TSMC Arizona chip plant will be a paperweight | Hacker News Reader

TSMC Arizona chip plant will be a paperweight

9to5mac.com

72 pointsironyman45 comments

I'm not an industry analyst, but the packaging issue seems a bit overblown.

Yes, packaging for this kind of high density design as an iPhone is very specialized, but it does not contain the precious IP and the security concerns it carries.

Also, iPhones and any costumer electronics at this level of complexity and scale will be assembled in China for the foreseeable future, and this is more like a stop in the middle of a journey than a round-trip.

Maybe this website is too Apple-focused, or maybe I'm missing something, but "older" technology makes the bulk of the ICs manufactured and is much more strategic than the latest Instagram filter accelerator processor. Maybe by value bleeding-edge nodes are bigger, but by volume they are not even close.

By the way, I won't miss if my car comes without infotainment with the latest GPU, but I surely hope the ABS processors are there. Or even the shitty microcontroller that every single button seems to hide nowadays.

And for most applications that do not require incredibly high density, there are plenty of packaging plants in the US (I google and got this [1] as a first result. Sorry, not an analyst). And even the advanced packaging seems to be getting attention that this article does not acknowledge [2].

[1] https://www.globalspec.com/local/4218/C_US [2] https://semiengineering.com/expanding-advanced-packaging-pro...

Facts known from the start seem to flow slowly into the public knowledge.

Arizona fab was always going to be a MegaFab, not a GIGAFAB® in TSMC terminology.

MegaFab's (25K+ WSPM/month) are for medium value, flexibility, cost, delivery precision, customers. US defense contractors for example. Not for some massive customers like NVIDIA or Apple.

Having worked in the semiconductor manufacturing industry for a decade, albeit a while ago now, this article seems a bit overblown. The packaging industry was moved to low-cost areas long before the semiconductor fabs were; the fab is generally considered the high-tech part. Not that packaging isn't important or specialized or difficult in its own way, but it's done in lots of countries, and it is not in the slightest unusual for it to be done in a different country than the wafer fab. Usually (though by no means always), a country with lower cost of labor.

I’ve heard that the problem isn’t the lack of skilled engineers in AZ, but the union trades. At Intel they have to prepare the location for an electrical pull three days ahead of time. The pull must be between 9am and 4pm, M-F. I expect the unions are making similar demands at TSMC, and the Taiwanese, just like every other American business owner are pissed. However the Taiwanese think that bringing over people to do the trade work themselves is going to improve the pace. That’s going to result in a plant which will be continuously sabotaged by the unions. What really needs to happen is the AZ Democratic Governor (or the two Democratic senators) need to step in and negotiate with the Unions. She probably doesn’t realize this is the most important issue in the state right now, and that she can make a difference.

Negotiating with the unions doesn't make any difference if there fundamentally aren't enough workers in the labor force at the currently needed skill levels. If they want more workers, you have to actually look at training programs or pulling from wider regional workers - maybe negotiate with union to take on more apprentices and accepting regional union members for the surge.

Scheduling work between 9-4 M-F is the reason for massive delays and budget overruns? Union tradesmen purely to blame?

Here’s a comment from another article linked in the OP to give another side to the story:

“I have intimate knowledge of this project from the beginning. There are multiple causes for this project being behind schedule and over budget. They all stem from TSMC themselves.

First, TSMC and their engineers are absolutely clueless about building to meet U.S. code. Within 8 weeks of the project starting, they were firing the Tiawan based engineers because they were clueless and could not communicate with the U.S. based partners.

Second, they are behind schedule because THEY OWE HUNDREDS OF MILLIONS OF DOLLARS TO SUBCONTRACTORS AND ARE NOT PAYING. Yes, HUNDREDS of MILLIONS. Contractors have stopped work because they have not been paid. No wonder they want to bring in out of country labor. They won't have to worry about paying anybody. I literally have been told "this is not how we do it in Taiwan. We just settle up at the end of the project."

This should be an insult and slap in the face to all U.S. taxpayers, not just the contractors and workers who are being taken advantage of by TSMC and our local officials who are supporting them and their actions. Yes, Mayor Gallego, Senator Sinema, Senator Kelly I am calling you out. Keep telling everyone how great a deal this is for us. So TSMC wants $50 billion from U.S. tax payers, they don't want to pay hundreds of millions of dollars to contractors who have done work, and now they want to bring in their cheap slave labor so they can circumvent the work stoppage by contractors they wont pay. Everything I said is true and verifiable.”

https://9to5mac.com/2023/06/29/tsmc-arizona-chip-plant-delay...

>"this is not how we do it in Taiwan"

Reminds me of the movie Gung Ho where the Japanese are trying to build cars in 80's America.

https://www.youtube.com/watch?v=hKNeFHBPgRo

or Chinese LCD factory in Ohio https://www.imdb.com/title/tt9351980/?ref_=ext_shr_lnk - American Factory

Is it an accurate documentary, or is it some over dramatized Netflix piece like Drive to survive?

I'd like to see other sources that they don't pay for subcontractors. I don't think 9to5 and its comments are good sources for semiconductor manufacturing.

At the same time, Japan's TSMC fab (JASM), building is on schedule and don't see such problems. Maybe lower salary and similar work culture contribute flawless building.

Yes you're probably right. The governor of Arizona doesn't realize what the most important issue in her state is right now, but you do.

She’s a politician. The most important issue in her state right now is the next election. That’s true for every politician.

That's what they keep saying, but other projects seem to be able to get stuff built just fine if they are paying market rates.

From what I hear, they are being extremely tight with their budget, while simultaneously demanding aggressive timelines. These things are for the most part, mutually exclusive.

Importing 500 skilled Taiwanese workers seems like a no-brainer. I don't understand this mentality that those wouldn't count as "US jobs". They would massively lift the local economy.

The article also seems incredibly short-sighted, but since it's coming from an Apple clickbait farm that's not too surprising.

Importing 500 skilled Taiwanese workers is a no-brainer business decision. But American politics is motivated by strange nativism, e.g. "They're taking our jerbs!"

I don't like the reflexive union bashing in the comments, but for some documentary perspective (which to some degree reinforces the objections of the people whose comments I'm complaining about), check out the film American Factory (https://en.wikipedia.org/wiki/American_Factory).

I rarely like documentaries but I liked this one.

I still don't know how they are supplying water to the thing, we don't have much water out here as it is...

It'll just become a issue later and another reason they can't be more of a thing there.

Likely another unsustainable pipeline off of the Colorado river if I were to venture a guess.

Isn't nearly all the water used by a fab recycled and reused?

There's plenty of silicon in the world that's not designed by AMD, Nvidia, or Apple. Recall, if you will, the 2021 chip shortage.

But it isnt as sexy as leading nodes, so we dont talk about it

Everyone knew that this was just a PR move. This was the intended result. Some policy wonk somewhere thought that this was all on the up-and-up, but everyone else involved, from Apple to TSMC, and all of the mid level politicians that showed up for the photo ops knew that this was virtue signaling for votes, kickbacks, and subsidies/tax breaks.

I think you might be giving mid level politicians too much credit

yeah, good point.

On-shoring manufacturing isn’t “virtue signaling”, it’s of massive strategic importance.

I, for one, think it’ll work out.

No. This was a strategic move.Strategy(specially when it works) looks superfluous but it is not.

There are strategic fields like food or semiconductors in which you need to assert independence or you will be blackmailed.

E.g Germany recently made strategic mistake using the gas from the Russian pipelines AND not building alternatives.

Had Germany built energy plants as alternatives, they would have not been used at all(because is is usually more expensive), so people like you would have called them "PR move".

Had Germany built alternatives that they could use, Russia would not have been as emboldened to invade Ukraine, and if Russia would had been stupid enough to invade anyway, Europe would not have paid the start of the war to Russia with overprices and Ukraine could have been given heavy weapons way sooner.

Europe has wasted tens of billions of euros by making such strategic mistakes.

The US is building semiconductors plants in the US. The objective is not to build semiconductors in the US if it is moreexpensive, the objective is be able to create semiconductors very fast if China invades Taiwan.

>Germany recently made strategic mistake using the gas from the Russian pipelines AND not building alternatives.

Recently?! The gas dependency was set in the 60's

Given the Chinese stated goal to recapture Taiwan and its recent mandates of state ownership, TSMCs reliance on Taiwan soil facilities is an existential threat to their business as a private enterprise. On the one hand keeping the semiconductor dependent on Taiwan solidifies their defense by being critical to global supply chains, on the other that may not be enough. As a business they need to prepare for the loss of their major manufacturing facilities. I don’t think this is a photo op, it’s a first step in a decoupling and the establishing of an off shore base of operations outside of their main base.

I find it sad issues like visas for Taiwanese experts are at the fore here. I would look at as an opportunity to bootstrap local talent through collaboration and some percentage of those visas will turn into citizens and permanent residents, aka Americans.

They aren't asking to bring their PHDs or material science people... my understanding is they are asking to bring project managers, foremen, construction workers, pipe fitters and electricians.

I live about 10 minutes away from the plant and anecdotally from what I can see of the Taiwanese employees, most seem to be people who are in jeans and cotton work shirts... not button downs or polos.

This deglobalization thing doesnt seem to be going so well.

Used to be that the US would just import scientists and engineers then develop stuff themselves. It'll be interesting in a few years to learn if this Arizona plant was just one head of the capitalist hydra and its success or failure didn't really matter all that much. I suspect that the US doesn't need to import technology in that manner to get good results; it never has in the past and I don't think that much has changed.

All the excitement over high end chips seems overblown. We appear to be talking about a couple of years R&D. Tactically interesting, it'll decide a lot of things in practice about who specialises in what. Strategically not that big an issue. The trade war manoeuvring is doomed to look silly; the US can't stop China and China couldn't stop the US anyone thought there was a real problem to solve.

I really don't think you can just import scientists and engineers and essentially develop TSMC. It's not the sixties anymore, Intel has a massive R&D budget but they are not currently able to compete with TSMC. Technology is also quite advanced, I doubt it's just a few years of R&D. In fact, it's so expensive, there are only a few factories on earth capable on producing those chips. China would have done it already if it's so simple.

> I really don't think you can just import scientists and engineers and essentially develop TSMC.

I think it's more a question of being willing to spend the money you need to stay at the cutting edge. Based purely on market logic, you can't build a TMSC because it would be too expensive and risky for modern-day capitalist investors to contemplate.

But if you have resources and don't care about making a profit in the market (like a government), you can do it eventually.

> It's not the sixties anymore, Intel has a massive R&D budget but they are not currently able to compete with TSMC. Technology is also quite advanced, I doubt it's just a few years of R&D.

I don't think Intel is "not able" to compete, but rather they took a costly mis-step. IIRC, TMSC took a more slow-and-steady approach, while Intel was too ambitious and its plan failed, so now it has to backtrack.

> In fact, it's so expensive, there are only a few factories on earth capable on producing those chips.

IIRC, they basically build a new factory each time the technology advances, so that's not as much of a limitation than it sounds.

Seems like starting an automotive manufacturing company would be far worse as a prospect yet they are popping up everywhere. That automotive industry has always been a case study in great ideas failing hard. If you succeed you get to compete with everyone else who has to fight and scratch for a profit while juggling one of the most complex undertakings imaginable. There are 150 year old established brands like Chrysler that have/are failing now.

The cost of a fab for each new process node is roughly double the cost of the previous node.

That’s not true for cars. In fact, starting a new car company is probably cheaper than a few decades ago, since most part manufacturing has been outsourced.

> That’s not true for cars. In fact, starting a new car company is probably cheaper than a few decades ago, since most part manufacturing has been outsourced.

Also most if not all the new car companies are building electric cars. My understanding is those are (in many ways) simpler to make than internal combustion engine cars.

Chrysler is only 98 years old, not 150.

That was a very similar investment landscape when TSMC started up - it’s a real shame that US capitalists are so short sighted. Or maybe they’re too cheap to invest when they have control of government taps via lobbying. Why invest your own capital when you can lobby for government handouts.

Taiwan Government own 48% of TSMC stock at it founding. TSMC is not a product of free market capitalism. It is a product of the Taiwanese state.

Source: Chip Wars (2022)

Then it's a shame the not only US capitalism and world capitalism is incapable of investing in the long term. Though I'm fine with government support as long as the outcomes return a fair share of the benefits of the private venture back to the government.

A few leaders can pull critical people over with them and help speed up projects. Samsung and SMIC poach TSMC R&D leaders fairly frequently and it makes the news over there.

Half the reason US started the semi war against PRC was because they knew it's pretty much that simple provided work force is culturally compatible. Ditto with foreign fab projects expanding in PRC at the time - without requirement of crazy subsidies to draw them in (big funds was for domestic indy). Industry assumed, PRC with right equipment and a few years of poaching TSMC, Samsung and whatever incumbent talent would be sufficient to get ball rolling. SKR basically did that from TW talent. IMO US can replicate east asian fabs quickly if it can replicate east asian work culture because that's the only way east asian fab owners knows how to run said fabs. They're not interested changing, because they weren't interested in expanding to US in the first place, since the business case never made sense. Huawei (possibly SMIC) 7nm did it in a few years with hands tie behind back.

> It's not the sixties anymore, Intel has a massive R&D budget but they are not currently able to compete with TSMC.

If it isn't a matter of budget then it is a combination of organisation, regulation and focus. All of those things can change fairly quickly. Under ideal conditions, Intel could get to TSMC's leading process node in a few years.

Why the need to shoehorn "capitalist" in everything these days?

Reply on news.ycombinator.com