Thread by @IanCutress

#IntelInvestorMeeting Thread - keynote presentations

I think someone exported this starting video at 720p.

Looks like it's an actual event with people on site.

People were invited. One day, perhaps.

3 hour event

Oh no

Oh no

He's wearing that Torrid t-shirt again

Please can someone tell him that the word 'Torrid' doesn't translate to other english speaking countries the way he thinks it does

Torrid is a new word for 'English'.

No come on, please. Is this your blind spot? Really?

Sometimes you have to tell your CEO where their blind spots are.

'We're ahead of schedule to where I thought we'd be when I joined'

Only Intel can spread across all these

Executing on strategy - future of growth begins today.

> Today? Not... last month?

>The six business units are a good breakout. Does the Custom Compute for blockchain section become a 7th?

Quecca-scale.

What. The. WTF.

I'd write all the buzz words Pat is saying right now, but it's not needed.

Driving Moore's Law. Note the 'Transistors per package' metric.

1 trillion transistors per package by 2030.

@CerebrasSystems was doing that with wafer scale in 2020. Just some perspective here that some (well-funded) moonshots already exist.

PG: I helped create USB. My daughter plugs in a stick and says 'thank you papa'.

<silence in the room>

What was... that...?

Enabling open ecosystems

Geographically resilient supply chains

Based on this graph, USA never led?

USA CHIPS Act, EU CHIPS Act, Intel going after incentives

Why is Alder Lake is quotes, but none of the others are?

Pat says he wants one of the Mobileye cars. He's the CEO, why can't he have one?

Five nodes in 4 years.

'Doesn't it take 2 years per node?'
'Intel will catch up quickly'

That's because all node updates become new numbers, rather than a +

Manufacturing has almost 'an unlimited budget' to get Intel back on track

PG: Grove used to say 'we gave you an unlimited budget and you still overspent'

<silence>

Process Leadership is based on perf/watt. It's a very important distinction to remember

Meteor Lake hasn't taped out yet. Later this year.

First SRAM Wafers of 18A. GAA/RibbonFET test chips

PG's 'Simba' moment.

PG: If you weren't amazed by Ann's presentation this morning, you're not a geek.

huh... what? Really? What is this statement?

New megafab location in Europe coming

Intel creates growth engines for the industry

More software engineers at Intel than Pat was managing at VMWare, a software company.

...aaaaand there's the religious reference.

Intel's software is a differentiator

How much $ is Intel making from software today?

PG: The Best Team in the Industry

Did you hear that $TSM $NVDA $AMD Samsung

PG: 17k new employees in 2021. The Brain Drain is gone.

PG: We invented OKRs

what?

Intel revenue growth expectations

PG: None of the following assumptions are heroic

Arrow Lake name now confirmed

Alder Lake 12th Gen
Raptor Lake 13th Gen
Meteor Lake 14th Gen
Arrow Lake 15th Gen

Xeon updates. Granite Rapids, Sierra Forest

Sierra Forest is an all E-core version of Xeon for 2024

New platform every 2 years. New product every year.

Emerald Rapids will have increased core count, but feature compatible with Sapphire Rapids

Alder Lake leverages 'Hybrid' architecture

Pat then uses that as a point to mention that Xeon will do similar, but all P-core and all E-core. No Hybrid for Xeon

Intel 3 process node ?

PG: 2024 is an inflection point in competitiveness

Networking and Edge: NEX

PG: Shifting networking compute on dedicated silicon, leaving CPU free to do normal work

AXG Revenue in 2021 was $0.7B
AXG Revenue in 2022 will be >$1B

So 4m GPUs for 300m up to 1.3b (?). How much revenue per GPU?

PG: 'Business ain't fair'

PG: HPC has been niche for 40 years. Now is mainstream.

PG: Developer ecosystem is pumped

Falcon Shores. One CPU chiplet, two Xe chiplets. ? Or a central IO die with Xe GPU, and two CPU chiplets?

Automotive

IFS - Intel Foundry Services

Strong pipeline of customers
Already running 30+ different test chips in 2022

What's the scale on this revenue graph? Does is start at zero? So a 2x increase in revenue by 2026?

OK so 2x from current offering, unknown revenue from future node offerings

18A on IFS in 2H 2024.

Add in Tower, available today. Wait, doesn't it close in 12 months?

Tower adds $1.5b revenue to IFS today.
EPS accretive.
But GMs are sub 20%.

What annoys me about this is PG stating that IDM and IFS are always positive feedback loops. If it goes negative, then it speeds up that spiral

Intel going to run factories on nodes much longer, requires build out of new facilities.

PG doesn't mention that it means lots of demand. One of the biggest fears is that market does a downturn in demand. PG defends this, saying they can bring more in-house. That's not easy

For example, if that happened today, you can't simply build the TSMC N6 GPU on an Intel 7 process and expect the same result overnight. It's a 3-5 year process to shift products from fab to fab.

PG: Not all these need to hit

1.0 model

2.0 model with IFS

PG: Use customer pre-pays to create capital flexibility

...who is prepaying? I mean, without MDF incentives

Pat teasing that CapEx intensity will decrease later in the decade.

What makes me think that if he wants to succeed, he can't let his foot off the gas.

PG: How do you know we are executing? We'll give you proof points with products on market and our roadmaps, like Alder Lake and Intel 7.

Me: That's not an answer. Those were in the pipe before you joined. Words are still words, not execution.

Roadmaps

PG: Aim to double revenue and double Intel's multiple. A 4x of shareholder intel value.

So, $800b market cap?

Sandra Rivera @SandraLRivera to the stage, talking datacenter and AI

@SandraLRivera

2x in 5 years?

More Ice Lake Xeon CPUs shipped in December than AMD shipped all year.

Just don't ask about ASPs and GMs

Sapphire Rapids, coming 2H22

Demo

The demo seems to not going well, they're showing the wrong videos

Intel: AI in the Xeon, no need for an Accelerator

Also Intel: But we'll also sell you an accelerator.

Here's the Future Xeon Roadmap. I'm glad we're getting roadmaps

SR: Intel 3 using higher density, higher performance libraries.

Which is it? high density, or high performance? Usually it's a scale. This is still FinFET based EUV, not GAA.

Intel can leverage its enterprise portfolio. No other company can do all of this at once. As a result, Intel is pushing combo Intel solutions for market value

$40B silicon TAM in AI by 2026. Seems low? Depends what you include

Now explaining what SGX again. Is it SGX or TSX that's been talked about, implemented, disabled, implemented, etc

Michelle Johnston Holthaus to the stage @MJHolthaus EVP of CCG

@MJHolthaus Number of PCs per household is increasing

Jim Johnson, Interim GM of CCG to the stage

Client revenue growing - still grew when Apple left and Intel divested two technologies (5G and memory?)

This graph is somewhat misleading. Plot back to 2018 when PCs were at a historic low. Are we really expecting >360m units per year for the next 5 years?

Cloud Client Experiences - immersive experiences through virtualization

Lunar Lake to use external chiplets

This confirms the Lunar Lake name then as well

Note that with Meteor and Arrow Lake, Intel won't have performance per watt leadership according to this

ADL is ramping fast

Sneak peak of Raptor Lake, 8 Perf cores, 16 Efficiency Cores

I can get more cores with Xeon Phu, just sayin'

'Up to double digit'. So less than 10%

Why is the AI M.2 module 'specific' to Raptor Lake? If it's M.2, it can be on any system

Meteor Lake: Next-Gen Graphics, tGPU (tile GPU). Shipping 2023

Back to Raptor Lake: Shipping 2H 22

Now ecosystem

Evo

CCG Summary

OK 15-20 minute break. Then:

Financial
Panel Q&A
Drinks

That last one, only if you're on site. Or maybe not.
It's 11:30pm here. Not sure if I'll say awake.

ok one more.

'People are worried about our low 50s Gross Margins. We have a path to strong mid 50s, high 50s'.

Wasn't the demand to push back to mid-to-high 60s?