Samsung loses Qualcomm contract to TSMC for making next-gen Snapdragon chips

[Skye Jacobs]

Editor's take: Samsung faces the challenge of balancing improvements in its foundry business with the demands of its smartphone division. The company's ability to navigate these complexities will be crucial for maintaining its leadership in both the semiconductor and smartphone markets.

Samsung Foundry has once again failed to secure a contract from Qualcomm for the production of its latest flagship mobile processor, the Snapdragon 8 Elite Gen 2. This setback marks another chapter in Samsung's ongoing struggle to compete with industry leader TSMC, which has been awarded the entire manufacturing contract for the upcoming chip.

The decision comes at a critical juncture for Samsung Foundry, which has been grappling with low yield rates in recent months. However, there are signs of improvement on the horizon. The company recently stabilized its 3nm production process, achieving acceptable yield rates, and is simultaneously pushing forward with the development of even more advanced 2nm chips. Samsung also has ambitious plans to produce 1.4nm chips by 2027.

Despite these advancements, Qualcomm's choice to partner exclusively with TSMC for the Snapdragon 8 Elite Gen 2 reflects the current state of the industry. TSMC's proven track record and reliability likely outweighed any potential cost savings that Samsung might have offered. This decision is particularly impactful given the widespread adoption of Snapdragon SoCs in flagship smartphones.

The loss of this contract doesn't spell the end for Samsung Foundry's aspirations. The company is expected to make another bid for the production of the Snapdragon 8 Elite Gen 3 when it enters development. To improve its chances, Samsung will need to successfully fulfill orders from other clients and further improve yield rates.

This development coincides with reports of a significant price increase for the Snapdragon 8 Elite Gen 2, which could lead to higher costs for premium smartphones, including Samsung's Galaxy series. This situation presents a dilemma for Samsung's mobile division as it may need to consider alternative strategies to maintain competitive pricing.

One potential solution could involve reviving Samsung's in-house Exynos chip line for future Galaxy devices, such as the S26 series. This move could help offset the impact of rising Snapdragon prices.

Alternatively, Samsung might explore partnerships with other chip manufacturers like MediaTek or make compromises in other components such as displays and cameras to offset increased processor costs.

Despite these challenges, industry experts anticipate that Samsung's upcoming Galaxy S25 series will remain among the top smartphone choices in 2025. The company's strength in other areas of mobile technology should support its operations as it continues to refine its foundry capabilities and chip production processes.

Permalink to story:

Samsung loses Qualcomm contract to TSMC for making next-gen Snapdragon chips

 

[yRaz]

Great, and now costs will go even higher because TSMC is getting to the point of being able to ask whatever price they want.

 

[Burty117]

Great, and now costs will go even higher because TSMC is getting to the point of being able to ask whatever price they want.

The thing is, or at least, as far as I can tell, TSMC not only worked hard to be where they are, they were kinda given the position as well, Intel failed to move on from 10nm, and Samsung just seems to always be struggling with yields.

Global Foundry were happy to sit on much older nodes, it just seems to me, not many foundries wanted to even compete.

 

[yRaz]

The thing is, or at least, as far as I can tell, TSMC not only worked hard to be where they are, they were kinda given the position as well, Intel failed to move on from 10nm, and Samsung just seems to always be struggling with yields.

Global Foundry were happy to sit on much older nodes, it just seems to me, not many foundries wanted to even compete.

I'm not questioning how hard they worked here or the skill of their employees. The thing is, the end result is going to end up in a monopoly on cutting edge nodes and we'll all suffer higher prices because of it.

 

[Burty117]

I'm not questioning how hard they worked here or the skill of their employees. The thing is, the end result is going to end up in a monopoly on cutting edge nodes and we'll all suffer higher prices because of it.

Only if the thing you're after is on a bleeding edge node, Just means now the "high end" of tech is slightly more out-of-reach to everybody.

Just like how 7nm/6nm and even 5nm is now trickingly down to much cheaper and budget orientated products, it'll just take longer for 3nm onwards to get cheaper.

To be honest, I don't see it as too much of an issue, because we're not getting giant performance leaps anymore, 5800X3D vs 7800X3D vs 9800X3D, yeah things are getting better but we're not talking old school days of 28nm vs 14nm type stuff.

If you can't afford the latest and greatest, the last gen item, that's now cheaper, probably gets the job done anyway.

We're now getting to the end game for Silicon, physics has kicked in, we can't squeeze out much more really.

 

[trparky]

We're now getting to the end game for Silicon, physics has kicked in, we can't squeeze out much more really.

Yep, now it's time for optimizations. Gone are the days where we can say... "Oh yeah, we'll just throw more hardware at the problem." And we know that a lot of games today are unoptimized pieces of crap.

 

[yRaz]

Only if the thing you're after is on a bleeding edge node, Just means now the "high end" of tech is slightly more out-of-reach to everybody.

Just like how 7nm/6nm and even 5nm is now trickingly down to much cheaper and budget orientated products, it'll just take longer for 3nm onwards to get cheaper.

To be honest, I don't see it as too much of an issue, because we're not getting giant performance leaps anymore, 5800X3D vs 7800X3D vs 9800X3D, yeah things are getting better but we're not talking old school days of 28nm vs 14nm type stuff.

If you can't afford the latest and greatest, the last gen item, that's now cheaper, probably gets the job done anyway.

We're now getting to the end game for Silicon, physics has kicked in, we can't squeeze out much more really.

I have 13 computers in my homelab, most are from between 2015-2020. Yeah, it's fine for 99% of what anyone needs to do. It's the reason that AMDs I/O die is made on a different node than the chiplets.

 

[Burty117]

Yep, now it's time for optimizations. Gone are the days where we can say... "Oh yeah, we'll just throw more hardware at the problem." And we know that a lot of games today are unoptimized pieces of crap.

Absolutely, when it comes to gaming specifically, it's why RT will become mandatory when the next generation of consoles comes about. Rasterization performance has hit a wall, and the only way to improve it at this point is to simply put more cores in, which is now more silicon, which is now more expensive than ever.

RT on the other hand, those RT cores have been doubling or trippling in performance every couple of years, give it a bit more time (probably just past 2030 at the rate things are going) and your average GPU will be handle RT no problem.

BUT, I do absolutely agree with you on unoptimized crap game devs are spewing out these days, there's dedicated YouTube channels now calling this nonsense out:

Some (not all) game devs really aren't trying very hard.

I have 13 computers in my homelab, most are from between 2015-2020. Yeah, it's fine for 99% of what anyone needs to do. It's the reason that AMDs I/O die is made on a different node than the chiplets.

Yeah exactly that, Really, higher end stuff getting more expensive only really affects those willing to pay for it currently, Plus these companies will need to do some math and decide whether selling less at a higher price (and a slight hit to your reputation) is worth it, or, instead of making 150% markup on it, lose a little bit on the bottom line but sell more.

Remember, none of these companies sell anything even remotely close to a loss, I'm sure it was calculated a couple of years ago but a 7950X cost somewhere in the neighbourhood of £70 to make, that includes packaging. I'm fully aware you need to add R&D costs, Marketing, Shipping etc... But you get the idea on the markup of these things. Look at Intel and the B580 pricing, they definitely aren't making 150% profit on that, probably closer to 50% since they've really gone in on the value side.

 

[Theinsanegamer]

You know, I'm really surprised AMD never made the jump to Samsung. TSMC is very expensive and if they are not building big GPUs, the consumer stuff doesnt NEED to be on TSMC, which could then be more dedicated to CPUs and AI GPUs.

sure, it's not as good, but its still serviceable (see ampere) and a LOT cheaper, which would help with AMD's 2-3% margins.

 

[kiwigraeme]

I saw a video the other day on YT by Chinese Observer - a very well funded, anti PRC propaganda unit

Start watching about 2minutes 20 seconds, When it shows all the companies and machines Micron use in Taiwan to produce their stuff. Just Japanese, european company after company + some other countries thrown in .

No matter what if you care or not for anti-PRC propaganda. Its really eye opening all the other companies needed to produce this stuff

 

[Mr Majestyk]

Global Foundry were happy to sit on much older nodes, it just seems to me, not many foundries wanted to even compete.

Gloflo serves a different market that does not require in anyway bleeding edge proicess nodes. The world is awash with appliances and cars that can get all their capabilites form mature and dirt cheap 12-28nm processes. TSMC does not serve that market. GloFlo would need tens of billions dollars in capital to even try to compete, something no one would fund given it would take them at least 5 years to get fabs built and validated and there would be no guarantee they wouldn't have similar troubles to Samsung.

 

[Burty117]

Gloflo serves a different market that does not require in anyway bleeding edge proicess nodes. The world is awash with appliances and cars that can get all their capabilites form mature and dirt cheap 12-28nm processes. TSMC does not serve that market. GloFlo would need tens of billions dollars in capital to even try to compete, something no one would fund given it would take them at least 5 years to get fabs built and validated and there would be no guarantee they wouldn't have similar troubles to Samsung.

This was pretty much, exactly my point, every company is so risk adverse, they gave TSMC the top spot, TSMC also worked hard to get there, but nobody bothered competing with them either.