Thread regarding Intel Corp. layoffs

I think CCG and probably DCAI roadmap should include both x86 and ARM CPU's in the design. This will help fill the fab as well lessons will be learnt integrating ARM ecosystem into products.

@2txm My hope has been that Foundry would pick up enough ARM business that some Product group (or maybe a new product group) would start designing based on ARM.

The company tried this before and like a lot of other things it should have stuck with, chose to abandon ARM for more high profit x86 products.

What they should have done (as with GPU) is stuck with it and used those high profit margin x86 products to compensate for learning the best product to fit these new markets. Somebody somewhere in ELT must have read Clayton Christensen.

They could have at least spelled it out to the Board using crayons or something.

With disaggregated client architecture it should be possible to replace x86 with ARM/RISC V cores

Intel's IDM 2.0 strategy is not limited to x86 architecture, and it is flexible enough to support ARM, RISC-V, or other custom cores in its foundry services.

IDM 2.0 refers to Intel's approach of expanding its manufacturing capacity and services, including offering foundry services to third-party customers. As part of this, Intel's manufacturing process, which includes advanced nodes like Intel 7, Intel 4, and future nodes, is designed to be adaptable for a variety of processor architectures, not just x86.

Thus, Intel's foundry services can accommodate ARM, RISC-V, or even custom cores that Intel’s customers (including Intel itself) might want to incorporate into their designs. This allows customers to design chips based on these architectures and have them manufactured by Intel using its process technologies, similar to how TSMC and Samsung offer foundry services for a variety of chip architectures.

Intel itself continues to focus on x86 for its own CPUs, but its foundry services are open to a broad range of architectures.

ARM is RISC and RISC-V will overtake it in time, just as (20 year old) ARM is overtaking (40 year old) x86.

The reason is that the newer architecture is better optimized, so takes less silicon to achieve the same computation power.

A person might think 'well the chips are really small' but that doesn't matter because this is about cost and profitability and the scale of the chips being produced.

So size is less relevant for low volume and more relevant to high volume.

Intel has tried to keep x86 relevant by reducing legacy support and optimizing the cores. In both PC and Servers the chips are less 'one size fits all' than they used to be.

That would be fine if they were able to be produced on a newer node than was available (at TSMC) to those hordes of ARM design firms.

Problem is, no customer is going to switch back to a proprietary supplier once they go to one that can be replaced by other vendors.

This is why the cash cow, having left the barn, is not coming back. Intel can chase it around the field and fight all those ARM hordes to try to milk what it can.

Can RISC-V displace ARM in mobile platforms? Hey for sure no. For the same reason ARM will have difficult displacing x86 code base written over last 40 years. If Intel sc--ws up badly in executing client/server roadmap it is possible. Otherwise, I think it is not possible anytime soon.

In client SoC, the CPU die size is small compared to uncore IPs like Gfx, NPU. I think ARM will have difficult breaking x86 moat if Intel executes well. This does not even include SW compatibility issues ARM has. For sure some erosion in x86 client market share possible.

I see it as the reverse, x86 is dead the long run. You need to flush the old Intel out of your head. The only hope is cost efficient fabs even is not the leading node being a US based strategic provider.

Thanks for your assessment. The major downfall of Intel will be associated with only putting development efforts into the X86 architecture. Hiring Pat as the CEO could be looked at as a bad move given his background in and bias towards X86 based products. ARM will continue to erode the X86 market share in desktops and laptops

Pat had previously described being able to get enough external customers for Foundry to enable it to stay at least on par with TSMC and at times be ahead.

That proved to be a bit of magical thinking and likely part of why the Board pushed him out, because as it turns out getting Foundry customers takes years and years.

It might be that as TSMC raises pricing for new nodes, or due to government pressure to second source to Intel, that Foundry is able to speed up the ramp but it will still take years.

All along the way, the company has to consistently prove it can be relied upon to deliver, and still make a profit. I could be wrong on this point but thought that the contracts were for good die, which would mean that Foundry would eat yield losses.

Even if/when Foundry ramps up with external demand, that does not make x86 competitive. It doesn't make any new platform choose x86 or an open source architecture. x86 has no competitive advantage over ARM at this point.

IDM 1.0 only worked for Product groups because x86 dominated its markets, and once that domination was broken (due to the mfg process falling behind) it can't come back.

It would be nice to see Intel use the time remaining, before x86 revenue really craters, to stand up an ARM product group and get real about the path forward.

It may simply be more feasible to sell off Product groups and move forward as a foundry. That part of what Pat was after was not wrong, even if his timeline was based on party dr-gs.

@wqo well put

I can't help but think that ELT and the Board know perfectly well (or at least have had it described to them with crayons) how semiconductor strategy works.

I assume they know that even if Foundry could magically get several nodes ahead of what is available to ARM firms, that only brings x86 up to parity and that no platform is going to subject themselves to 1 (or 2 if you count AMD) supplier using a proprietary architecture when they can have far more market power being supplied countless design firms using an open architecture.

More magical thinking: The only hope for x86 irrelevance is if it suddenly is the only architecture that can power some new, must have platform.

Even NVDA won't keep dominating AI forever, for the same reason.

So I think the strategy (which they will likely never publicly state) is to extend and pretend, to slow down market share loss till the scale becomes too small to sustain.

Problem is that if PC goes ARM then that could result in a sudden, steep decline, and even before then the capital and skills to maintain product and Foundry development will fall below the critical mass needed to keep pretending.

I also think ELT and the Board are unwilling to deal with this because the path forward would result in a much smaller company, unable to afford their compensation.

Wouldn't be the first company to implode because management chose not to see what was not in their short term interest to see.