Intel’s entire turnaround plan hinges on this one new chip family – Clearwater Forest pictured, Intel’s first 18A chip slated for high-volume manufacturing
Posted by imaginary_num6er@reddit | hardware | View on Reddit | 132 comments
dc_chilling17@reddit
Hilarious article.
Intel looks super competitive across the board now on products and improving rapidly.
There is a good chance Pats vision comes true and they hold both the design and performance crown again in coming years.
They have a massive amount of cash coming in: 36B of chips act grants + 11B of low cost loans + 2.1B from Ohio + 16B from private equity…
Unfortunately nobody that writes these articles read 10-Qs and 10-Ks. Intel is much healthier and the street would have you believe.
_Cracken@reddit
Is this what a "super competitive Intel" looks like for you ??
Intel is not leading in anything at all other than their bleeding market share.
In servers, they finally caught up to AMD in core count, but they are still loosing to AMD.
In desktop, their brand new Arrow Lake is loosing badly to Ryzen 9000 series and intels own prior generation CPU's. Arrow lake is getting horrible reviews across the board.
Don't even look into the shitshow of Intel discrete graphics Pat's "vision" was that by 2025 AMD would be in rearview mirror, it's not happening cause the main products for 2025 are already out, and they are not cutting it.
Although intel has been doing bad for nearly a decade now, and still is, i don't believe in the doom and gloom. Crazy to see Intel made more revenue in 2016 than they do today.
i believe we haven't seen the worst yet. Ask yourself this when you take a look at intel's roadmap and compare it to the competition: is there anything on there that indicates intel will take back leadership ? Also ask yourself: in recent decade who has been best at actually executing their roadmap, intel or the competition ?
clingbat@reddit
I don't know why everyone is hyped about Intel's 18A when TSMC's 16A isn't that far behind and they have a much better track record execution wise over the past decade than Intel. It would be so Intel to spend a ton on the latest machines from ASML and still fuck it all up in the end honestly.
theQuandary@reddit
Professional gamblers don't always win. Sometimes they have very long losing streaks. The real point is that they win more than they lose.
Intel has spent most of its history ahead of everyone else by 1-2 nodes. Unfortunately, discovery is a dice roll. Sometimes you find what you want very quickly and sometimes it takes a very long time.
TSMC was a year or so behind schedule with N3 and seems to have lost steam on N2 and 16A.
If the TSMC dice rolls continue to go bad and Intel's continue to go well, Intel stands a very good chance of taking the lead on fab tech.
It's too early to tell without insider info and certainly too soon to bet everything on either Intel or TSMC.
soggybiscuit93@reddit
16A is way further behind. Not worth comparing.
clingbat@reddit
A year isn't way further... And that's IF Intel stays on schedule.
soggybiscuit93@reddit
If roadmaps are followed, A16 is closer to the release of 14A than it is to 18A
Hendeith@reddit
People are hyped up about 18A, because it's supposed to be first competitive node from Intel since TSMC released N7 and also first mass production node that uses GAA transistors and BPD.
TSMC A16 is not nearly as interesting and it's planned for 2026, same year that Intel plans 14A for - which in turn would be world first High NA EUV node.
AceVenturaHoldingsCo@reddit
Just DM’d you
XYHopGuy@reddit
Intel has struggled to execute. If 18a fails they will bleed money, stock price will tank, employees will leave and the company will slowly unwind. Not sure if call that healthy but they're unlikely to run out of money in the near term.
Their TSMC based CPUs do seem nice.
DYMAXIONman@reddit
I think the doom is overblown somewhat. If their foundry remains useful for 3rd parties, it will still generate revenue. What their design team is now doing is no different than AMD. All they will need to do is choose the best node for their new products.
_Cracken@reddit
They just did that with Arrow Lake. It's on a better and more expensive TSMC node than AMD, and yet they loose to AMD, not by a little, but by a lot.
Invest0rnoob1@reddit
They have Microsoft and Amazon as foundry customers
wonder_bro@reddit
The AWS deal is not a “Foundry win” in the true sense. It is still a chip designed and built by Intel for AWS. However it is significant in that AWS has shown confidence in 18A technology
Strazdas1@reddit
If it keeps the foundry busy and intel profitable, does it really matter whose designing it?
wonder_bro@reddit
It does to an extent because a foundry charges customers by Wafer starts per week or month but in this deal where Intel is basically selling the final product they can only price the final product so they might lose out on revenue on bad dies. A pure play foundry can charge customers for hot lots or the velocity of running wafers which in this case will come from Intel Foundry itself rather than the customer. Intel cannot also charge AWS for any design flaws or Litho/Reticle stepping revisions if the design is done internally
Exist50@reddit
They do not have Amazon as a foundry customer. AWS is buying a product from Intel's networking group, that happens to be planned for 18A.
Invest0rnoob1@reddit
What do you think 18A is? 🤔
Exist50@reddit
A node also used by Intel's internal teams... That's a win for Intel design, not foundry. At least no more than Panther Lake is.
winniedemon@reddit
Given that Intel design can choose to go to TSMC, any manufacturing on 18A is technically still a win for Intel foundry.
Exist50@reddit
In theory, yes, in practice, not really. They needed half a decade of foundry failure, teeth pulling, and the intervention of Jim Keller to even do ARL/LNL. And they wanted SRF on N3 as well.
ExeusV@reddit
Any source on Keller impact on LNL?
Exist50@reddit
Keller did a few things. 1) Provide the political push for Intel to go to external foundries, 2) "Motivate" P-core to adopt a 21st century design methodology, and 3) Push Atom into the spotlight/make Intel adopt hybrid.
None of those are exclusive to LNL, nor are they a direct influence, but there's some credit that's due.
ExeusV@reddit
How do you know?
ResponsibleJudge3172@reddit
Everything good ever is blamed on Keller, who left too long ago for me to take the claim on Intel seriously. Like blaming him if Zen 6 succeeds
BoltTusk@reddit
Hopefully better than Samsung 2nm
thegammaray@reddit
From Intel's press release: "As part of the expanded collaboration, Intel will produce an AI fabric chip for AWS on Intel 18A, the company’s most advanced process node."
Obviously "AI fabric chip" is a vague term, but do you have a specific reason to think it's an off-the-shelf part from Intel's networking group? If so, what product is it?
Exist50@reddit
Yes, note the wording. "Intel will produce" is not the same meaning as "AWS is designing on". For example, try this substitution: "Intel will produce a client PC chip for Dell on Intel 18A". They're being rather deliberately vague, just as they were for the Ericsson design win.
One yet to be released. We're talking 2026+ year. And it's either off the shelf, or semi-custom, but most likely a relationship like Intel's IPU/DPU and Google. Single definitional customer, but Intel-designed and ultimately intended for a broader market.
thegammaray@reddit
What you've said about the wording makes sense w/r/t who's designing the chip, but it doesn't support the hard distinction you're trying to make:
Why not both? AWS is no stranger to piecemealing different suppliers' products together or to custom chips manufactured elsewhere than Intel, so if Intel's design manufactured on Intel 18A is an attractive product to AWS, then AWS must have some degree of confidence in 18A. And if Intel 18A is manufacturing volume for AWS that 18A wouldn't have been manufacturing otherwise, then the fact that AWS's money is reaching the foundry by way of the design teams doesn't make the money any less green to the foundry, right?
The underlying question is: Will 18A be an acceptable/workable alternative to whichever TSMC process (different for different products, obviously) would be used for any given chip? The AWS collaboration seems to be evidence for a positive answer, regardless of which Intel division has the AWS contract.
Exist50@reddit
That's the only context in which the term "foundry customer" has any meaning. Intel themselves say they have 4-ish foundry customers. If they were including anyone buying any Intel product, they'd have hundreds of companies, and millions of consumers! You and I could be "Intel Foundry customers"! Obviously, that's not how people are using the term.
If that's how you're looking at it, it's true to an extent, but also pretty redundant. Panther Lake will be more complex and coming earlier, so what would this chip tell us that Panther Lake doesn't already? Obviously, we know Intel's client customers are also working under the assumption it will reach market.
There's also the fact that companies have always planned around Intel's stated roadmap. They did for 10nm, Intel 4, etc. Didn't stop those nodes from having trouble.
Yes, at the end of the day, money is money, but then why would Intel themselves try to make such a big deal of getting external customers? Answer: because it's a proxy for the free-market competitiveness of Intel Foundry. Right now, no one willingly designs on Intel nodes by choice, because they're too difficult to work with, too expensive, poor roadmap execution, etc. Their own design teams have been fleeing as much as they're allowed to (PCH starting with ADL-S, MTL's SoC/IO, ARL/LNL compute, entire GPU roadmap...), and what remains they get heavily compensated for post-financial split. This is reflected in Intel Foundry's $7B operating loss.
So the market wants to see Intel Foundry get profitable, non-coerced design wins, because that would indicate much greater confidence in the nodes, and thus a path to profitability on both internal and external deals.
thegammaray@reddit
I'm not saying AWS is a direct foundry customer. I'm saying the AWS contract can still be a win for Intel's foundry.
Laptop OEMs are generally at the mercy of chip design companies, but AWS isn't; if AWS needed a different product, they'd go get that one instead. Since Panther Lake samples aren't publicly available for analysis, the fact that AWS -- which presumably has access to early manufacturing samples/metrics -- has contracted AI fabric chips built on Intel 18A provides us with a data point that the mere existence of Panther Lake currently doesn't.
I can't speak to what motivates Intel's PR strategy, but from a business model standpoint, the foundry just needs to bring in enough cash to subsidize the R&D and operating costs. Whether that cash comes in from external customers directly or from Intel's design teams through contracts with the design teams' external customers isn't a crucial distinction. In fact, if the foundry is being paid to manufacture Chip X regardless, would Intel prefer that Chip X was designed by an external customer, or by Intel's internal design teams? The second one means that the design team is making money on top of the foundry's manufacturing revenue. (Of course, the foundry is likely charging Intel's design teams a lower rate than they'd charge external customers, but the margins are higher in design anyways.) Either way, money is money for the foundry.
The fact that Intel's playing catch-up to TSMC is the whole point of the article at the top of this thread. Intel's server CPUs and consumer CPUs moving forward are both set to be manufactured on 18A. Again, that's positive evidence for 18A, not negative.
Exist50@reddit
Sure, but that's not the claim that I responded to above.
The main alternative is going all-in on Nvidia. It's worth it for AWS to take a gamble. They're not betting their business on Intel.
Again, it's a plan for a product order, and for something that hasn't even taped out yet. They won't get manufacturing details for that, though they will get samples once the chip is out of the fab.
Correct, but since they split, the same economic incentives imply. A node that can sell for a profit to 3rd parties is also one that can be sold within Intel.
Yup, Intel would definitely prefer it to be completely internal, all else equal. But that doesn't show the world that they have a working node.
It's...a bit more complicated than that, actually.
thegammaray@reddit
Obviously they're not betting their entire business on Intel, but what would they be gambling with if not the fruits of their business... They're betting money, and they're betting it on Intel.
AWS is absolutely not entering into this contract without an inside track on Intel's 18A manufacturing progress. It's a red herring to suggest that AWS's decision doesn't reflect that data simply because AWS hasn't received samples of the specific AI fabric chip they've contracted.
Apologies, but I can't figure out what point you're trying to make here.
...Why not? Intel's already showed everyone that Intel's fine with contracting TSMC for manufacturing if Intel's own manufacturing isn't competitive, so from where I'm standing, their decision to use their own manufacturing is evidence that they have a working node.
How so?
Exist50@reddit
They're not betting anything they can't afford to lose, or don't have alternatives for. It's certainly much better for Intel than no deal at all, but not as much a vote of confidence as it's being sold as.
Of course they would, so long as the contract allows them to back out, or even get a breakup fee, if Intel fails to deliver. It would be highly unusual for a fab to disclose yield numbers to customers of their customers.
Basically, if a node is good enough to want, it's good enough to make a profit on.
There's still a bias to internal nodes. While TSMC is an option, it requires a really compelling benefit for the business to justify it, and that isn't the case here.
TSMC has and will maintain a competitive advantage over Intel nodes. So long as that gap remains, there will be strong business incentives to make products at TSMC. That's all I'll say on the matter.
StoneFlowers1969@reddit
You are not informed. This was a design win for IFS, theres a design team in IFS that works on customer custom designs. This has nothing to do with Intel Product side.
Exist50@reddit
Everything you said in this comment is wrong.
StoneFlowers1969@reddit
I work in that group actually and everything I said is correct. CCG is not designing these chips.
Exist50@reddit
NEX, not CCG. And if it's a foundry win, why doesn't Intel state outright?
StoneFlowers1969@reddit
This will be my last reply to you clown not NEX nor CCG will be taking this design through tapeout that responsibility will fall under IFS.
Exist50@reddit
Remains? They've been struggling to get any 3rd party traction. And even in their optimistic scenario (which hasn't been playing out), the vast majority of Intel Foundry's volume would be from Intel's own design teams.
Asleep_Holiday_1640@reddit
They have been struggling to land 3rd party traction because 18A is not yet HVM ready and there aren't yet 18A products to prove the silicon is healthy.
When that happens as soon as say next year, you will see them land more customers.
It is important to put things in context.
Helpdesk_Guy@reddit
They have been helplessly trying to establish any viable foundry-business for over a decade now (2011-2024). It never materialized, and they never really could hold onto any substantial contracts nor foundry-customers.
So what makes you think, that 18Å is going to be any different? Honest question though!
That has been the hope-fueled yet pretty much futile go-to answer ever since, you know …
Exactly! Intel failed on everything foundry for over a decade straight! Their failure to acquire any customers is literally 'nothing new'.
Exist50@reddit
So that's late '25-early '26. At which point the competitive node from TSMC (N3) will have been on the market for how long? And if they start designing then, that means we wouldn't see 18A products till realistically '28 or later, at which point it'll be rather outdated. That strategy is not going to win major orders.
Basically just Microsoft so far. Assuming that deal holds. Who else are you thinking of?
Puzzleheaded_Fox3546@reddit
TSMC's node that's competitive with 18a is N2, which is going volume production in 2025-2026, so the same timeframe. Why would people only start designing afterwards? That begins a long time in advance. Potential customers are already evaluating it today.
Exist50@reddit
No, it's N3. Note how Intel's using N3 for Panther Lake graphics, and more tellingly, their own 2026 AI flagship.
The comment above proposed that customers are waiting to see 18A in the wild before adopting it. So the start line would be PTL/CWF's release. And that claim has merit. Anyone starting now is taking a bigger risk, which is why the uptake has been so tepid.
Puzzleheaded_Fox3546@reddit
Why would they release a node that only competes with TSMC's current node (whose successor is being released next year)? That doesn't make any sense at all. The whole point of dropping 20A was to focus on a leading edge node.
Because of limited capacity. Which is a different issue. If 18a only competed with N3, there'd be no point in using it at all for Panther Lake when they could use N2 instead. They aren't because they will have 18a.
DYMAXIONman@reddit
Rumor is that 18a will beat n3 but lose to their n2 in density. It's worth remembering that when TSMC comes out with a new node most manufacturers aren't using it. RDNA 4, Blackwell, Battlemage, and Zen 5 all used/using TSMC 4nm.
clingbat@reddit
This is because Apple has bought out capacity of TSMC's latest node in 1 year exclusivity agreements for the their last two nodes...
It's not for lack of desire, but rather Apple locking everyone out with their deep cash reserves.
Puzzleheaded_Fox3546@reddit
That's usually because Apple has bought out most of the capacity and cost because of the competitiveness around securing capacity with TSMC's latest node.
Exist50@reddit
Because that's the best they have, of course. What else can they do? At minimum, it's more competitive than Intel 3 vs N3.
The point in dropping 20A is that it could not fulfill its role as a demonstration vehicle for the health of the fabs. And Intel doesn't have the money for such a pointless PR exercise.
Lol, it has nothing to do with capacity. I have no idea where this belief came from. If they could move it in-house, they absolutely would.
PTL could have gone all-in on N3, but they didn't for the same reason that MTL used any Intel 4 at all. The loss of foundry volume, higher costs, and effort to make the design pivot would hurt Intel as a whole more than the benefit to the product. Remember, Intel based their planning on much more aggressive PnP and schedule projections.
ARL was compute on either N3B (816 or 68) or 20A (68). PTL is, iirc, 18A for compute/SoC, N6 for IO, and N3E for bigger iGPU. Also, might have a lower end GPU on an Intel node (Intel 3?), but not 100% on that.
Puzzleheaded_Fox3546@reddit
You're wrong on the process nodes for ARL and PTL unless you can give me a source saying otherwise.
20a capacity was always going to be limited, because of yield, cost, etc, which is why N3 was being tapped for the iGPU of ARL, which again was cancelled with 20a.
Sorry, this is ridiculous.
Exist50@reddit
Well you can wait for PTL to release and see for yourself. As for ARL/LNL, I called that at launch.
Wait, what? You think ARL as a whole is canceled or something? First, N3 hasn't been planned for the ARL GPU for a very long time. Its purpose was, and remains, the primary node for the CPU tile. ARL will launch with an 8+16 LNC+SKT CPU on N3B.
It's what they've done with every node since 14nm...
aminorityofone@reddit
Time will tell. Keep in mind that people have been saying intel will get back on their feet for a decade now. It has yet to happen and things just keep getting worse. Such as the 13th and 14th gen fiasco
Objective_Today_2962@reddit
Remains? They haven't produced a single chip as a boundary and basically only the US government has signed onto a firm commitment, but that's for political reasons.
imaginary_num6er@reddit (OP)
DarkGhostHunter@reddit
It sounds really grim for Intel, but that’s what happens when you become complacent for a decade. ARM is mature, RISCV is rising in applications, AMD is soaring, and TSMC is the king (and recently expanded to US soil).
The missed mobile. They missed cryptoboom. They missed AI. And still have fans to maintain.
I wonder what would happen if 18A fails. Will NVIDIA buy them? Or even AMD? May be the saudis? Or the Chinese (and get blocked to hell)?
WJMazepas@reddit
They are far away to being sold. They still are the kings in laptop sales and a lot of their parts still use their foundry.
BlurredSight@reddit
Also patents, DOD contracts, and smaller tech even as simple as WiFi chips
RTukka@reddit
All of which is already baked into Intel's stock price. If 18A runs into serious problems, you can be sure the price will drop further.
Already Intel's market cap is only about half of Qualcomm's. AMD is at about 2.5 times Intel's value. And Nvidia? They're worth 30+ Intels.
I don't think governments would let Nvidia or AMD buy Intel even if they were inclined to do so. They blocked Nvidia from buying Arm, after all. But just because Intel still has cash flow and valuable assets doesn't mean they can't get bought, or go bankrupt.
Digital_warrior007@reddit
It's very unlikely that 18A should run into some serious problems at this point. 18A defect densities are below that level. It's at the final flow flush stage for high volume manufacturing. But I have serious doubts about the volume they can achieve considering they also cater to some external customers, including Microsoft and Amazon. As of now, I think they just have one single fab that can manufacture 18A wafer.
onlyslightlybiased@reddit
You can sell as many laptops as you want, you have to actually be making good margin on them though.
poopysprinkles@reddit
The US needs a domestic semiconductor supply chain for defense reasons. I believe Intel would be bailed out by taxpayers before being allowed to truly fail, and certainly before being bought by a foreign interest.
onlyslightlybiased@reddit
Realistically though, Intels manufacturing is just one part in a huge supply chain for semi conductors, the US is nowhere near ready and probably won't be for another 10 years at least.
poopysprinkles@reddit
Yea and I'm not saying otherwise, I'm just saying that allowing Intel to fail or be sold to the Saudis or whatever does not advance those efforts.
HCharlesB@reddit
I think they're already producing parts that go into defense so this is the likely outcome if they continue to stumble.
NoobFace@reddit
They're only kings in laptop sales because AMD hasn't been investing in their laptop channel at a level that'll let them satisfy current demand.
If OEMs cant' get x86 chips because Intel has foot gun syndrome and AMD doesn't have the capacity...I'm imagining we're about to see a metric fuck ton of ARM options from the big guys.
WJMazepas@reddit
They still are Kings. In my country, I can find some laptops with a 5700U or gaming laptops with something from AMD. Otherwise, it's all Intel.
All companies are basically buying Dell laptops with Intel here.
AMD has to improve their laptop game by a lot to become the leader in that market
horrorwood@reddit
It was the same in the UK, but within the past year I am seeing more and more AMD laptops.
theQuandary@reddit
Intel didn't miss mobile. They launched Atom in 2008. Mobile is more simple, but they were probably starting design around 2004 which was a full 3 years before the iPhone changed the game.
The problem was x86.
Intel didn't miss the cryptoboom. They invested heavily in Larabee/Xeon Phi. The problem was that they couldn't make the cores small enough to be competitive with Nvidia/AMD. It was after the failure of Phi that they switched to a traditional GPU. This doesn't seem to be a problem with the Larabee design though as most of the new AI startups are using a similar design with RISC-V.
The problem was x86.
Intel did miss the start of the AI boom, but that's because they pivoted to a new GPU ISA too late because they've bet far too much on x86 and the Wintel monopoly.
Nointies@reddit
If 18a fails fabs get spun off.
Exist50@reddit
Realistically, they get shut down and sold for parts/real estate. There'd be no one to sell to.
JDragon@reddit
Besides American taxpayers, given how Raimondo has been handling things.
falcongsr@reddit
She's right though...we can't rely on foreign entities for advanced nodes.
JDragon@reddit
We can't rely on Intel either, apparently!
(I'm just being flippant, if 18A fails Foundry being spun off into a government-funded entity is probably the best outcome for both Intel and "national security" interests.)
falcongsr@reddit
I'd bet on it happening, especially if the office of the POTUS doesn't switch parties.
Visionioso@reddit
That’s not what happened. Taiwan doesn’t fund zombies. They saw a well performing company and gave them all the support they could to let them perform better. They never supported zombie companies hoping thongs would turn around.
Strazdas1@reddit
TSMC was never a zombie, but it got funded by government long before it was a leading node manufacturer.
falcongsr@reddit
I know. It's too late now.
Strazdas1@reddit
Not just TSMC. All 3 of major Intel competitors are strongly supported by the state both monetarely and legally (by adapting laws).
Exist50@reddit
I mean, it's worked perfectly well so far...
Nointies@reddit
Well yeah, after they get spun off of course.
jaaval@reddit
What gets spun off? 18a is really the first marketable product. If it fails they will have to wait for several years before there is anything to sell.
Nointies@reddit
Yeah thats why they get spun off into their own company which then just fails.
jaaval@reddit
That's not how anything works. "Spun off" to where?
aminorityofone@reddit
Spun off as split. Just like how AMD split from their own foundry business.
jaaval@reddit
So by selling it to UAE state investment company?
Shikadi297@reddit
Well yeah basically
Nointies@reddit
That is in fact, how it works. Spinning off the fabs to be their own company is entirely possible.
it might still 'own' it, but it wouldn't be a part of the intel core business, they could then be valued seperately.
jaaval@reddit
If they own it it is part of their business and they will be legally required to report it as part of their financial results.
upvotesthenrages@reddit
No, that's not how that works, at all.
MS own part of Apple, that doesn't mean they include that % of Apple revenue in their quarterly reports.
A subsidiary might be owned by Intel, but legally it's its own entity and does its own reporting. Similarly, it can be sold off or get investors just for that part of the business.
jaaval@reddit
The key there is “part”. A partly owned company is not a subsidiary but an investment. When they own more than half (meaning they get to make all the decisions) it becomes a subsidiary.
Both gaap and ifrs rules require subsidiaries to be reported in the consolidated report of the parent company.
upvotesthenrages@reddit
True.
I assumed OP, by saying "spun off", meant that it would be sold, either partially or wholly.
jaaval@reddit
The problem we were talking about was that there is nobody to buy it unless they can make it work.
Shikadi297@reddit
It will be like global foundries most likely
ThankGodImBipolar@reddit
Intel 3 seems like a pretty good product based off Granite Rapids reviews today (not sure that Intel is taking many orders for that process though).
Exist50@reddit
It's basically a more expensive N5/N4 competitor. Which is way better than they were on Intel 7, but not enough to be worth using for any significant 3rd party volume. At least until they have a lot more robust analog IP portfolio.
edparadox@reddit
It's worse than that.
When Ryzen came out, which was absolutely not a secret, since AMD bet the whole company, everyone and their mother knew they were releasing something, the did not and/or could not really compete properly over the course of 7 years while they were the uncontested king the previous 10-15 years.
Moreover, they were the only ones not being fabless and now, not only they never use this as leverage, apart from the "confidential aspect", it's their make or break moment?
For now, with all the cash they have, even if their fabrication capabilities are not good, they could even go fabless.
The thing that worries me, is that, very much like their CPU architecture, their fabrication tech is not up to the task,
aminorityofone@reddit
But reddit is convinced that Intel is fine and all the news stories about buy outs and failings are nothing to worry about.
callanrocks@reddit
AMD buys them and opens up x86-64 to avoid the inevitable antitrust suit.
king_of_the_potato_p@reddit
Amd wouldn't be allowed to buy them, it would fall under monopoly for desktop.
the_dude_that_faps@reddit
The doom and gloom over Intel is mind boggling to me. AMD was in muchuch deeper shit before Zen launched AMD they basically had no assets to pay their debts with. They had sold fabs, they sold Radeon mobile. They even sold their offices and then leased them to just have that extra bit of cash.
Intel in comparison has over 100 billion in assets which it can use to further leverage themselves. They still have a lot of revenue. It comes down to spending and efficiency.
They will get there. It's not a matter of if, but rather a matter of when. Who I'm honestly worried more about is AMD. While their designs are good in general, they have a tough time outcompeting their rivals I'm everything else that matters. And margins are looking very bad.
Visionioso@reddit
No one doubts Intel’s design team. The problem is the foundry, it’s expensive as hell as it’s underperforming. And since Intel is an IDM the foundry’s shit nodes are hurting the design team too.
theQuandary@reddit
I'm the exact opposite.
A lot of Intel's historic perf/watt advantage from the design team actually came from being 1-2 nodes ahead of the competition.
Today, AMD matches Intel in perf/watt. Qualcom has surpassed Intel in perf/watt. Apple blew past Intel 4 years ago.
But the fabs are different. Intel's fab engineers are amazing. Every fab you can name has had bad nodes or even had multiple bad nodes in a row (looking at you Samsung). The only question is if you can recover and stay in the game. Intel seems to be recovering rapidly and may even retake their lead again.
Strazdas1@reddit
AMD would have filed for bancrupcy if Zen 1 wasnt as wonderful as it turned otu to be. They had to sell their foundries to stay afloat long enough to finish Zen 1.
the_dude_that_faps@reddit
They sold the fabs way before Zen was even in anyone's mind. The fabs were sold under the leadership of Héctor Ruiz in 2009. Lisa Su was a few years away still from taking leadership and steering the company back towards high performance CPUs.
aminorityofone@reddit
Ah yes, this nearly 10 year old argument. When is When? Its been nothing been issue after issue and delay after delay. Altera is being spun off, mass lay offs, the stock option changes, canceled their shuttle program and they even canceled the free food.
kotori_mkii@reddit
So the exact same things as every other tech company? Cause they're nearly all cost cutting right now. Guess we just won't have technology in a few years.
Strazdas1@reddit
Better find some cave real estate as we go back to pre-industrial world /s
the_dude_that_faps@reddit
Nothing of what you said is a sign that they are dead or contradicts what I said. Spending needs to be controlled, which is where layoffs are coming from. They are still producing designs at a breakneck speed in comparison to AMD. Hell, Meteorlake isn't even a year old and we're getting Lunarlake, Arrowlake-H and Arrowlake-S. And let's not even talk about servers with granite Rapids now a few months after Sierra Forest.
They are iterating very rapidly. All signs of them not running out of steam at all.
Look at it from this perspective, if Intel sold their fabs tomorrow, what's stopping then fro continuing to execute? Are they in such a bad place that their products are inpurchaseable? Their client division sold more than all AMD combined last quarter. And that includes datacenter where AMD is stomping Intel.
Exist50@reddit
I'm not sure if you remember those times, but the doom and gloom was way worse for AMD then. Hell, people didn't even really talk about them much outside of graphics.
They're leveraging that a lot already.
einmaldrin_alleshin@reddit
AMD had a 50% gross margin last year. However, they reinvested all of their profits, so the net profit is barely positive
Objective_Today_2962@reddit
What a dumb article. Everyone (Intel CEO included) knows if 18A fails Intel is heading for bankruptcy.
WJMazepas@reddit
Oh, so they have been operating at losses for many years now?
I didn't knew that
monocasa@reddit
They've been shedding projects left and right to stay afloat, but they don't have much left there to grab extra budget from. Their new data center chips (which are where the margins are that keep them afloat) looks like they're going to be outclassed just a month later when Epyc Turin hits the streets.
The amount of capital being juggled is as much a liability as it is a benefit. It can all come crashing down very, very quickly.
jaaval@reddit
Intel has plenty of money and is in no risk of going bankrupt for the foreseeable future.
monocasa@reddit
They wouldn't be shedding everything not related to core business, and publicly stating that the whole company is riding on 18A's success if that were actually true.
You can very quickly go from billions in the bank to billions in the red when actually investing that capital in the business.
That's why leading edge foundaries have essentially dried up to places where the host government sees value in propping them up, and not always successfully at that.
jaaval@reddit
They are a public company and their financials are public. What they are doing is slowing down spending, that is very very very far from any risk of going bankrupt.
Helpdesk_Guy@reddit
If you honestly think, just because some company is publicly listed, its financials are clear for everyone to see and get deep insight into, you outright delusional.
Intel has been cooking their books with everlasting cross-financing and their industry's finest Financial engineering™ since time immortal – That's how they could conceal their extremely costy and loss-making side-shows such as Optane with several Billion US$ lossess, their Atom disasters with $12-15Bn or their ARC-disaster with also several Billion US$ in losses ever since.
Also, that's how they to this day conceal their sneaky rebate- and incentive-programs towards their OEMs. All that is NOT publicly known. Same as General Electric and others cooked their books ever since, despite being publicly listed …
monocasa@reddit
Their public financials are a good chunk of what I'm basing my statements on.
For instance, their net income trend is 2021: $19.9B, 2022: $8.0B, 2023: $1.7B. That is not a good trend line. And that's with shedding ~$3B or around 15% in operating expenses in 2023, otherwise they'd already be losing money. They have cash reserves, but that money is already pretty heavily leveraged. Businesses don't just keep billions in a bank account; it's collateral for loans. That house of cards falls apart real quick.
One-End1795@reddit
So how is it dumb, then? That is the same thing the article says....
Objective_Today_2962@reddit
Because we don't need an article to say 1+1=2.
One-End1795@reddit
Oh, because EVERYONE ON THE PLANET already knows the ins and outs of the Intel situation. Yeah, I should've remembered that they teach that in school everyday, and that everyone reads every single article on the state of Intel the second they are published. Oh, and listen to the earnings calls, too, and read the earnings report. Man, the author IS stupid. /s
Objective_Today_2962@reddit
The intel situation has been front page news around tge world recently. It certainly isn't limited to some niche audience. Anyone on a computer hardware site absolutely should be aware.
One-End1795@reddit
Yeah, and I am sure everyone knows that Intel showed the chip for the first time at a private invite-only event held last week. Of course, they do; they know this because they were born with that knowledge. Your argument is not defensible; please stop digging.
Objective_Today_2962@reddit
The specific chip they show is irrelevant; it's the process itself that matters. Intel also showed 20A ARL chips and we see how that went..
Fast_Wafer4095@reddit
Maybe he tried to say it is not really news since it is telling us what we were already aware of?
Astigi@reddit
Intel 18A is their last stand, if not a TSMC competitive alternative, IFS is finished
PMARC14@reddit
Idk doesn't even have to be a truly TSMC competitive node alternative just has to do a bit better than Samsung.
ResponsibleJudge3172@reddit
It doesn't even have to be better than Samsung. Just good performance vs cost and volume. Volume is what Intel is lacking lookingg at Intel's own charts
Geddagod@reddit
Intel released expected capacity charts a while back, maybe close to a year ago, and I think it was a bit shocking how little advanced node volume they appeared to have. Intel 7 was their highest volume node until mid 2026. With the pausing and delays of many of their fab plans, it appears as if the situation is only going to get worse.
SteakandChickenMan@reddit
Only surprising if you don’t know how they do product ramps.
wrestlethewalrus@reddit
„If intel can‘t sell it‘s product, it is DOOMED.“
This is the kind of insight I come here for.