Their revenue is higher than AMD’s but their costs are a multiple of AMD’s costs.
Really bad example; GloFo gave up on leading-edge nodes because of the expense; they simply can not raise enough cash to justify attempting to create a competing node that may or may not be successful and attract customers.
The only ones would do it and make it a viable bleeding edge company would be a collation of the big tech companies. Microsoft, Google, Apple, Amazon who both need enormous amounts of silicon but are also smart enough to know that sourcing all your chips from a single vendor (TSMC) will at best result in them getting ripped off with monopoly tier pricing.
That's not Intel's only problem; they have many. Also, having higher revenue doesn't paint the full picture as they have to have higher revenue since they have higher fixed costs and capital investment costs; for this reason, at least for a few quarters now, AMD has had better margins (and better consistency that last several quarters, interestingly).
I'd be more comfortable if Intel just ripped off the Band-Aid and sold IFS off into it's own independent company and just contracted out it's chip manufacturing to them, just like AMD did with GloFo.
That's why TSMC has been the "Gold Standard".
I think the issue is Intel would take a massive loss in that scenario, which is where I believe you are correct in that Intel is playing the Wall St. game with the maneuver. I don't think they want to do what is essentially a sale at the bottom, and are hoping that by getting outside investments they can later spin it off for a gain without having to put all the money in themselves to take that gamble.
They had plenty of years in the past to spin it off, the fact that it's taking them this long to decide to go down this route is baffling.
I think the AMD/Global Foundries case is a textbook example of how this could go wrong. Global Foundries never delivered a competitive node for AMD until after many years of Intel stagnation. Almost as soon as Global Foundries pulled that off and AMD had returned to profitability, Global Foundries gave up on bleeding edge.
The lack of EUV machines really threw a wrench in the works. Intel had to actually move some machines during the MTL launch cycle which is certainly not ideal. Intel 3 is a long term node, but I do think the combination of needing to push to 18A which was to be their first leading edge node for third parties and the limited EUV machines has impacted availability.
I agree with you there. This move would have made a lot more sense when Pat first stepped in.
How would it have made sense then? They were in the middle of massive capital expenditure for EUV machines and not using industry standard EDA for any of their production nodes. It would have been pretty much immediately killed as a potential leading edge provider. The only ones who would have benefited from this are the investor class who love killing companies in the name of profit.
I believe it was Gelsinger himself that said Arrow and Lunar silicon is tsmc, Intel is packaging only.
Nope the only official word at all came from the foundry update as Intel has been unusually quiet about ARL.
and here's the other:
The first quote uses the word "primarily" implying that not all would be external. There are no other tiles for any other product currently being made by Intel other than the Compute Tile. This is why some of us have figured it's possible that a single die is being made on 20A (Intel typically has at least 3 for each generation across desktop/mobile).
One recent rumor claims this for Arrow Lake's compute tiles:
Intel could have used the proceeds from the spin off of Mobile Eye, the sale of the modem business and other assets sales to pre-fund the Foundry and gotten off their plate and been able to re-focus on design while letting Foundry re-focus on production. I personally think they would have had an easier time back then getting outside investors on board and would have made a nice story for the "CHIPS and Science ACT". Now it feels more like desperation.
I don't even know where to start here. Nothing you said makes any sense at all. No company is going to dump billions into something that isn't healthy right before selling it off. Not only is it foolish, but it would likely end up with lawsuits they would be likely to lose.
Intel already did this, they dumped billions into it and are still doing so. Whether they did it up front or did it over 3 years the end result is no different.
How is this different than today? The plan for 4 nodes in 5 years had already been announced, the plan was, and still is, catch up to TSMC and then pass them while opening up the foundry to new customers. The plan is no different today than it was 3 years ago.
Ignoring the inflammatory comment, I've done investments for a long time and spent a good part of my early career in automotive with is an extremely capital intensive business. I understand the need for up front capital in order to produce a product. That up front capital goes a long way to attracting investors. For example just 4 years ago investors were pumping billions into unprofitable EV makers with even a slim glimmer of hope of profitability. Companies like Rivian are still not profitable, but they are well funded by investors because there is a path to profitability and had a well capitalized balance sheet. All investors really care about is risk reward, a well capitalized company that loses money is fine so long as the path to profitability does not take longer than the time it takes to run out of cash. If we are saying Intel Founder has zero path to profitability then this conversation is moot, as no matter what Intel does, the foundry business will fail.
It's actually quite a bit different because continuing to pump money into part of your business is a lot different than dumping a ton of money in and then selling.
What you're suggesting wouldn't be shoring up a balance sheet however. It would just be dumping a bunch of money in with no guaranteed path to returns beyond the sale. It's also not about misleading it's about the responsibility US law has put upon publicly traded companies to do what's best for investors.
Investment in leading edge is extremely expensive and has been something Intel's investors have been complaining about for years. The returns are only guaranteed if you execute, but given the state of semiconductor it's possible to make a profitable business in pretty much any segment as long as you have some sort of niche. Intel could afford to take a chance on the long game because of the amount of money they were sitting on and the profits from other parts of the business.
Personally I wouldn't compare investors creating a bubble (Tesla's huge valuation based on nothing tangible) in a fledgling industry (meaning EV specifically as part of the Tesla pitch was no doing things like regular car companies) to an established industry with known profit centers. Had all car companies doing EV work benefited from this bubble I might agree with the comparison, but that absolutely isn't the case.
As I've already explained it could be profitable today. That doesn't mean it would end up being a good thing for the industry or consumers though.
No, it's clear. There will be no Intel silicon for Arrow and Lunar. It's just that Intel PR doesn't want to make a big deal of that (obviously). Intel has officially cancelled 20a and 18a won't be ready until next year; possibly in time for a refresh gen.
If you open the article you linked, at the beginning it links to the announcement from Intel:
Nothing more specific about the fate of Arrow Lake on 20A was stated.
At IFS he was referring to mobile CPUs there has been no confirmation of any other sort. You posted the article that leads to the Intel article I quoted.
You presume to know what would've happened had AMD not sold it. Where would they have gotten the operating cash to keep turning out new designs? Even with the spinoff, they still had several rounds of layoffs. I think they had already cut to the bone, so don't tell me they could've just "trimmed the fat".
Consistently? Most of their availability problems occurred due to the runaway success of their Zen 2-based EPYCs, followed by the pandemic, and got sorted out by late 2022. It's not as if Intel didn't also have some availability problems, if you look back to 2019 and the pandemic years.
Samsung killed off its core design group. These days, its SoCs are just using IP licenesed from ARM. I'm not saying there's no competitive intelligence that could leak from Samsung's fabs to their product division, but Apple is so far ahead that maybe they aren't terribly concerned about this?
Not sure what your point is, because they design chips that go into embedded, self-driving (via MobilEye), laptops, desktops, and servers. The only market they don't compete in is phones, and Apple is pretty much the only competitor out there who both designs their own chips and sells end-user products containing them. Everyone else who's designing chips is in the chip game, therefore direct competitors.
So, then you have absolutely no idea how hard it is for someone to reverse engineer IP from the masks that actually reach the fab? I'll give you a clue: it's a lot harder than just disassembling some code you get in binary form.
The irony is that GloFo would've made multiple returns on its investment, if it had a mature 7 nm node during the chip crunch from 2020 to 2022. It just shows how stupid and short-sighted their decision was to kill of that node, especially when it was nearly production-ready.
Some of them already announced deals with IFS and I expect more would follow, for that reason. As much resources as those companies have, even they can't shoulder the burden of supporting an entire modern fab on their own.
Facebook
Twitter
Instagram