←back to thread

US Intel

(stratechery.com)
539 points maguay | 9 comments | | HN request time: 0.001s | source | bottom
Show context
themgt ◴[] No.45026515[source]
I’ll be honest: there is a very good chance this won’t work .... At the same time, the China concerns are real, Intel Foundry needs a guarantee of existence to even court customers, and there really is no coming back from an exit. There won’t be a startup to fill Intel’s place. The U.S. will be completely dependent on foreign companies for the most important products on earth, and while everything may seem fine for the next five, ten, or even fifteen years, the seeds of that failure will eventually sprout, just like those 2007 seeds sprouted for Intel over the last couple of years. The only difference is that the repercussions of this failure will be catastrophic not for the U.S.’s leading semiconductor company, but for the U.S. itself.

Very well argued. It's such a stunning dereliction the US let things get to this point. We were doing the "pivot to Asia" over a decade ago but no one thought to find TSMC on a map and ask whether Intel was driving itself into the dirt? "For want of a nail the kingdom was lost" but in this case the nail is like your entire metallurgical industry outsourced to the territory you plan on fighting over.

replies(19): >>45026609 #>>45026778 #>>45026847 #>>45027040 #>>45027203 #>>45027671 #>>45028085 #>>45028186 #>>45029665 #>>45029679 #>>45030185 #>>45031538 #>>45032843 #>>45034153 #>>45034357 #>>45034925 #>>45035444 #>>45035539 #>>45037189 #
wat10000 ◴[] No.45027040[source]
Intel was the best until fairly recently. Then they still looked like the best to a non-expert observer, and then still looked at least competitive until even more recently. The modern world changes too fast for our governments to adapt to. Especially when we're talking about state of the art semiconductors and our leader was born before the invention of the transistor.
replies(1): >>45027160 #
scarface_74 ◴[] No.45027160[source]
Intel hasn’t been “the best” since the world cared more about mobile in 2010. There GPUs have always been also ran. It just wasn’t a big deal until crypto and later machine learning.

Even for integrated graphics, Intel has been behind Apple’s/TSMC ARM based processor before the Mx based Macs.

replies(5): >>45027338 #>>45027401 #>>45027704 #>>45034308 #>>45035357 #
qwytw ◴[] No.45027704[source]
Well on desktop and server they pretty much had no competition until the late 2010s or so. So they were the best by default.
replies(2): >>45027916 #>>45028105 #
1. wat10000 ◴[] No.45027916[source]
And it took a long time for mobile CPUs to be considered important. Arguably mobile CPUs still aren't considered more important than desktop, even though they obviously sell more.
replies(1): >>45027993 #
2. qwytw ◴[] No.45027993[source]
Mobile CPUs are almost a commodity these days and fairly low margin especially compared to server chips. Most people really don't care what chip does their phone has and its almost always "good enough" relative to the price.

IMO that's much less of a case for laptop and desktop (let alone server). Even if people don't understand the technical details e.g. Apple's superior performance per watt (or its implications at least) is something a lot more people notice.

replies(2): >>45028050 #>>45034379 #
3. JustExAWS ◴[] No.45028050[source]
This is demonstrably not true. TSMC is ahead because of the volume of mobile and that happened on the back of a lot of investment from Apple - who does make high end chips.

Intel focus on low volume high end chips is another reason they are behind.

replies(1): >>45028192 #
4. qwytw ◴[] No.45028192{3}[source]
> This is demonstrably not true.

In what way?

TSMC doesen't design or sell the chips. If they have limited capacity they will of course charge more for manufacturing mobile chips if they can sell the capacity to Nvidia/AMD/Apple instead.

ARM chips (and that's pretty much by design based on ARM's business model) are close to being a commodity.

Apple is of course an exception but they are not directly part of the CPU market. And ARM and Qualcomm are barely bothering trying to compete with them because there doesen't seem to be a lot of point. They themselves are pivoting to datacenter because there is just more money to be made there.

> Intel focus on low volume high end chips is another reason they are behind.

I guess that's complicated. It seems like an optimal strategy if you are a chip designer (e.g. Nvidia or AMD vs Qualcomm). Not so much if you are a fabricator. Of course Intel being both makes things a lot harder for them.

replies(2): >>45028437 #>>45034385 #
5. JustExAWS ◴[] No.45028437{4}[source]
> TSMC doesen't design or sell the chips. If they have limited capacity they will of course charge more for manufacturing mobile chips if they can sell the capacity to Nvidia/AMD/Apple instead

That’s the entire point. Intel is behind because their vertical market strategy of using their own fabs only for their own chips doesn’t give them enough volume compared to TSMC who has volume because they are a foundry.

> ARM chips (and that's pretty much by design based on ARM's business model) are close to being a commodity.

ARM doesn’t manufacture chips. The entire argument is that it’s a strategic interest for Intel to manufacture chips in the US. ARM is irrelevant to this conversation.

> Apple is of course an exception but they are not directly part of the CPU market. And ARM and Qualcomm are barely bothering trying to compete with them because there doesen't seem to be a lot of point. They themselves are pivoting to datacenter because there is just more money to be made there.

Apple, Nvidia and to a lesser extent all of the companies that are designing chips and using TSMC as a foundry are more relevant than x86 chips.

Between phones, tablets, watches, and Macs, Apple, etc alone sells more devices with Arm chips than PCs and servers sold by Intel.

They have enough scale to fund leading edge processing.

Qualcomm is a bigger seller of processors than anyone since every mobile cellular chip that I’m aware of except for the very few designed by Apple for the low end 16e is sold by Qualcomm.

> I guess that's complicated. It seems like an optimal strategy if you are a chip designer (e.g. Nvidia or AMD vs Qualcomm). Not so much if you are a fabricator. Of course Intel being both makes things a lot harder for them.

That’s exactly the issue. Intel the chip designer would be better off if they used TSMC and Intel the fabricator would have a lot more funding if other chip designers trusted them enough to use them as a foundry.

Every company that both tries to be vertically integrated and a “platform” fails at one or the other.

Google - The Google Pixel is an also ran hobby project. But they are relatively successful with their products across iOS and Android

Apple - a great vertically integrated product. But no one uses Apple Music on Android unless they are an iPhone family with the one off Android user. iTunes has sucked from day one on Windows and Safari for Windows was rapidly abandoned.

Microsoft - the Surface laptops have gained some traction. But sales are miniscule in the grand scheme of things.

But you notice in the case of Google and Microsoft, they aren’t crazy enough to manufacture their hardware.

replies(2): >>45029449 #>>45030381 #
6. qwytw ◴[] No.45029449{5}[source]
My only point was that mobile chips are pretty much a commodity these days and not a particularly lucrative market if you want to maximize margins and growth. The fact that there are way more ARM chips sold in a year than x86 servers doesen't prove much.

> Qualcomm is a bigger seller of processors

Yes, even for their highest end mobile chip (Snapdragon 8 Elite) the estimated OEM price seems to be under <$200.

A midrange AMD EPYC 9004 chip is $4000-6000. Presumably the gross margin is also higher. So hardly a fair comparison. Of course Intel seems to be propping up it's server marketshare by dropping its margins, so I wouldn't be surprised if they are lower than Qualcomm's.

> But you notice in the case of Google

They design their own mobile SoCs though which seems like a fairly serious effort for a hobby project (although being an off the shelf design like almost all ARM chips besides Apple's maybe not)

7. ac29 ◴[] No.45030381{5}[source]
> every mobile cellular chip that I’m aware of except for the very few designed by Apple for the low end 16e is sold by Qualcomm

Not sure on volume, but Samsung and Mediatek are major players too. I dont think Google has used anything from Qualcomm since Pixel 5 (2020).

8. SlowTao ◴[] No.45034379[source]
Reminds me of the desktop/server processor space in the late 90's. At least this time we have a dozen vendors that are all using ARM instead of x86/POWER/DEC Aplha/MIPS/IA-64 & ARM. More chance of surviving together rather than making multiple moats.
9. Panzer04 ◴[] No.45034385{4}[source]
Datacenter is not a monolith, and neither is it guaranteed to generate more profit or revenue.

IIRC intel makes 2-3x in client sales what it does in datacenter.