Intel has a Pretty Big Problem..

[adamsleath]

there are many sources.

aggregate sales data isn't always easy to find. always assumptions about this that or the other "source" being accurate.
so called "market research" organisations.

peddies for example.

going by the current hype, here elsewhere, toober sentiments. regardless it has been about spruiking the latest hardware releases...for decades.

pick your search engine. google isnt the only one.

Peddies:

• Q1 2024
  PC CPU shipments decreased by 9.4% sequentially, but increased by 33% year over year. JPR suggests that this may indicate the market has stabilized and returned to its normal cyclic behavior.

iGPU penetration
JPR predicts that iGPUs will dominate the PC segment, with penetration expected to reach 98% within the next five years.

I can surmise from this that there are a **** tonne of pc's being churned out compared to the number which have AIB's in them.

market hype is powerful. just look at the ETH mining craze and gpu's. even now there are people still offloading multiple gpu's.

=
in the past ive sold a bunch of stuff, pc stuff, that is still relevant in the hype circus. also i have a bunch of defunct e-waste, yesterday's merch. sitting in my cupboards. i know others who have done this at a scale multiple times more than myself. gets a bit old. the habit.

 

[TrixP10]

there are many sources.

aggregate sales data isn't always easy to find. always assumptions about this that or the other "source" being accurate.
so called "market research" organisations.

peddies for example.

going by the current hype, here elsewhere, toober sentiments. regardless it has been about spruiking the latest hardware releases...for decades.

pick your search engine. google isnt the only one.

Peddies:

• Q1 2024
  PC CPU shipments decreased by 9.4% sequentially, but increased by 33% year over year. JPR suggests that this may indicate the market has stabilized and returned to its normal cyclic behavior.

iGPU penetration
JPR predicts that iGPUs will dominate the PC segment, with penetration expected to reach 98% within the next five years.

I can surmise from this that there are a **** tonne of pc's being churned out compared to the number which have AIB's in them.

market hype is powerful. just look at the ETH mining craze and gpu's. even now there are people still offloading multiple gpu's.

=
in the past ive sold a bunch of stuff, pc stuff, that is still relevant in the hype circus. also i have a bunch of defunct e-waste, yesterday's merch. sitting in my cupboards. i know others who have done this at a scale multiple times more than myself. gets a bit old. the habit.

can you add in the DC numbers to that oh wise one

 

[adamsleath]

datacentre.
you can have that one if you like. i looked at it some time ago.
last i saw nvidia blitzing with hopper/blackwell. amd gaining. intel still in the game due to legacy support, momentum, customer base etc.

The global data center market size was valued at USD 219.23 billion in 2023. The market is projected to grow from USD 242.72 billion in 2024 to USD 584.86 billion by 2032, exhibiting a CAGR of 11.6%. North America dominated the global market with a share of 39.16% in 2023.

Source: https://www.fortunebusinessinsights.com/data-center-market-109851

North America held the largest data center market share in 2023. The rapid adoption of cloud services, AI, and big data applications has fueled a surge in data center demand in the region. Enterprises are increasingly utilizing generative AI and other advanced technologies, which necessitate enhanced data processing capabilities and robust infrastructure. Substantial investments from key players, such as Schneider Electric, IBM, Cisco, and others, fuel market expansion.

Asia Pacific is expected to grow at the highest CAGR over the projected timeframe. The region experienced higher growth that fueled demand for data center capacity globally, and is expected to continue, with total supply projected to increase from 11.1 GW in 2023 to 26.7 GW by 2028. Furthermore, with a growing population and growing digital demands, Indonesia has become a key market. Major investments are being made to develop hyperscale data centers, particularly in Jakarta and Eastern Java. For instance,

In May 2024, Cisco Systems launched its first edge data centers to expand its security footprint in Indonesia. This facility helps customers in financial services and the public sector to align with local data regulations and compliance requirements.

Source: https://www.fortunebusinessinsights.com/data-center-market-109851

other players that are not in the DIY PC market hype:

Schneider Electric, ABB Ltd., IBM Corporation, Cisco System, Inc., Huawei Technologies Co., Ltd., and Hitachi Ltd are the top players in the market.

Source: https://www.fortunebusinessinsights.com/data-center-market-109851

what we/I hear about in the tooberland is elon musk. google meta etc. nvidia. nasdaq darlings and the red tulips of the capital price spiking stocks.

Nvidia is interesting.

 

[arknor]

Intel might be too big to fail — Washington policymakers are already discussing potential solutions if the chipmaker cannot recover

Too big to fail.

www.tomshardware.com

 

[Rroff]

Not too big to fail, too much of a national security issue for the US for them to fail, but there is only so far before the government would have to step in.

 

[humbug]

Intel might be too big to fail — Washington policymakers are already discussing potential solutions if the chipmaker cannot recover

Too big to fail.

www.tomshardware.com

However, sources say policymakers are shying away from lump sum bailout payments as the White House did for Chrysler and General Motors in 2008. Instead, one of the solutions they’re considering is a government-encouraged private-sector merger, potentially with other Intel rivals like AMD or Marvell.

Right, the government can't just keep throwing money at Intel which they then use to undermine AMD's competitiveness, AMD would rightly have serious legal grievances about that. Pat would love that... given me £20 bn, i spend it on undercutting AMD and now you're going to have to give me another $20 bn so that i can keep doing that because if i don't i fail.

AMD and others have to be part of the solution, Intel will have to like it.

 

[TrixP10]

Right, the government can't just keep throwing money at Intel which they then use to undermine AMD's competitiveness, AMD would rightly have serious legal grievances about that. Pat would love that... given me £20 bn, i spend it on undercutting AMD and now you're going to have to give me another $20 bn so that i can keep doing that because if i don't i fail.

AMD and others have to be part of the solution, Intel will have to like it.

I was reading an article that was suggesting that Gov. should have to be repaid and / or the Goverment would part own the company.

Seems that ATM Intel is a bottomless pit.

 

[TrixP10]

Cash Flow & Liquidity​

As of Sept. 31, 2024, Intel had cash and cash equivalents of $8.78 billion, with $46.47 billion of long-term debt. In the first nine months of 2024, Intel generated $5.1 billion of cash from operating activities compared with $6.8 billion a year ago.

For the quarter, AMD generated free cash flow of $496 million. It ended the quarter with net cash and short-term investments of $4.5 billion and $1.7 billion in debt.

 

[AthlonXP1800]

Intel Fails To Achieve Gaudi AI "$500 Million" Revenue Target, Showing Slacking Performance In The AI Segment

Intel's Gaudi AI accelerators haven't managed to hit its "modest" quarterly revenue, as Team Blue fails in attracting market interest.

wccftech.com

 

[AthlonXP1800]

US Policymakers Are Reportedly Open To Potential Intel "Merger Deal", As They Explore Options To Pull Team Blue Out Of The Danger

Intel's financial troubles have started bothering the US, as they are now exploring options for recovery, including a potential merger.

wccftech.com

Hopefully US government will grant Nvidia permission to acquire Intel so Nvidia will have license to design and manufacture x86 CPUs better than Intel did after failed ARM takeover.

 

[Dg834man]

US Policymakers Are Reportedly Open To Potential Intel "Merger Deal", As They Explore Options To Pull Team Blue Out Of The Danger

Intel's financial troubles have started bothering the US, as they are now exploring options for recovery, including a potential merger.

wccftech.com

Hopefully US government will grant Nvidia permission to acquire Intel so Nvidia will have license to design and manufacture x86 CPUs better than Intel did after failed ARM takeover.

Just imagine the nightmare pricing.

 

[KompuKare]

US Policymakers Are Reportedly Open To Potential Intel "Merger Deal", As They Explore Options To Pull Team Blue Out Of The Danger

Intel's financial troubles have started bothering the US, as they are now exploring options for recovery, including a potential merger.

wccftech.com

Hopefully US government will grant Nvidia permission to acquire Intel so Nvidia will have license to design and manufacture x86 CPUs better than Intel did after failed ARM takeover.

Would AMD grant Nvidia A64 rights, or would Nvidia have to stick to i386 compatibility?!

 

[humbug]

Would AMD grant Nvidia A64 rights, or would Nvidia have to stick to i386 compatibility?!

They would, because while i386 is near obsolete its not completely obsolete, in the same way that Intel cannot operate without AMD64 AMD also can't really operate without i386, together they make up X86_64.

IMO the best way forward is for all the big US tech firms to have a stake in Intel foundries, that is separate from Intel's chip IP.

 

[Nathlm]

Not too big to fail, too much of a national security issue for the US for them to fail, but there is only so far before the government would have to step in.

How far do you think the government will go to save intel?

I think they're probably already looking elsewhere, as I don't think Intel is viable anymore unless it gets its act together

 

[Grim5]

Not too big to fail, too much of a national security issue for the US for them to fail, but there is only so far before the government would have to step in.

Yep Intel is lucky the market right now would make the US government bail them out. If the market was different; say if AMD had its own fabrication in the U.S then Intel would be up the shoot

The other problem is Intel still makes chips the US military uses today - quick example: many U.S fighter jets are using computers with Intel processors in them and they can say goodbye to new and replacement parts if Intel disappears.

They need to do a cost benefit analysis to see how much electronics needs to be replaced with new devices on their military systems - and what I do know is while the newest stuff like the F-35 and 6th gen planes is designed to be modular, a lot of old military equipment is the opposite of modular and would simply not function without the original parts - the F-22 for example runs on an Intel i960MX processor and can't function without it

 

[jigger]

The amount of money it would take to buy Intel would be staggering. Politics and competition issues aside, I’m really not sure any one firm could manage it financially and the timeframe for a takeover would probably be decades. I think government intervention or some kind of partnership/stake hold, as Humbug said would be most likely, if anything.

 

[humbug]

So we know now that TSMC 3nm wafers are costing Intel $24,000 a pop, the CPU is a girthy 243mm, its bigger than my 5nm GPU!

Out of a 300mm wafter, which TSNC's 3nm are they are getting 242 CPU's, because its huge, / that by $24,000 = $99 per CPU, that's not including the PCB, foveros packaging and substrate, assembly ecte... (Intel use an interposer, AMD don't) this thing is $200+ before its even sold in to the first stage of the supply chain.

That is a pretty big problem.

Assuming AMD are paying full price for 4nm that is $18,000 a pop, they are getting 862 eight core dies per wafer, so 431 sixteen core CPU's, $41 per CPU, the 6nm wafers are dirt cheap now, so add $10 for the IO die to that, about $50, an 8 core CPU is only about $30 in wafer costs.

 

[humbug]

There is another thing, IMO Intel MCM designs are pointless.

The reason AMD went MCM is for modularity, they can use better lithography for the logic dies and older cheaper lithography for the analogue dies given those don't scale anyway, so what you have is the CPU dies are on 4nm and the IO die on 6nm, you save costs like that.
More over you can construct them like lego, 8 core CPU you put one logic die on it, 16 core is 2 logic dies, 32 core 4 and so on, theoretically there is no limit to how many logic dies you can lego on to a PCB, the PCB size being the only limit.
You can also use them for other products, for example Strix Halo is a 40 CU APU with 16 cores, all they do is make a 40 CU GPU tile and glue 2 eight core logic dies on to it.
AMD only make one logic die for all and any application. A singular Lego brick for everything.

So AMD's MCM designs were well-thought-out and very intentional, they didn't do it just for the sake of doing it.

Intel's MCM packaging doesn't work like that, nothing on the 285K is compatible with anything but its self, you can't use any of the tiles for any other product type, so what's the point in them?

Its like Intel can see the success of AMD's MCM idea but haven't even begun to understand why. Them mocking it all those years ago i think proves that.