CCDs output is mainly depend on how much of TSMC's output AMD can buy.
Obviously they'd never get 100% of the output, but there was a race between TSMC and Intel: can Intel produce 600 million Zen 3 CCDs equivalent dies?
Soldato
- Joined
- 25 Sep 2009
- Posts
- 10,010
- Location
- Billericay, UK
Trolling or just ignorance? He talks about AMD making a mistake divesting it's fab business but forgets AMD only sold that off so they could raise some much needed working capital and clear their debts so they could avoid bankruptcy.
Associate
- Joined
- 29 Jun 2016
- Posts
- 543
Oh absolutely. There is literally no one who isn't highly trained. Think of cleaning a clean room
Associate
- Joined
- 29 Jun 2016
- Posts
- 543
Awesome data! I thought TSMC might have 30% of revenue in 5nm and 7nm, so lovely to get some firm numbers. I would guess this is because TSMC are undeniably the world leaders in cutting edge nodes, and that other fabs will have higher revenue from older nodes.
Soldato
- Joined
- 25 Sep 2009
- Posts
- 10,010
- Location
- Billericay, UK
A remarkable turnaround given all the troubles that had moving on from 28nm which they seemed they were stuck on for an eternity
Based on the general themes and tone of his posts, I'm going with the former. Literally every post is about how AMD is inferior to Intel in some capacity, and as each point he raises is proven false, some other spurious claim is made.
The layman, this may look like the truth.
Dig a litle deeper, you'll see Intel currenly out-supplies (and thus outsells) AMD by roughly 20:1. AMD have no manufacturing capacity of their own.
This recent news of Intel buying Globalfoundries, for $30 billion - will let Intel out-supply (and out sell) by roughly 50:1.
Intel have already won.
AMD had a really god design with Zen. Ryzen 5000 series are the best choice of CPU currently for most people/uses, but AMD simply can't produce enough CPU/APU/GPU for all markets/segments worldwide to have any kind of long term impact.
It doesn't matter how great a deisgn of CPU/GPU/APU you have, if you have to beg/compete with Apple, Nvidia and all the other companies for wafer allocation at TSMC. Apple will always get priority. AMD just can't produce enough to compete with Intel in terms of sales, supply.
Last edited:
I don't see this as a good thing and I doubt this will ultimately get past regulators. There are only a handful of companies with this capability and Intel swallowing up one of them has implications on the competitiveness of the market, particularly where there is a global shortage....
If intel want to supply other companies, I'd rather they built up that capacity themselves rather than just swallowing up a competitor.
Intel don't actually have the cash to buy them so this will be a share deal meaning they are just using their market value to effectively erase a competitor.
If intel want to supply other companies, I'd rather they built up that capacity themselves rather than just swallowing up a competitor.
Intel don't actually have the cash to buy them so this will be a share deal meaning they are just using their market value to effectively erase a competitor.
Last edited: