Samsung to Produce Tesla Chips in $16.5 Billion Multiyear Deal

[Xebec]

This story and What does it have to do with Strongarm?
I think it's a bit unreasonable
Apart from that, instead of strongarm, there must have been a way to create x86 chips with ultra-low power segments by focusing on low power consumption designs.
I don't think it would have been impossible if I had worked hard at that point.
In fact, the ATOM for the early days of mobile In fact, I don't think the ATOM SOC for the early days was bad.

I can understand and agree if you say this Intel has started the foundry business earlier.
Even TSMC may have become an untouchable existence.

Agreed - not impossible, if Intel had gotten a foundry going in the mid-to-late 2000s, (my guess is) TSMC would have been a bit slower getting to 7nm and 5nm, even with large GPU / Bitcoin / AI silicon commitments.

Intel was in the process of selling StrongARM/XScale or had already signed the deal around the time Apple approached Intel to produce chips for the original iPhone. XScale was sold in 2006, and the iPhone launched in 2007, but of course discussion for the iPhone chip fabbing would have been a few years before 2007.

The early ATOM SoC was bad from a performance perspective compared to desktops, but so were the early iPhone and Android ARM based CPUs. Atom kind of stayed weak because Intel already had higher end silicon, but iPhone/Android ARM devices improved greatly over time vs. Atom because they had to get better.

But yes - the iPhone pitch was maybe the best opportunity for Intel to really get a Foundry type business off the ground. They were already used to fabbing mutiple architectures and IP sets (StrongARM/XScale alongside x86, and later they'd of course pick up FPGAs).

 

[tomatoma]

Agreed - not impossible, if Intel had gotten a foundry going in the mid-to-late 2000s, (my guess is) TSMC would have been a bit slower getting to 7nm and 5nm, even with large GPU / Bitcoin / AI silicon commitments.

Intel was in the process of selling StrongARM/XScale or had already signed the deal around the time Apple approached Intel to produce chips for the original iPhone. XScale was sold in 2006, and the iPhone launched in 2007, but of course discussion for the iPhone chip fabbing would have been a few years before 2007.

The early ATOM SoC was bad from a performance perspective compared to desktops, but so were the early iPhone and Android ARM based CPUs. Atom kind of stayed weak because Intel already had higher end silicon, but iPhone/Android ARM devices improved greatly over time vs. Atom because they had to get better.

But yes - the iPhone pitch was maybe the best opportunity for Intel to really get a Foundry type business off the ground. They were already used to fabbing mutiple architectures and IP sets (StrongARM/XScale alongside x86, and later they'd of course pick up FPGAs).

Fundry business is likely to be larger than architecture issues.

In reality, it takes time for Apple's in-house designed A4 to come out.
Even if you use Intel's factory, you don't necessarily need to use Intel's products or architectures (e.g. x86 or strongarm/XScale).

Also, even if Intel had taken the hegemony, Looking at the industry as a whole, the appearance of EUV is likely to be a bit delayed, so I think the timeline is not that different from the real world.
In fact, it took a long time because the full exposure process shifted to EUV after 2020…

 

[tomatoma]

I want Intel to try hard...
let me dream

 

[hist78]

A little more on perhaps why Samsung:

Curious how Musk will react if Samsung yields in TX follow their pattern of low yields in SK on advanced nodes.

View attachment 3399

Since when did Elon Musk or Tesla become a semiconductor manufacturing guru?

 

[tomatoma]

Since when did Elon Musk or Tesla become a semiconductor manufacturing guru?

No, Elon uses these words, but I guess it just feels like Tesla will be sponsored by factory automation.
For example, using Tesla's robotic technology, automation of manufacturing factories

 

[MKWVentures]

I am struck by how people think that Intel "Missed" the opportunity.

Intel tied for this as hard as they could. Tesla chose Samsung. I'm not sure it was a very dificult choice

If we assume Pay wasnt incorrect when stating every customer was running text chips or looking at Intel years ago, then it is now clear everyone looked at Intel and chose someone else.

 

[Xebec]

No, Elon uses these words, but I guess it just feels like Tesla will be sponsored by factory automation.
For example, using Tesla's robotic technology, automation of manufacturing factories

That would be interesting if Samsung ends up using some Teslabots for the Texas factory..

 

[Xebec]

I am struck by how people think that Intel "Missed" the opportunity.

Intel tied for this as hard as they could. Tesla chose Samsung. I'm not sure it was a very dificult choice

If we assume Pay wasnt incorrect when stating every customer was running text chips or looking at Intel years ago, then it is now clear everyone looked at Intel and chose someone else.

(I know you have a lot of good info I don't have...)

There are a LOT of ways to screw up on contract bids..

Do you thnk Tesla actually trialed chips on Intel 3 or 18A before deciding?

 

[hist78]

No, Elon uses these words, but I guess it just feels like Tesla will be sponsored by factory automation.
For example, using Tesla's robotic technology, automation of manufacturing factories

If that's the case, I think starting now, Elon Musk and his staff will receive coffee or tea at a nearby Starbucks, but will never be allowed to set foot in an Intel or TSMC fab.

 

[Daniel Nenni]

I do not think Elon Musk and Lip-Bu Tan would be good business partners for this kind of venture. Vey different business styles and personalities. Unless Musk bought Intel outright and mushed it into his xAI conglomeration. It also said the deal was signed before Lip-Bu took over.

“Samsung agreed to allow Tesla to assist in maximizing manufacturing efficiency. This is a critical point, as I will walk the line personally to accelerate the pace of progress,” Musk said on X, and suggested that the deal with Samsung would likely be even larger than the announced $16.5 billion.

I have not done the math but it is hard to believe that Tesla will buy that many wafers. Maybe his xAI will be involved as well? And SpaceX? I also question having Elon Musk next door and him the walking manufacturing line. It sounds disruptive but not in a good way. Elon seems to chase shiny things.

If you compare this to the deal Apple did with TSMC it would probably be a complete opposite. Maybe Elon will fall in love and buy Samsung Foundry like he did with Twitter? Twitter lost $1.1B prior to the acquisition in Q4 2022.

Samsung Foundry did $18B last year but how much of that was internal Samsung? Operating loss was $2.4B in 2024 and $3.6B in 1H 2025. Sounds like it is ready for a Musk makeover!

 

[DanX]

It's because Samsung is willing to lower the price.

IFS needs to reassert its position.

I know because when we talked to Samsung ,they were literally begging us to tapeout in their fab.

IFS is the #3 choice for us since they don't have a PDK.

 

[Xebec]

I have not done the math but it is hard to believe that Tesla will buy that many wafers. Maybe his xAI will be involved as well? And SpaceX? I also question having Elon Musk next door and him the walking manufacturing line. It sounds disruptive but not in a good way. Elon seems to chase shiny things.

FWIW, Tesla's AI6 is meant to be a universal chip - for the Tesla robots, Automotive full self driving, and also AI inference and supercomputing. No doubt intended for xAI Grok too (+ Tesla Dojo).

There's a pretty large upside potential, but.. it's only potential without guarantees.

P.S. The media keeps referring to the chip as the A-16 rather than AI 6... kiinda funny how they keep all parroting the mistake.

 

[Xebec]

It's because Samsung is willing to lower the price.

IFS needs to reassert its position.

I know because when we talked to Samsung ,they were literally begging us to tapeout in their fab.

IFS is the #3 choice for us since they don't have a PDK.

Curious - No PDK at all? which process?

 

[DanX]

Curious - No PDK at all? which process?

BTC chip needs 14A, I don't think they have a PDK.

 

[Daniel Nenni]

Curious - No PDK at all? which process?

The Samsung 2nm PDK was a mess so this was not a PDK based decision. Tesla worked with both TSMC and Samsung on previous designs. I really think this is a Texas driven deal since Tesla is big in Texas which is why Elon Musk moved there. I also think Elon wanted control of a fab and only Samsung would do that. I hope it all works out because this deal reaches out to 2033/34. A lot can happen during that time in the semiconductor manufacturing world.

$16.5B / 8 years = $206.25 million per year

Apple paid TSMC roughly $22B in 2024
Nvidia paid TSMC roughly $14B in 2024
AMD paid TSMC roughly $8B in 2024

Intel needs that type of partner for 14A, not Tesla. Just my opinion of course.

BTC chip needs 14A, I don't think they have a PDK.

Yes, there is a preliminary PDK for 14A out in the wild. Besides, Intel 18A is competitive with Samsung 2nm and TSMC N2. Intel also has better packaging than Samsung.

TSMC N2 vs Intel 18A vs Samsung SF2 Wiki - Semiwiki

Category TSMC N2 Intel 18A Samsung 2nm (SF2) Target Mass Production 2H 2025 (risk production late 2024) H2 2024 (internal); early 2025 for IFS customers Late 2025 – risk production Transistor Type GAA nanosheet (TSMC’s implementation) RibbonFET (Intel's GAA design) MBCFET™ (Multi-Bridge Channel...

semiwiki.com

 

[DanX]

As I figure it, IFS has absolutely no grasp of the fab ecosystem.
If they are serious about fabs, BTC miner customer should be their first customer.

 

[DanX]

Apple paid TSMC roughly $22B in 2024
Nvidia paid TSMC roughly $14B in 2024
AMD paid TSMC roughly $8B in 2024

Intel needs that type of partner for 14A, not Tesla. Just my opinion of course.

This is really wishful thinking.
You are the backup plan and the second source. You should expect 10%.

 

[siliconbruh999]

Yes, there is a preliminary PDK for 14A out in the wild. Besides, Intel 18A is competitive with Samsung 2nm and TSMC N2. Intel also has better packaging than Samsung.

Even if they have competitive tech their design rules are difficult to adapt to not to mention Samsung has Cheapest wafers TSMC has the best best wafers Intel is in kind of middle a good wafer.

This is really wishful thinking.
You are the backup plan and the second source. You should expect 10%.

They need 4-5 companies putting $1 Billion order outside of Intel products.

 

[hist78]

The Samsung 2nm PDK was a mess so this was not a PDK based decision. Tesla worked with both TSMC and Samsung on previous designs. I really think this is a Texas driven deal since Tesla is big in Texas which is why Elon Musk moved there. I also think Elon wanted control of a fab and only Samsung would do that. I hope it all works out because this deal reaches out to 2033/34. A lot can happen during that time in the semiconductor manufacturing world.

$16.5B / 8 years = $206.25 million per year

Apple paid TSMC roughly $22B in 2024
Nvidia paid TSMC roughly $14B in 2024
AMD paid TSMC roughly $8B in 2024

Intel needs that type of partner for 14A, not Tesla. Just my opinion of course.




Yes, there is a preliminary PDK for 14A out in the wild. Besides, Intel 18A is competitive with Samsung 2nm and TSMC N2. Intel also has better packaging than Samsung.

TSMC N2 vs Intel 18A vs Samsung SF2 Wiki - Semiwiki

Category TSMC N2 Intel 18A Samsung 2nm (SF2) Target Mass Production 2H 2025 (risk production late 2024) H2 2024 (internal); early 2025 for IFS customers Late 2025 – risk production Transistor Type GAA nanosheet (TSMC’s implementation) RibbonFET (Intel's GAA design) MBCFET™ (Multi-Bridge Channel...

semiwiki.com

"$16.5B / 8 years = $206.25 million per year"

Should be $2.0625 billion or 2,062.5 million per year.

 

[DanX]

Even if they have competitive tech their design rules are difficult to adapt to not to mention Samsung has Cheapest wafers TSMC has the best best wafers Intel is in kind of middle a good wafer.

This is what I mean when I said they have no idea how fab works.
If you are #2, let alone #3, I would want to have R&D at 10% cost.
You are not going to put 150% RD for a second source.
Which means I will choose Samsung over Intel every time.