Samsung delays $37B Texas chip plant with no customers in sight

Alfonso Maruccia

Posts: 1,845   +556
Staff
Not Looking Good: Samsung is one of the few companies in the world with true chipmaking capabilities. However, the foundry division of the South Korean conglomerate is struggling significantly in its efforts to expand processor manufacturing operations in the US.

Samsung is reportedly delaying the completion of its chip manufacturing plant in Taylor, Texas, due to a lack of customers for its foundry services. According to industry sources cited by Nikkei Asia, the South Korean corporation's ambitious plan to launch a major new US foundry operation is already backfiring.

The facility was originally slated to open in 2024 but has now been officially pushed back to 2026. Documents from Samsung C&T, a Samsung affiliate overseeing construction, show that the project was nearly complete as of March – reaching 91.8 percent completion. However, Samsung has not begun installing chip manufacturing equipment, as it would currently sit idle.

A supply chain industry source reportedly confirmed that Samsung is in no hurry to finish the Taylor plant. Local demand for advanced chip production is insufficient, the source said, and the process nodes originally planned for the facility are now largely outdated. To move forward, Samsung would need to significantly revamp the site's capabilities, an expensive undertaking with no guarantee of short-term return on investment.

Samsung initially planned to manufacture 4nm chips at its Taylor, Texas, facility but later announced plans to upgrade to a more advanced 2nm process node. The company originally committed $17 billion to the project, but over the past few years, that investment has ballooned to $37 billion across the broader Texas region. This includes $4.7 billion in subsidies granted under President Biden's CHIPS and Science Act.

However, the new Trump administration is actively working to dismantle the CHIPS Act and other semiconductor funding programs established during Biden's term. Compounding the challenges, Samsung has reportedly faced poor manufacturing yields with its 2nm process, another factor contributing to delays at the Taylor facility.

According to TrendForce analyst Joanne Chiao, Samsung's 2nm yields have improved, but the company continues to face internal and external hurdles in expanding its US foundry presence. The Taylor plant is expected to produce chips for smartphones, PCs, and consumer electronics, yet current demand in those sectors remains weak.

Samsung continues to trail behind TSMC in the global foundry market. The Taiwanese giant dominates production of advanced chips such as GPUs and AI accelerators, driven by strong demand from data centers and tech firms. While the Taylor facility may eventually support limited production thanks to government subsidies and tax incentives, Chiao suggests that Samsung's original ambitions are likely postponed well into the future.

Permalink to story:

 
Covid caused some sort of panic reaction about insufficient supply. Too bad, nobody realized situation was temporary and new fabs take years to complete. Now we have oversupply. Just like I and many others predicted.

Some predictions just cannot miss.
 
It doesn't help that the federal employees in charge of managing CHIPS funding have been fired.
I'm going to try to avoid being political here, but I have a hard time understanding what is going on with the CHIPS act and WHY. Trump dismantling the CHIPS act doesn't help bring chip manufacturing to the US, but he said he would and he hasn't replaced it with anything. The Chips act kinda sucked, but it was better than nothing, but not we have nothing and tariffs actually make construction costs for fabs more expensive than they were pre-CHIPS.

I'm not trying to argue, I just want some clarity on the issue. It's a really bizarre situation we're in now
 
Covid caused some sort of panic reaction about insufficient supply. Too bad, nobody realized situation was temporary and new fabs take years to complete. Now we have oversupply. Just like I and many others predicted.

Some predictions just cannot miss.
We don't have 'oversupply'. What we have is not enough production using modern lithography and this massive fab was built to produce 4nm.
 
We don't have 'oversupply'. What we have is not enough production using modern lithography and this massive fab was built to produce 4nm.
As said on article, even 2nm would not be guaranteed success. If there is really demand for modern processes, Samsung 2nm (probably roughly on par with TSMC 2nm) should have very high demand. But no.

TSMC, Intel and Samsung all have "modern" processes available and even once outsold TSMC 7nm is not that bad even today.

Also there was supposed to be shortage on not so bleeding edge processes but again, there is basically no real shortage of anything right now.
 
As said on article, even 2nm would not be guaranteed success. If there is really demand for modern processes, Samsung 2nm (probably roughly on par with TSMC 2nm) should have very high demand. But no.

TSMC, Intel and Samsung all have "modern" processes available and even once outsold TSMC 7nm is not that bad even today.

Also there was supposed to be shortage on not so bleeding edge processes but again, there is basically no real shortage of anything right now.
There IS a shortage, it is just on TSMC's newest nodes. Nobody wants their chips from Samsung fabs because their nodes are inferior to the competition. This has been the case over half a decade at this point.

And because Samsung already has volume problems with its order book, it can't afford to cut prices to compete.
 
There IS a shortage, it is just on TSMC's newest nodes. Nobody wants their chips from Samsung fabs because their nodes are inferior to the competition. This has been the case over half a decade at this point.

And because Samsung already has volume problems with its order book, it can't afford to cut prices to compete.
Well, if you can give sources, I'm happy to admit but https://www.tomshardware.com/tech-i...-cites-lack-of-us-subsidies-and-waning-demand

Basically 1.5 years ago TSMC decided not to put another 3nm fab or at least delay it because there is not enough demand. And at this point they don't anything better than 3nm on mass production. Does not sound like shortage for me.
 
I'm going to try to avoid being political here, but I have a hard time understanding what is going on with the CHIPS act and WHY. Trump dismantling the CHIPS act doesn't help bring chip manufacturing to the US, but he said he would and he hasn't replaced it with anything. The Chips act kinda sucked, but it was better than nothing, but not we have nothing and tariffs actually make construction costs for fabs more expensive than they were pre-CHIPS.

I'm not trying to argue, I just want some clarity on the issue. It's a really bizarre situation we're in now
As I understand it (well, that's a big word in this case as it makes no sense) is that tariffs are the answer to everything. A 100% tariff on chips not made in the USA.
Instead of luring companies to the US with incentives they get punished for being elsewhere. The whip instead of the carrot on the stick.
To be fair, I like the reasoning better. Paying some of the richest and most profitable companies (heh maybe not Intel) millions of tax payer dollars seems silly. But if I was TSMC or Samsung I think I'd rather charge US customers a lot more than the rest of the world and keep things as they are.

I was hoping with all the additional capacity being constructed that prices would drop hard but looks like to avoid that they simply don't finish the factories.
Imo if Trump wants domestic production he should have made deals for Intel like he's been making them for Boeing. Get Qatar or some other place to somehow pay for it.
 
Covid caused some sort of panic reaction about insufficient supply. Too bad, nobody realized situation was temporary and new fabs take years to complete. Now we have oversupply. Just like I and many others predicted.

Some predictions just cannot miss.

As I recall, the worst of the chip problems in the US wasn't with anything built on advance processes. Ford, GM, and the rest of the automakers, for a start, were either shutting down assembly lines or cutting options or parking cars half finished due to a lack of power IC's and similar chips. The reason for that is China has pretty much taken over the market for these products. TSMC Samsung, and a few other foundries are much like US companies that quit building products once the can't make enough money on them.

With the number of companies that fold or don't last, I've never quite understood the concept of making good money on a product and discontinue it because you're not making ENOUGH money.
 
Covid caused some sort of panic reaction about insufficient supply. Too bad, nobody realized situation was temporary and new fabs take years to complete. Now we have oversupply. Just like I and many others predicted.

Some predictions just cannot miss.

Yeah, once the supply backlog got addressed, demand returned to normal levels.
 
I'm going to try to avoid being political here, but I have a hard time understanding what is going on with the CHIPS act and WHY. Trump dismantling the CHIPS act doesn't help bring chip manufacturing to the US, but he said he would and he hasn't replaced it with anything. The Chips act kinda sucked, but it was better than nothing, but not we have nothing and tariffs actually make construction costs for fabs more expensive than they were pre-CHIPS.

I'm not trying to argue, I just want some clarity on the issue. It's a really bizarre situation we're in now
Donald instead wants to have more steel mills and phone assembly plants
 
Ford, GM, and the rest of the automakers, for a start, were either shutting down assembly lines or cutting options or parking cars half finished due to a lack of power IC's and similar chips. The reason for that is China has pretty much taken over the market for these products.

I work for a company that does some custom board designs; would you believe buying the capacitors we need are currently on a (checks supplier site) 54 month backlog?

Unfortunately, a certain fruit company gobbles up all the supply, and everyone else are second-class citizens when it comes to fulfilling orders.

With the number of companies that fold or don't last, I've never quite understood the concept of making good money on a product and discontinue it because you're not making ENOUGH money.

It's how modern finance works: There's only so much floor space to dedicate to producing a good. If that space could be reallocated towards producing something that could potentially yield *more* profit, then you obviously kill that product.

Seriously, floor space utilization is absolutely a metric used. My company recently threw out a couple million (probably north of $10M) of test equipment we've accumulated over the years for testing/troubleshooting our products. I'm currently trying to troubleshoot a problem for a customer, and guess what: No working logic analyzers (we have *one*; we used to have like 20). Because "the floor space wasn't being used to enhance shareholder value". (And yes, this is *exactly* the reason given).
 
The real cause is samsung 2nm yield is low.

nvidia apple etc. prefers to have options, at least near at tsmc quality.
apple can use it for non pro iphones.
nvidia can use it for 70 class geforce chips
 
I work for a company that does some custom board designs; would you believe buying the capacitors we need are currently on a (checks supplier site) 54 month backlog?

Unfortunately, a certain fruit company gobbles up all the supply, and everyone else are second-class citizens when it comes to fulfilling orders.



It's how modern finance works: There's only so much floor space to dedicate to producing a good. If that space could be reallocated towards producing something that could potentially yield *more* profit, then you obviously kill that product.

Seriously, floor space utilization is absolutely a metric used. My company recently threw out a couple million (probably north of $10M) of test equipment we've accumulated over the years for testing/troubleshooting our products. I'm currently trying to troubleshoot a problem for a customer, and guess what: No working logic analyzers (we have *one*; we used to have like 20). Because "the floor space wasn't being used to enhance shareholder value". (And yes, this is *exactly* the reason given).
And tsmc could buy expensive euv machines using money from that fruit named company.
Meanwhile Intel money bosses did financial "innovations"
 
And tsmc could buy expensive euv machines using money from that fruit named company.
Meanwhile Intel money bosses did financial "innovations"
They could, but then prices would drop due to the increased supply, and leave then with an excess of equipment once they move on to the next process node. Both bad for TSMC.
 
Back