Pat Gelsinger's grand plan to reinvent Intel is in jeopardy Intel's fledgling foundry business is in trouble. The division is bleeding billions each quarter and now the chipmaker has revealed that it won't even manufacture parts on its own long-awaited 20A node. TSMC will be eating Intel Foundry's lunch through much of 2025, while Chipzilla focuses on ramping production of its …
Given that Intel / AMD enjoy de facto monopoly on the ISA, I wonder why CMA and the likes are not breathing at their necks to open up the specification for fair competition to emerge.
It's clear that Intel is no longer capable to progress and AMD cannot be left to its own devices. Regulators should step in.
But I guess CMA has more important things to do.
I'd say even that was charitable. We've *already* had Windows-on-ARM a few years ago in the form of Windows RT, and that went nowhere. In typical MS fashion they ditched it like a hot potato when it wasn't the instant success they wanted.
But more importantly here, RT was just one of countless "throwing-mud-at-the-wall" attempts at non-x86 Windows that MS released over the years, including numerous incompatible variants of Windows CE, plus Windows Phone 7 (CE-based), Windows Phone 8 (similar name, but completely incompatible NT kernel) and RT's successor Windows 10 Mobile.
All ditched and abandoned, along with any owners who were lured into trusting MS's promises.
But if we don't want to rely on MS's support (because only an idiot would at this stage), apparently Windows 10 and Windows 11 on ARM *do* support x86 through..... "software emulation". Ha ha ha, no. (Some poor mug is going to get screwed over when they buy one on the assumption it can run x86 Excel (or whatever) anyway and finds it's like wading through treacle.)
You'll forgive me for my scepticism, but in MS's case it's been fully earned. Maybe Windows on ARM *will* eventually take off.
But until I see that actually happen, I'm not wasting any mental effort on the possibility that *this* time will be different.
It's more like their third release, and it's very nice.
I've been running Windows on ARM on a Surface Pro X for about five years as my main travelling machine, and I've just replaced it with a Surface Pro 11 running a Snapdragon CPU. It's very lovely indeed, don't believe all the sneering you read from people you have barely touched the machines.
GJC
Sadly, I think Intel have long forgotten how to design CPUs. All they do is incrementally improve (or worsen) existing designs, and have done for decades.
I have known a handful of ex-Intel refugees, and they all independently told me that the company is a cult - anyone who dares have a new idea is marginalised.
This is the danger of being a semi-monopoly, or at least having enough of market share that you can get away with this sort of thing, and of being in that position thanks to a single product line.
It makes you conservative in the protection of the status quo and that product line, and it insulates you from the need to change until the ground has already shifted so fundamentally- usually due to a paradigm shift- that it's far too late for you to change.
It also isn't a healthy mentality when you want to compete in other markets where you *aren't* necessarily in that position.
Trying out some of the ideas people's stuff sounds a much better idea to me.
Intel used to make LOTS of stuff that didn't necessarily turn into product. Not dead investment, the ideas generated feed into the next round of real products.
They honestly might do better thinking about the eventual, more or less inevitable transition from X86 to ARM.
I remember intel spending most of the 90s playing stupid monopolist games and giving huge discounts to shops that would exclusively sell intel chips. Like 20-40% off wholesale, when margins on tech were 10-15%! They were literally given a new lease on life with the "core" aka revamped pentium 3 architecture after the P4's cosmetics finally wore off and you could see what a pig it was.
Exactly. They weren't suing because of a good faith prediction or estimate of success that didn't pan out.
They were suing because Intel misrepresented something that was already known to them at the time, which meant the company's perceived value and price of it's shares was higher than it would have been, with investors who bought them at that time losing out when that knowledge became public and the share price fell.
For more detail, here's a link to a previous article. And in response to someone saying "Did they not read the bit about 'stocks may fall as well as rise'", I posted:-
> Devil's advocate, but what the article claims they actually allege is that "Intel's management made "materially false and misleading statements" relating to the success of its foundry business".
> I don't know enough to tell whether that claim- and the entire legitimacy of the legal complaint it underpins- is legitimate or not.
> But assuming it is then it sounds like they're saying that Intel- or rather its management- *knowingly* made misleading statements about something they *already* knew not to be true at the time.
> That being the case, if they then bought shares at a price that was higher than it would otherwise have been- as a result of those misleadingly-positive statements- and later saw the share price fall (to what it otherwise should have been in the first place) when the truth was revealed, they could still reasonably claim to have lost money as the result of the false statements put out by the Intel execs.
> That would be significantly different to the case I suspect many here are imagining, where the fall in share price was due to a business investment or plan that simply hadn't worked out.
In the 19th century Great Britain had a law that no foreign vessels could be loaded in British ports. This ensured Britain always had a strong maritime industry, even when France could ship more cheaply.
Intel Foundry's problems would be solved overnight if the federal government banned chip imports from overseas.
Their problems would start overnight if major foreign governments banned chip imports from the US in response. Which they almost certainly would.
And that's not discounting the possibility of further actions against US companies.
The US might be powerful, but it's not so powerful that it could get away with simplistic macho fantasy moves like that. At least, not without serious consequences that would likely far outweigh the benefits.
Apologies for that; I posted that on my phone, and the titles are often- if not usually- just copies of what was being replied to anyway, so I must have missed it.
There are plenty of people around here who *would* say what you said and completely mean it, though.
No need for heavy-handed "banning". Industrial Policy using Import Duties suffice.
Of course this is still a delicate thing, because there exist powerful forces who give a rodent's backside about national or other security interests. It would also help to have a mentally capable president in charge.
Dependence on Taiwan and South Korea is a Strategic Risk that should be dealt with.
The US still has a version of this. The "Jones Act" requires that all goods transported by water between U.S. ports be carried on ships that have been constructed in the United States and that fly the U.S. flag, are owned by U.S. citizens, and are crewed by U.S. citizens and U.S. permanent residents.
I appears the problem is they are NOT making things. Spending 9 Billion operating a FAB would be OK, if it made 10 billion in product to sell. But currently it's not making new chips for Intel, and no one else seems to be buying (yet). So they are trying to sell product that they haven't worked out how to produce yet. Hence bleeding $$.
There are/will be plenty of ARM- and other ISA-based designer-manufacturers around. Apple, Amazon and the like. Apple has created an ARM-based CPU which has more single threaded horse power than any other (public) chip designer.
Nicely coded application programs run just as nicely on ARM as they run on x86. Just rent them from Amazon, Hetzner or many other cloud vendors. Or buy an Apple notebook. The latest RPI will also do for many personal server needs.
https://www.cloudpanel.io/blog/hetzner-offers-80-core-arm-based-server/
https://www.hetzner.com/press-release/arm64-cloud/
At some point investors will have to consider if Intel's leadership is misleading them materially as to the health of 18A.
I hope for Intel's sake that things aren't bad, but two major customers turning up their nose at it is not a good sign.
On the upside, Lunar Lake and Arrow Lake made on TSMC are looking strong - great adverts for TSMC's 3nm process. Probably not what Intel want to be showing off, but it's clear that they have not got the volume in their own fabs to make them.
And volume of leading edge nodes has always been Intel's problem - or rather, they would keep a node advantage in the past by ramping very very slowly in one particular product area (usually mobile) until they sorted the volume issue out after a year. It was always a fake process lead, and it fell apart eventually.
Gotta say, skipping 20A for 18A reminds me a lot of Intel's bravado in the past on mode shrinks, 10nm being one that failed spectacularly because they went in too fast.
difficult....... shock news to anyone who's thought "lets wave a magic wand and poof the products appear"
TSMC have done the hard work in making good chips, its going to take Intel the same amount of time and cold hard cash to bring their foundries upto TSMC spec... if they can. before the investors get spooked and do a runner/ make intel spin off the foundries before they can make a profit
I'll predict Intel will never be able to bring their foundries up to TSMC spec. Intel were only ever half in on the manufacturing, they lack the massive scale and decades of focused competence than TSMC have. In the last quarter, Intel made a Foundries operating loss of $2.8bn, on sales of $4.4bn. Allowing for the cost of capital (not in that operating loss) it's pretty clear that Intel were giving the Foundries product away. in the same time period TSMC sold near enough $20bn of product, and made a net profit of $7.6bn.
I can't think of any instance where a company so far behind, so uncompetitive as Intel Foundries has managed to turn that around and create a world class business of comparable scale to the market leader. Even if investors were tolerant (which Wall Street is not), Intel's culture is well known, and actively mitigates against change and doing things differently. It takes a good 7 years of persistent and intense effort to change an embedded corporate culture, but Intel clearly don't even recognise the problem.
...it is the further "horzontalization" of business. A long time ago IBM made everything from semiconductor CPUs to databases. IBM owned factories assembled everything. They also made spinning harddrives. Then came Intel, Seagate, Microsoft, Oracle, SAP and destroyed this "vertically integrated" business.
Now it seems Intel's model of vertical integration is no longer viable.
As noted before, it's time for industrial policy of the smart kind. Production must be moved to the US and to western Europe, some way or the other. Just throwing money at Intel apparently does not cut it. Of course both Biden and vdLeyen are rather clueless folks. They cannot think of an IT Airbus.
Intel is right to focus on high-end nodes like 18A instead of 20A. Everybody wants the highest density so they can claim the most powerful AI-chip to please the stock market and to up their share price. A substandard process simply doesn't cut it.
In the end there's really no choice but to continue no matter what investors say since it's too dangerous for the U.S. and the West to rely on a firm that's located on a disputed island which may get invaded almost any minute. Intel must march on and get it right, no matter how much money needs to be spent.
If Intel can get nose to nose with TSMC and China starts making the wrong noises Intel's profits will simply explode beyond anything imaginable. Right now may be a good time to invest in Intel.
Gelsinger knows what he's doing. By splitting the products division from the foundry, Gelsinger set up Intel Foundry for one of two outcomes: (1) It thrives. (2) It fails, in which case Uncle Sam will certainly bail it out. So it's heads Intel wins, tails taxpayers lose. It's how Wall Street never loses.
Cardboard box? You were lucky. We lived for three months in a rolled-up newspaper in a septic tank. We used to have to get up at six in the morning, clean the newspaper, eat a crust of stale bread, go to work down t' mill, fourteen hours a day, week-in week-out, for sixpence a week, and when we got home our Dad would thrash us to sleep wi' his belt.
This post has been deleted by its author
(Apart from China and Taiwan) where is anyone investing in innovation in uP manufacture?
Maybe someone needs to do a Elon Musk approach to this and ignore every expert in the old optimum in order to find a new optimum.
Germanium? Cryogenic? Direct optics? 3D printed substrate? Perovskite?
(I'd be Interested if you can add more off-the-wall tech for this)
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
