News Lip-Bu Tan's first speech as Intel CEO focuses on innovation and working with foundry customers

[Admin]

Intel CEO Lip-Bu Tan made his first keynote speech, laying down the path he intends to take for the company and soliciting customer feedback.

Lip-Bu Tan's first speech as Intel CEO focuses on innovation and working with foundry customers : Read more

 

[jam200nd]

First time contributor, long-time reader. This is the right decision by Lip-Bu Tan. Intel needs to hold on to its parts that support the vision of the company and grow it organically. I'm not a fan of the Wall Street Gurus and others who say split up the company. Many times when this happens the cash flow falls for all segments. Instead of synergy you have failure and lost investments. My view.

 

[ekio]

"Focusing on innovation", from the company that keeps using the same rotten obsolete ISA x86 for the past 47 years... Nice one.

 

[User of Computers]

rotten obsolete ISA x86

In what way is x86 rotten and obsolete? Point me to an objectively better ISA that has mass market adoption.

 

[redgarl]

First time contributor, long-time reader. This is the right decision by Lip-Bu Tan. Intel needs to hold on to its parts that support the vision of the company and grow it organically. I'm not a fan of the Wall Street Gurus and others who say split up the company. Many times when this happens the cash flow falls for all segments. Instead of synergy you have failure and lost investments. My view.

Sorry, but this whole strategy is Pat strategy, and they fired him for this exact same strategy.

Honestly, I called that Intel downfall will be due to their fabs in 2020, and I am now more convinced than ever in 2025.

They will fail and a new CEO will change nothing about it.

Intel is a leader of nothing while trying to do everything. They should focus on CPUs and drop GPUs and fabs or they will eventually lose their cashflow to follow their business model (bribing OEMs) for maintaining their mobile marketshare. At the minute they will lose mobile, it will be the end of intel and their fabs is their main cost driver because they can`t match TSMC leadership.

 

[redgarl]

In what way is x86 rotten and obsolete? Point me to an objectively better ISA that has mass market adoption.

I just replied to you with a similar reply you made, asking for an explanation...

There is something called google you know...

 

[bit_user]

I'm not a fan of the Wall Street Gurus and others who say split up the company. Many times when this happens the cash flow falls for all segments. Instead of synergy you have failure and lost investments.

In the long term, it makes a lot more sense than for Intel to be this weird beast that's a foundry + design house, joined at the hip. I'm sure the financial argument for splitting them will win the day. There's just no good reason why such a capital-intensive manufacturing business should drag down the design side's profitability, compared to what other fabless semiconductor companies are achieving.

For the time being, the best option is clearly to have the CPU business help support the foundry, until it's self-sustaining with a viable roadmap which keeps it at the leading edge (yes, that's shade I'm throwing at Global Foundries).

 

[bit_user]

"Focusing on innovation", from the company that keeps using the same rotten obsolete ISA x86 for the past 47 years... Nice one.

They do AI on some non-x86 architectures, like their NPU (powered by SHAVE DSPs) and Gaudi. Their GPUs seem to be maturing nicely, as well.

A few years from now, I think it should become clear which direction they're going to take with their CPUs. For the time being, it looks like they'll do battle, wielding APX and AVX10.2 as their weapons. Should be an interesting fight.

 

[bit_user]

Sorry, but this whole strategy is Pat strategy, and they fired him for this exact same strategy.

I think if Pat had set more realistic expectations and done a better job on execution, he might still be there. That said, it's entirely possible that the transition Intel needed couldn't be completed under the tenure of a single CEO - that when they inevitably had a bad quarter, the investors would run out of patience and demand blood.

 

[gg83]

The title could swap pat gelslinger. That's exactly what he was focusing on.

 

[User of Computers]

I just replied to you with a similar reply you made, asking for an explanation...

There is something called google you know...

Yes, but Google doesn't know everything, nor does it really support your argument, so I was looking for something straight from the horse's mouth.

 

[bit_user]

Yes, but Google doesn't know everything, nor does it really support your argument, so I was looking for something straight from the horse's mouth.

Not directly answering your question, but I found some interesting materials I thought were relevant.

Here's a diagram depicting the instruction encoding of x86 (32-bit):

Source: http://pnx.tf/files/x86_opcode_structure_and_instruction_overview.pdf[/b]
​

This diagram shows the opcode format of ARMv7 (32-bit):

Source: http://emucode.blogspot.com/2010/09/decoding-arm-instruction-set.html
​

Sorry, I looked for 64-bit versions of each, but didn't find anything as good.

Anyway, the diagrams aren't directly comparable, but if you understand what they're each showing, it's clear than ARM is a couple orders of magnitude simpler. In fact, it's so cheap to decode that when ARM started dropping support for the 32-bit ISA, it got rid of mop caches. These caches are normally used to bypass the decoder, for instruction sequences the CPU recently executed. Except, with a decoder so cheap, you no longer need a mop cache and can just devote that die space to logic which does actual work!

 

[TerryLaze]

In the long term, it makes a lot more sense than for Intel to be this weird beast that's a foundry + design house, joined at the hip. I'm sure the financial argument for splitting them will win the day. There's just no good reason why such a capital-intensive manufacturing business should drag down the design side's profitability, compared to what other fabless semiconductor companies are achieving.

For the time being, the best option is clearly to have the CPU business help support the foundry, until it's self-sustaining with a viable roadmap which keeps it at the leading edge (yes, that's shade I'm throwing at Global Foundries).

Oh great, intel should become another one of the companies that only provide for servers and gets scalped for desktop due to low units produced for it and high demand.....just like nvidia and amd, great!

Also, if you think that the Fab side can become self-sustaining then how would it be dragging down the design side?

 

[bit_user]

Oh great, intel should become another one of the companies that only provide for servers and gets scalped for desktop due to low units produced for it and high demand.....just like nvidia and amd, great!

I don't follow.

Also, if you think that the Fab side can become self-sustaining then how would it be dragging down the design side?

The chip design business is a lot like software, with relatively low headcounts, low amounts of depreciating assets, shorter return-on-investment, and higher profitability. For investors, they like to see this business decoupled from the capital-intensive manufacturing business that requiring lots of up-front investment and longer turnover. I'm not a financial guru, so I can't explain the nuances of why, but I think it has something to do with different types of investors being interested in those different types of businesses.

 

[TerryLaze]

I don't follow.

TSMC is booked out for a few years, intel would have no chance in hell to make enough CPUs if they didn't have their own FABs, if they spinn off the FABs they have and they become successful they will have no chance in hell to make enough CPUs because everybody else will be making things there.

The only way for intel to make enough CPUs is to have their own FABs so that they can control the volume.
Without their own FABS they would only be able to trickle out CPUs/GPUs.

The chip design business is a lot like software, with relatively low headcounts, low amounts of depreciating assets, shorter return-on-investment, and higher profitability. For investors, they like to see this business decoupled from the capital-intensive manufacturing business that requiring lots of up-front investment and longer turnover. I'm not a financial guru, so I can't explain the nuances of why, but I think it has something to do with different types of investors being interested in those different types of businesses.

So your main concern is to keep the investors happy....

The only investors in intel, that I know of, at the moment are the ones for the FABs that invested specifically for the FABs.
Who is investing into intel on the design side?!

 

[bit_user]

TSMC is booked out for a few years, intel would have no chance in hell to make enough CPUs if they didn't have their own FABs,

Arrow Lake and Lunar Lake are made in TSMC fabs and only packaged by Intel. Of their latest products, only their server CPUs are still being made in their own fabs, right now.

if they spinn off the FABs they have and they become successful they will have no chance in hell to make enough CPUs because everybody else will be making things there.

That's far enough in the future that hopefully their foundry will have enough capacity to cope, by then. Anyway, it'll just put them in the same boat as everyone else making CPUs. So, they won't be at a relative disadvantage.

So your main concern is to keep the investors happy....

Investors own the company. They ultimately decide what happens. I'm just trying to explain their thinking, as best I understand it.

The only investors in intel, that I know of, at the moment are the ones for the FABs that invested specifically for the FABs.
Who is investing into intel on the design side?!

It's a publicly-traded company, so it'd be a long list and we can't really say why any of them invested in it.

 

[TerryLaze]

Arrow Lake and Lunar Lake are made in TSMC fabs and only packaged by Intel. Of their latest products, only their server CPUs are still being made in their own fabs, right now.

Yeah and you think that if intel had to make all of their products at tsmc then it would give them better availability?!

Anyway, it'll just put them in the same boat as everyone else making CPUs. So, they won't be at a relative disadvantage.

No, no disadvantage, they will be in the same boat as everybody else, as I said, being scalped and not having enough to go around for consumers.

Investors own the company. They ultimately decide what happens. I'm just trying to explain their thinking, as best I understand it.


It's a publicly-traded company, so it'd be a long list and we can't really say why any of them invested in it.

Stockholders are not investors, investing in stocks is not investing in the company, other than the initial cash injection when first going public, or when releasing even more shares to the public, there is very little gain for a company.

Also intel has been doing stock buy backs since 1990, so they are not fully publicly traded, it wouldn't make sense for stockholders to buy back stocks from themselves for themselves...
So the bottom line is, who knows how much of intel is public and how much private, if the private part is more than 50% then screw everybody, they can do whatever they want. (within reason, because they would still not want to make stockholders too angry)

 

[bit_user]

Yeah and you think that if intel had to make all of their products at tsmc then it would give them better availability?!

I'm just pointing out how easily TSMC seems to be handling that amount of production volume. I can't say whether the additional load from the server CPUs would be a problem, but with enough planning, perhaps not. Rumors were already swirling about Intel's buy of TSMC N3 wafer capacity back in like 2021, so that would've given TSMC some time to build out the necessary production capacity.

Stockholders are not investors, investing in stocks is not investing in the company,

I'm not going to argue semantics. The shareholders are the ones who vote on the big decisions, including the board composition. The board is who ultimately calls the shots about how the company is run.

Also intel has been doing stock buy backs since 1990, so they are not fully publicly traded, it wouldn't make sense for stockholders to buy back stocks from themselves for themselves...

That doesn't matter, when it comes to governance. When they buy back publicly-traded shares, those become non-voting shares, unless they're re-issued. It doesn't make the company any more autonomous.

 

[TerryLaze]

I'm not going to argue semantics. The shareholders are the ones who vote on the big decisions, including the board composition. The board is who ultimately calls the shots about how the company is run.

That was my question, who actually knows every shareholder of intel.

That doesn't matter, when it comes to governance. When they buy back publicly-traded shares, those become non-voting shares, unless they're re-issued. It doesn't make the company any more autonomous.

Re-issued means that even you realize that they are not gone, as long as they exist whoever owns them has that amount of vote in any matter.

You could be right though, if they go to someone who has the majority share they become deciding shares and anybody else would have to vote against if they so want.

 

[bit_user]

That was my question, who actually knows every shareholder of intel.

What I've found is that about 45% are retail investors. You can also see a list of the biggest shareholders.

Re-issued means that even you realize that they are not gone, as long as they exist whoever owns them has that amount of vote in any matter.

Oh they can also be dissolved, in which case they are gone. You have to look at the details of the buyback, in order to see what happened to the shares.

You could be right though, if they go to someone who has the majority share they become deciding shares and anybody else would have to vote against if they so want.

I don't know what you mean by this. If they company does a buyback, the shares don't just go to a big shareholder. The only exception to that might be when stocks are given to employees, as part of their compensation package or through exercising options. I think that's usually done by issuing new shares, which dilutes the equity of existing shareholders. So, if a company is doing buybacks, some of those shares might get re-issued to employees, instead.

 

[ekio]

In what way is x86 rotten and obsolete? Point me to an objectively better ISA that has mass market adoption.

x86 was designed almost 50 years ago, with designs that didn't age very well over time.
To compensate that fact, it was patched with tons of extra instructions, while keeping backward compatibility.
The result is a Frankenstein set of instruction that is absolutely horrific and a huge waste of unneeded specs to implement (some modern ones, some, old ones, some completely outdated ones...)

Arm v8/v9 and risc-v64 are made from clean sheets and rely on much more efficient foundations (risc instead of cisc, meaning simpler/more efficient instructions and more consistent rules.

x86 is only being used a lot because, for years, Intel bullied competition and prevented innovation by destroying healthy competition, and forced PCs being x86, just like Microsoft forced PC to be Windows based. Now that the Intel mafia is down, innovation can start again.

 

[TerryLaze]

Oh they can also be dissolved, in which case they are gone. You have to look at the details of the buyback, in order to see what happened to the shares.

So you don't know what's happening with them but we keep discussing it....

I don't know what you mean by this. If they company does a buyback, the shares don't just go to a big shareholder. The only exception to that might be when stocks are given to employees, as part of their compensation package or through exercising options. I think that's usually done by issuing new shares, which dilutes the equity of existing shareholders. So, if a company is doing buybacks, some of those shares might get re-issued to employees, instead.

So you don't know what's happening with them but we keep discussing it....

Usually when a company goes public they (the owners/founders) keep the majority share and only sell/make available up to 49% of the company.
So buy-backs would actually literally be buy-backs of the founders buying back parts of the company they sold.

x86 was designed almost 50 years ago, with designs that didn't age very well over time.
To compensate that fact, it was patched with tons of extra instructions, while keeping backward compatibility.
The result is a Frankenstein set of instruction that is absolutely horrific and a huge waste of unneeded specs to implement (some modern ones, some, old ones, some completely outdated ones...)

That's not because of some nefarious business, it's because PC users don't want to throw away their PCs with every or every other new OS like you have to do with android devices.

x86 is only being used a lot because, for years, Intel bullied competition and prevented innovation by destroying healthy competition, and forced PCs being x86, just like Microsoft forced PC to be Windows based. Now that the Intel mafia is down, innovation can start again.

Apple went from using arm to using x86, not because they where bullied by intel but because arm was terrible back then, now apple went back to arm because now it's tolerable.
Amiga used arm and then went bankrupt because they had no chance to compete on performance, same for acorn with their archimedes and any other computer that used arm back then.

 

[bit_user]

So you don't know what's happening with them but we keep discussing it....

Don't try to pin this on me. You're the one who raised share buybacks, claiming they had governance implications. I pointed out that they shouldn't affect governance. You have yet to demonstrate otherwise.

So you don't know what's happening with them but we keep discussing it....

We can only have a discussion if I understand what you're saying. I said that to give you an opportunity to explain your point, because otherwise I can't provide a meaningful response.

Usually when a company goes public they (the owners/founders) keep the majority share and only sell/make available up to 49% of the company.
So buy-backs would actually literally be buy-backs of the founders buying back parts of the company they sold.

That's not how it works. A buyback is the company buying back the shares, not the founders. If the founders wanted more equity, they would just buy them on the open market and it wouldn't be termed a buyback.

Apple went from using arm to using x86, not because they where bullied by intel but because arm was terrible back then, now apple went back to arm because now it's tolerable.

No, Apple's Macintosh computers went from using Motorola 68K -> PowerPC -> x86 -> ARM.

They first adopted ARM for their iPods, then phones, then tablets, and finally laptops/desktops/workstations.

Amiga used arm and then went bankrupt because they had no chance to compete on performance,

Amiga used Motorola 68K-series. After bankruptcy, others licensed the branding and OS, who developed PowerPC models, but Amigas stuck with 68K long after Apple. From what I'm reading, the 68040 performed somewhat competitively with its contemporary, the i486. The 68060 likewise performed in the same ballpark as the Pentium. I think Amiga's problems weren't primarily due to performance, but I'm no expert on the subject.

 

[AICl0ud]

Just like the former Intel CEO, who loved to engrave his name on electronic chips, Lip-Bu has "just as much" (no) vision for the future.

He too ignores a potential four billion customers: smartphone users. And we're not even talking about tablets.

But Microsoft does it, so why not Intel?
In short, Intel has changed its director, but that's it.

Meanwhile, ARM, which has a monopoly on smartphones and tablets, is now attacking PCs.

Money doesn't make you smarter, especially when that's the only goal!

 

[bit_user]

He too ignores a potential four billion customers: smartphone users. And we're not even talking about tablets.

Intel tried making phone SoCs about a decade ago. It didn't work out and they lost $Billions in the process.

https://www.anandtech.com/show/10288/intel-broxton-sofia-smartphone-socs-cancelled
​

At least it helped them develop their E-cores, I guess.