Hacker News new | past | comments | ask | show | jobs | submit login
Semiconductors Demand 30% Above Supply, 20% Year-on-Year Growth (anandtech.com)
198 points by ksec on Feb 25, 2021 | hide | past | favorite | 165 comments



Equipment companies are running at an all-out sprint to ship fabrication equipment. Applied Materials, the world's largest by revenue, has 1,390 open positions. #4 Lam Research has 570.

The equipment space is brutally cyclic, akin to the oil services industry. Revenue can jump as much as 60% year over year, only to plummet 40% the next.

It's always worrisome to see these macro-trends, feasting now, but famine sure to follow...


I think this is the problem. The downcycle for the semi industry should of came around 2-3 years ago. The issue is that never came and the industry is growing at an even more ridiculous rate. The semi industry is single handed buying a good chunk of the OEM equipment space these past 3 years. Just get a THK rail quote and look at awe at the one year+ lead time.


This plus secular trends due to the shortage _itself_ inducing further future plant development at home, simply for redundancy as a hedge against tail risk...think we're closer to the start of a 5-10 year boom than the end


There was a huge boom in equipment in 2017 the 3D NAND demand thundered in. Semis over-invested as they frequently do, and the equipment biz then softened in 2018-2019. 2020 was solid and now 2021 is off to the races again, thanks (sadly) to the technology accelerating effects of COVID.


Are you working in toolmaking by any chance?



An actual RnD, or engineering?

In the link you point to the need of "EU fab." Who do you think will buy its chips?

Germans for their cars?


System engineering. We sell subsystems and components to the equipment companies.

Chips are sold globally for the most part, regardless of where they are made. EU, US, Japan, Taiwan & Korea made chips are typically leading edge and sold globally. Some China-made chips are only used in China, but they have a large export market for low-cost chips.


> Chips are sold globally

Well, they aren't. I see really a lot of people even inside the industry not noticing the elephant in the room.

By some digits, China is 60%+ of the world semiconductor market, and much higher than even this digit if you remove secondary products from the calculation.

> System engineering. We sell subsystems and components to the equipment companies.

What kind of? Do you have a niche market, or you sell whatever toolmaker want to custom order?


For example, nVidia's chips are made in Taiwan and Korea, and end up in devices sold globally. It's true that a large % of the assembly of circuit boards and electronic products are made in China.

We make control system computers. It's very, very niche. Typically, what we sell to one equipment company, we can sell to others. It's very difficult for companies that are outside of the sub-supplier ecosystem to break in. The controllers are usually customized to some extent, then we sell a few hundred annually of that design.

'A few hundred a year' is a home run product.


Actually, I am surprised that EU isn't competitive in semiconductors. The value of the euro is pretty low considering that you get access to a high quality workforce that ought to get at least 50% higher salaries. This is how Germany has become an export nation. Adopt a currency whose value is being pushed down by neighboring countries. This makes German exports more competitive in the global market. Why has this not worked out for semiconductors?


why is the semi industry so cyclical? do you see this changing?


The Semiconductor industry is driven by major market economic forces, like most. It's not terribly volatile. If you sell capital equipment to the semi makers, though, there's a compounding effect that makes the annual revenue for equipment makers very volatile. There are few Semi players, and their capital spend $s are huge and 'lumpy'. As in $10B for a new, state-of-the-art fab. 70% of that $10B is spent on capital equipment. There are very few players that can spend at that level, Intel, TSMC, Samsung & the Chinese government. What you wind up with is the spend schedules being grouped together in time. I.e., they all buy at once. Or, they all stop buying at once.

The most recent roller-coaster peak was actually driven by technology innovation. NAND Flash chips were made via a '2D' process until about 2016 when '3D' processes were perfected. That meant many more bits of flash could be built for the same real estate on the chip. It also required many more deposition and etch process steps for the same output. This drove Samsung, Micron and Toshiba to all spend huge $ on equipment at the same time in order to capture market share in this new tech. Then, they realized they'd overbuilt and turned off the spend.

The end result: Our business was +90% 2016-2017. Then -60% by mid-2018.

We thought it had stabilized from 2010-2016. But in reality, the consolidation of players simply made the spend numbers larger and we got clobbered by 3D NAND. Next up, COVID. There will be a cliff, only question is when.


fascinating. thanks for sharing. this explains why it's so lumpy. why hasn't someone tried to consolidate at the equipment layer and counter the forces from above?


> Applied Materials ... has 1,390 open positions. #4 Lam Research has 570

With this many openings they must be having problems actually getting stuff done. But as you suggest, by the time they get a good chunk of these openings hired for they'll be looking to cut heads again. People who have been through one cycle before probably aren't inclined to go through the buzzsaw again and so move onto other industries.


Maybe they are just planning on enhancing their manufacturing capabilities. Quote from their Quarter call: "Applied Materials is also trying to tap into efforts to bring more chip manufacturing to the US and reinforce the nation's leadership in chips."

And the majority of the job openings are CA and TX


The semiconductor shortage is being amplified by the US-China tech sanctions.

This will create a massive overinvestment in the sector where China will build capacity to meet its own need and gain independence of US controlled supplies.

At the same time western countries will invest massively in the sector to keep its technogical lead.


Packaging is hot. The question how hot it is?

Buddies of mine want now explore a niche of very low end litho market for a brand new entry.

I want to throw them an idea of capitalising on very cheap interposers, but don't have much confidence in how long advanced packaging will be able to attract hot money.


Given the strategic importance of semiconductors and how expensive new fabs are and how long they take to start production (if you start planning a new state of the art fab now it'll take ~5 years before it's ready to produce chips) it seems like perhaps we need to look a public-private partnerships here in the US. The Chinese government is funding fabs. While we tend to be squeamish about such things here in the US, it's likely we need to start looking at doing this. At the very least the US Gov could step in to capitalize two or three companies to start foundries similar to TSMC. When they get on their feet they would become fully private entities that would pay back any government loans as they become profitable. I'm not sure we have the time to wait to see if Intel will decide to go big into the foundry biz - they seem to vacillate on this pretty regularly.


The DoD's Industrial Capabilities report to Congress[1] that got posted on HN last week recommended exactly this:

> "Outside of the United States, foreign governments and their citizens pay the lion’s share, one way or another, of the cost of building the fab. The companies do not. They take on the other massive set of costs: running the fab. The hard truth is that if the United States does not start doing the same, our nation will continue to see its historically low share of chip production continue to decline to irrelevance."

[1] https://www.businessdefense.gov/Portals/51/USA002573-20%20IC...


The Chinese government is willing to spend what it takes to win and this is the reason why they will win. The US government is way too reluctant to spend in this regard. The Chinese government is investing hundreds of billions in artificial intelligence, hundreds of billions more in renewables, and so on. I don't see Congress giving a blank check to scientists and engineers, like China does.


Maybe eventually, but at the moment they lag some generations behind US/Japan/Taiwan/Korea. This, in spite of the fact that China has been dumping money into the industry for years.

There’s a narrative in the west that China is “willing to spend what it takes to win”, but at the same time, China’s spending is tied up by internal politics (also, corruption, pork, kickbacks) as much as US spending is. China is also, for example, willing to spend as much money as it takes to make a mega-blockbuster movie and yet keeps churning out flops (again, due to causes that are intimately connected with internal Chinese politics).

If you look at major US boondoggles like the F-35 program you might think of it as a partial success (the plane works) and yet a massive waste of money (we could have spent the money better), and some of these massive initiatives in China suffer from the same problems.


Exactly. Most narratives hang everything on simplifications like “willing to spend what it takes to win.” Reality is more complex and there are serious drawbacks to any real life examples.

A better comparison mould have been circa 2000 "asian tiger" policies of south korea, taiwan, etc. The idea was to subsidize industries with massive growth potential... on a 20-50 year horizon. On that sort of a timeline, F-35 program might be considered an ok success. More expensive than necessary, but better than not having invested. For this to make sense, fighter jets need to be a massive growth industry, and adjacent to other massive growth industries. Not sure if the US has appetite for generational investments in industries.

A lot goes into success and failure at something like this than the "correct theory" though.

Part of the idea behind letting the semiconductor industry decline is that it frees up capital and human resources to work on higher ROI stuff. It's hard to get a high ROI manufacturing, because manufacturing has a high "I." Building factories is capital intensive.

It's possible to make a decent case that it succeeded too. The US absolutely dominates the high ROI, high margin industries: Software, financial services, pharma. Design... Intangibles. You can't make a money machine like FB, Pfizer or JPMorgan if (a) marginal costs are real or (b) you need to invest capital in order to grow.


> Part of the idea behind letting the semiconductor industry decline is that it frees up capital and human resources to work on higher ROI stuff.

Yeah. That’s a very good point, something I haven’t really thought of. I think we could also invest more in semiconductor design, both in making more sophisticated designs and in making semiconductor design accessible to more people (both of which are already happening). I’m talking about “design” in contrast to manufacturing/process concerns. If the process changes slow down a bit, I wonder if we can get ASICs with lower NRE.


> The Chinese government is willing to spend what it takes to win and this is the reason why they will win.

But it's Taiwan that is doing the chip fabs in TSMC, not mainland China. While they do have some capacity in the mainland, it's best to not conflate the two when discussing the geopolitical situation.


I dont think the parents meant TSMC. China is indeed spending a hundred billion on establishing a Semi Conductor supply chain industry. If you count from 2013 ish, they expect to spend a total of $150B+ in 2025. From NAND, DRAM to General Fabs and other up and down stream in the supply chains. That number was projected in 2017/2018. I would not be surprised if they are pouring in even more resources now.

Luckily the Moat around leading edge, cost competitive DRAM / NAND and Modern Foundry is much much larger than I thought. Hopefully the world is not too late to react to it.


Well we did it for the F-35.

Defense pork/corruption is a liability in economic war. We need to wake up and realize that China is kicking it into high gear.


Do note that there is also plenty of pork/corruption in China. Massive amounts. We just understand the US pork/corruption better because we can read English-language news sources.


And because the English language news sources are allowed to publish such things before the government approves them.


The F-35 seems to be selling like hotcakes, from looking at the orders american allies have for the plane. Lockheed is essentially a wing of the department of defence and lot of the "project overruns" are actually dollars spent on R&D and technology development. Defense pork is bad, but giving billions of dollars to aerospace, electronics, and radar engineers is a fantastic way to maintain an economic edge, especially when Lockheed, SpaceX, and Boeing dominate the global market thanks to US government funding


> I don't see Congress giving a blank check to scientists and engineers, like China does.

What are you on about? Have you seen the defense budget in the us? It’s massive by any metric you look at. A huge chunk of that is pure R&D.


The problem is that when you have political allocation of capital, you create a very inefficient economy. Sure, if the goal is to copy someone else that has proven a successful approach you can make good progress with subsidies. But after you drive them out of business, you are stuck with this politically directed R&D agenda that has no one else to copy but is not able to innovate on its own. You quickly lose your way with inefficiencies.

This must be weighed against the fact that subsidized businesses will outcompete less subsidized businesses, and so if you want to bankrupt your foreign competitors you can do that by subsidizing your own domestic firms.

One approach is to not trade or impose targeted tariffs on government subsidies to help balance the scales. That was basically the idea of the 90s era trade agreements, which hasn't worked out so well, as targeted tariffs don't work. You apply a broad, long lasting blanket tariff on all goods coming from the offending country or nothing.

Another approach is to jump in with massive subsidies of your own and create a race to the bottom where every government has to subsidize any industry that some other government is subsidizing.

I'd like to throw restrictions on foreign capital inflows into the mix. A combination of banning capital inflows, some subsidies, and some broad, permanent tariffs against bad actors like China might be the best option to deal with the "spend anything to win" problem. Really we don't know.

What I don't want to see is governments rushing in to subsidize industry in a race to the bottom. That is bad for innovation as well as for democracy.

What the free trade crowd needs to realize that free trade is vulnerable to the subsidize-attack which can destroy your own industries, and right now they have an ideological commitment that is blinding them to this reality.


Yeah, for all the worry about (digital) semiconductors, the US’s solar cell capacity has atrophied. There are a few large thin film fabs, like First Solar, but NO operational large silicon cell fabs now that Panasonic has pulled out of the Tesla Buffalo factory (I hope Tesla replaces them) as part of their global reorienting away from cell manufacture. Violet Power is hoping to change that, but it’s remarkable how much the US relies on imported solar cells, with almost no domestic capacity.

Makes sense as the cells are a commodity and extremely cheap (7¢/Watt...). They’re a minority of solar panel costs (15-30¢/Watt) which are themselves a minority of installed solar costs (70-100¢/Watt utility, above $2/Watt residential), but it leaves the US heavily reliant on imports for the Green New Deal or related initiatives.

Violet power is interesting as they’re setup across the street from a shuttered polysilicon plant in Washington state IIRC. They’re planning to be vertically integrated in panel manufacture (& eventually use that silicon plant) for cells with extremely high efficiency (20 and eventually 30%) cells with a 50 year warranty lifetime (double the normal 25). It’s a good approach for a US company, IMHO. They want to eventually get to 5GW/year output capacity. Which sounds like a lot but is minuscule compared to what we need.

To transition to electrified everything, the US needs 1TW average electricity. We currently produce ~500GW average, 300GW of which comes from fossil fuels. Nameplate solar capacity is only about 10-20% of average electricity output (depending on lots of factors including seasonal effects), so to replace the 300GW average of fossil fuel electricity, we’d need 1500-3000GW of solar nameplate. To electrify everything, we’d need an additional 500GW average, for a total of 4-8 Terawatts of additional nameplate solar.

To do that over 20 years would require 200-400GW per year of solar cell and panel capacity output. And this is just the existing US demand. We need to be investing a ridiculous amount in solar factory capacity RIGHT NOW if we’re serious about climate action.

(Luckily solar panel factories only cost about 20¢/(Watt/year) in capital cost even in the US... maybe about the same for combined cell/wafer/silicon capacity for a total of 50¢/(W/year), so 200GW/yr vertically integrated factory capacity may only cost $100 billion to build out which is not that much when you think about it... Except we aren’t doing it.)


Look at aircraft manufacturing in China for a counter-example to the "winning" part.


> willing to spend what it takes to win and this is the reason why they will win.

Excuse me

"Win" what exactly? Why even use such adversarial polarizing language when we are talking of overall human development?

Huge segments of the US live in economic precarity, do you feel like you have "won"? After all that was the point of the Cold War, wasnt it?


The US has never won any fight by their government being great. They've won a lot of fights by US capitalists grinding their opponents into powder and selling the remains for fertiliser.

If the governments go head to head, China wins easily. Their government is stuffed with PhDs, scientists and engineers. The US government has lawyers.


China's government has a lot of people with engineering degrees. That doesn't mean they're engineers, they were aiming for party positions from birth.


Here in the USA we're more focused on the important things, like passing identity politics legislation.


Turing would be delighted to read this.


As someone protected by that legislation who also works in aerospace operations: good.


I'm curious is gender identity and/or orientation [1] not already a protected class where you work?

[1] https://en.m.wikipedia.org/wiki/Executive_Order_13672


Of course, so important.


To the people it protects it’s extremely important.


...citation?


Just hours ago [1] H.R.5 — 116th Congress (2019-2020) was passed in the house. Also known as the Equality Act. It appears well intentioned, great name, but some of the demands are gonna be a challenge, specifically, bathroom accomodations, and potential to disrupt traditional women's programs.

[1] https://www.congress.gov/bill/116th-congress/house-bill/5/te...


Seems pretty unlikely to make it through the us senate no?


Who knows. House was narrow so senate will be a battle.


Yep, I work in hardware design for defense. They dropped the Cadence license (no RFIC design for me), but I got to take Unconscious Bias training yesterday.


> I'm not sure we have the time to wait to see if Intel will decide to go big into the foundry biz - they seem to vacillate on this pretty regularly.

Historically, that's because the "foundry biz" has been small. If you check the revenue numbers in that chart, you'll note that TSMC is now larger than the rest of the foundry industry combined. They brought in almost $13B last year.

Intel 2020 revenue was $78B.

There just hasn't been incentive to go into that market. At best it would be a way to squeeze a few extra percent out of idle fab capacity. Now, obviously the trajectory here is not good for Intel. But that's because of technical details of fab process, it has nothing to do with their decision not to chase the foundry market.


> They brought in almost $13B last year.

Oh wow that's tiny for the largest foundry. And I gather that they are not just in the fab business but trying to integrate vertically into being an OSAT as well.

Edit: I mistook the number as revenue, not earnings. For apples-to-apples comparison, TSMC's annual revenue is closer to $40B.



> it seems like perhaps we need to look a public-private partnerships here in the US.

Wouldn’t be the first time... https://en.m.wikipedia.org/wiki/SEMATECH


Are you saying that Chinese semiconductor companies are outcompeting US semiconductor companies because they have greater access to capital? That doesn't seem right to me.


> Are you saying that Chinese semiconductor companies are outcompeting US semiconductor companies because they have greater access to capital? That doesn't seem right to me.

Suppose you have two companies in an industry where the start up cost is ten billion dollars. One can borrow ten billion dollars at 0% interest for the next five years, but they have to pay it back. The other just gets ten billion dollars from the government that they can keep.

Obviously the second company has a competitive advantage. The first has to charge higher prices in order to pay back the loan, which they can't do if the second undercuts them in the market.

This has been China's business model for decades. Then once all the competition is out of business they can raise prices to the point of profitability, but not to the point that anyone who hasn't already paid off their fixed costs can enter the market, and use the profits to do the same thing to the next industry.


Isn't that equivalent to starting up with VC money (and stocks, later)? You get billions to start up, and if you fail you don't have to pay anything back. It's the business model of Amazon, Uber, Tesla, Airbnb...


All of your examples are companies that were pioneers in immature markets. And then if you succeed, the VCs own your company. It's not charity.

Whereas China will give a Chinese company billions of dollars to unseat an incumbent in a mature industry and expect nothing in return except for that industry to be in China.


> and expect nothing in return

Aside from the expectation of unwavering loyalty to the Communist Party of China, you mean.


Is that not what I said?:

> except for that industry to be in China.


Fair, but the implication seemed to be that China was somehow not expecting ownership over the corporations in which it invests (since there was an implied contrast with VC investment where said ownership is expected).

That is: it goes a fair bit beyond merely keeping companies in a given geography.


Are you saying that the Chinese government is subsidizing the entire semiconductor industry to drive US companies out of business?


Are you saying that the Chinese government is not subsidizing the entire semiconductor industry in China to drive companies outside of China out of business?

They have a documented history of doing this in multiple industries over the course of many years. Notice how substantially all of the consumer electronics devices that used to be made in the US or Japan or Korea or Germany are now made in China.


There's a big difference between receiving an amount of government subsidy, and the government funding almost everything with nearly no budget limits.

The former is definitely true, though China is not the only country engaging in such a practice.

The latter claim needs sources to back up.

If you look at the Chinese chip industry, then you see that a lot of money comes from the private sector. Huawei, Xiaomi, etc are investing a ton of money in semiconductor companies. SMIC got a lot of funds through a stock market IPO. Clearly it's not all government money.

There's also a big difference between investing in an industry with the goal of dominating the world and eliminating competition, vs the goal of becoming a sovereign equal. Why is it necessarily the former, and not the latter?

The reason why the Chinese government invests in many sectors, is because they want to be independent instead of being at the mercy of foreign sanctions. If it weren't for that, they'd happily buy from foreign suppliers because that's cheaper than developing the technology themselves.

There is another documented history: that of the US sanctioning China, preventing China from accessing all sorts of western technology. For example, the US denied China of space technology. So now, China has wholly self-developed space technology. China isn't selling space technology below production costs to kill competitors. China invested in this sector in order to safeguard their access to space. https://www.businessinsider.com/nasa-china-collaboration-ill...

The US forced TSMC to not supply to many Chinese companies, and TSMC isn't even a US company. The US is trying to kill the Chinese semiconductor industry, making it wholly reliant on foreign suppliers. For years, Chinese companies have bought from foreign semiconductor suppliers because those are better. But the US has now turned this matter into an existential crisis, a matter of survival. Is it then any wonder that China, both public and private actors, are now pouring billions into the domestic semiconductor industry?


> If you look at the Chinese chip industry, then you see that a lot of money comes from the private sector. Huawei, Xiaomi, etc are investing a ton of money in semiconductor companies.

A private sector which got so successful not least because government systematically robbed IP from competitors (Nortel...). Yes, western politics also messed this up, but no these are not "private" sector operations in the same way Tesla is.


> government systematically robbed IP from competitors

You do realize that's exactly how the US became an industrial power right?

The difference being that most companies were happy to provide access to their technology in exchange of reaping higher profits from penny on the dollar Chinese labor and expedited logistics

But ofc, you never hear that, or the fact that we are talking of a different country with different laws, the US is not entitled to enforce its laws in other nations after all


> If you look at the Chinese chip industry, then you see that a lot of money comes from the private sector. Huawei, Xiaomi, etc are investing a ton of money in semiconductor companies.

Where do you think those "private companies" got the money to build their businesses? Notice also that it's domestic (i.e. controlled by the Chinese government) "private companies" doing this.

China isn't allowed to just nakedly subsidize everything under trade agreements, so they have to do money laundering like that in order to get away with it.

> Clearly it's not all government money.

It doesn't actually have to be. All it has to be is enough government money to put competitors out of business.

> There's also a big difference between investing in an industry with the goal of dominating the world and eliminating competition, vs the goal of becoming a sovereign equal. Why is it necessarily the former, and not the latter?

Because the former results in one country being effectively the sole source in the world for certain products, e.g. cell phones, whereas the latter results in a domestic industry but not the destruction of all foreign industry in that market.

> The US is trying to kill the Chinese semiconductor industry

The US is trying to prevent the world semiconductor industry from becoming monopolized by China the same way they've been doing to so many other industries, because semiconductors have significant national security implications.


> Where do you think those "private companies" got the money to build their businesses?

By... selling stuff to people?

I'm sure they get government subsidies, but that's not all of their money, or even a majority.

> so they have to do money laundering like that in order to get away with it

I'd like to see some proof that 1) they launder money and 2) that it's in significant enough amounts as to allow companies to adopt a kill-the-rest-of-the-world strategy.

The latter part of your post basically boils down to "action X is evil when they do it, but not when we do it, because we're the good guys and they the bad guys". Where's the proof? Has China actually killed competition via years of state-funded undercutting, or is it just speculation that they might?


The offshoring of the production of consumer electronics devices is mostly the result of the free-trade policies of said countries, coupled with market forces. The growth centred around manufacturing has been the basis of Chinese prosperity and requires no speculating about devious motives to account for.

As for the Chinese government's massive spending and subsidization of key industries, it's largely due to being free of the particular ideological constraints which force the US government to entrust everything to the market (except the military industry) which has predictably led to brittleness and vulnerability.

The American approach does resemble the Chinese one in a sense. Whereas the Chinese government prefers to directly subsidize e.g. semiconductors, the US government pours unfathomable quantities of capital into a military capable of dissuading any country from capitalizing on the US's dependency on any of their industries. When this approach breaks down, as in the case of China, things tend to get very dicey.


> The offshoring of the production of consumer electronics devices is mostly the result of the free-trade policies of said countries, coupled with market forces.

This fails to explain why everything moved to China and not Mexico or Brazil or India.

> As for the Chinese government's massive spending and subsidization of key industries, it's largely due to being free of the particular ideological constraints which force the US government to entrust everything to the market (except the military industry) which has predictably led to brittleness and vulnerability.

Entrusting everything to the market works great when it's a level playing field. When a country is doing things that would be an antitrust violation if done by a company, that requires another country to stop them. Typically through the use of trade sanctions, i.e. tariffs.

> The American approach does resemble the Chinese one in a sense. Whereas the Chinese government prefers to directly subsidize e.g. semiconductors, the US government pours unfathomable quantities of capital into a military capable of dissuading any country from capitalizing on the US's dependency on any of their industries.

In other words it's completely different?


I understand now. Unfettered market freedom is the best way to organize things and at the same time, large-scale government intervention is an unfair competitive advantage.

> This fails to explain why everything moved to China and not Mexico or Brazil or India.

Of course, no industries moved from the US to Mexico, Brazil or India over the same period as China's rise.

> In other words it's completely different?

Yes, in the sense that a portion of massive Chinese government spending goes directly into producing things that people need, and almost all of massive US government spending goes into funding imperialist violence.


Tariffs don't do much to hurt the other country. What they do is hurt yourself by making your consumers pay more for things.

If you want to make your country more competitive, you want Asian-style industrial policy and export discipline - as found not just in China but in Korea and Japan.


I'd like to see this documented history.

You mentioned electronics, but have they raised prices?


The trend in electronics prices is to decline over time, so you get the same result by not reducing prices as much as the ongoing reductions in cost.

Check how much a piece of consumer electronics costs in China vs. the US. They could charge the same amount.


> They could charge the same amount.

Go tell apple this, they would be very happy to be able to be able to charge even more for what they produce from China, they would be quite pissed to know that someone along the production chain is making too much money where it could be them instead

The only difference between let's say Xiaomi or US located electronics is that the US ones are very, very happy to pricegauge the hell out of their customers, that's why so many are happy to buy in alibaba rather than let's say Amazon or eBay


Are you saying that western electronics companies could theoretically produce at a price lower than what Chinese companies charge today, if only those western companies didn't get killed off? Proof?


They’ve done it in the security industry. Look at Hikvision for an example, who have recieved billions of USD in direct loans, and even more billions of USD in credit from the Chinese government.


Loans... credit... meaning that they have to pay it back some day. How is this different from any other company in any other country lending money to grow? The original claim was an unlimited pumping of state money, which never has to be paid back, with the purpose of killing the competition.

And now that HKVision has 'won', have they raised prices and are they extorting the market, as the original claim said?


That would be great if they were as it would make Chinese silicon cheaper for the US since they are trading partners.


Review the part where they only do this until the competitors are out of business. It's called predatory pricing and is an antitrust violation when companies do it.


I'm saying that the Chinese government has determined that semiconductors are strategically important and is willing to pour yuan into fabs and developing their own process equipment industry. Semiconductor manufacturing is a very capital intensive business and the short-term market view in the West isn't well suited to remaining competitive in that space. Silicon Valley has very little to do with silicon anymore because our markets decided to pour money into other areas where the payoff is quicker.


Thats great for them and great for US semiconductor companies since we are trading partners.


That would be fine if the US and China were great friends and allies - like if it was the UK instead of China that was the semiconductor powerhouse. But things look stormy on the horizon. Even discounting a military conflict with China (not likely anytime soon, I think) an economic conflict seems very possible. As China begins to manufacture a larger share of semiconductors they could easily decide to favor their own internal industries that are customers of said semiconductors and not allow state-of-the-art chips to be sold outside (or to certain countries). We (the US) have done similar things in the past.


> But things look stormy on the horizon

What do you even mean?

Be explicit

It certainly wasn't China the one that started the tarrifs debacle throwing the world economy into disarray


No. Dependence upon East Asian semi fabrication is a liability even if you're best buddies (as with Taiwan). TSMC was founded by Morris Chang -- hard to find a more reasonable, US-friendly, capitalist person in the industry -- but every time a butterfly flaps its wings in Asia-Pacific, we have to worry about the sanctity and continuity of our semiconductor supply chain.

NVDA just reported earnings and, while booming, it was not the blowout that you'd expect in the context of the greater demand picture for their products. That's in an environment where you don't have open multinational warfare in Asia-Pacific and their manufacturing partner TSMC is firing on almost all cylinders (water shortage excepted).

Fabs in the US and elsewhere in the Americas are more defensible in times of war, and more likely to prioritize demand from US companies when supply tightens.


Anyone can own stock in NVDA no matter where they live.


I genuinely don't understand the link between your comment and my own.


It is right.

US public companies are judged by the efficiency and timeframe of their returns on capital, which means that big long-term projects have to be carefully planned and justified.

That's not the case when the government gets involved. We've already seen China build a lot of other infrastructure for no reason. Highways, skyscrapers...

Additionally, Chinese companies operate in an environment where it's OK to steal foreign IP as long as the Chinese government approves of your business activity. So it's not just capital that the government uses to fertilize growth in its corporate nursery.


> That's not the case when the government gets involved. We've already seen China build a lot of other infrastructure for no reason. Highways, skyscrapers...

China urbanized 550 million people in the last 30 years.

They aren't building cities and highways to stand empty. They are building them in advance of demand. Sometimes the demand arrives a few months, or a few years after the infrastructure is in place. [1]

All those ghost cities you read about last decade? Most of them are now brimming with people.

In contrast, here in the US, we don't build housing until a decade after we need it (and when we do, the supply is anemic), so we have a crippling housing shortage in all major metro areas, which leads to sky-high rents. And through all that, we still have the occasional real estate bubble exploding, in a way that takes the entire world's economy down with it.

[1] https://en.wikipedia.org/wiki/Under-occupied_developments_in...


No corporation would finish construction of a factory ten years prior to the existence of demand. Companies like to lay out capex as close to income generation as possible. That comes from a shareholder-driven pressure for operational efficiency. It looks bad if Samsung (for example) spends $30B on a fab for which there is little demand. Fab utilization needs to be as close to 100% as possible; even at high utilization, the company doesn't break even on the fab for several years. Even if you assume 100% utilization, a fab may take a decade or more (from the time ground is broken on the facility) to break even on the investment.

In contrast, governments can leg out risk on big capital-intensive projects. Read the citations in the wikipedia article you referenced (avoid the blog post and the China-based publication) and you will see what I am talking about. A government's ability to build infrastructure without knowing whether there will be demand upon completion is how we end up with projects like HZMB that don't serve an existing commercial or private purpose.


The thing is, in China, government can create that demand, by moving government jobs to that city (thus bootstrapping the employment market as a primary employer), or having a state-owned enterprise open a branch office there (again, bootstrapping the employment market.)

Once there's a reason for people to live there, other services and businesses organically move in, to serve those people's market needs.

It's not just a matter of risk tolerance - it's also the ability to coordinate between employers and landlords, to optimize city planning.


That's a good argument for a different conversation. No sarcasm. It's a fair point but it takes a bit of a detour from the topic of semiconductors.

My point vis a vis semiconductors is that firms don't leap before they look. Governments can and do. You say they do it because they can guarantee demand. I say they can do it because they are more patient and risk tolerant. Either way, the logical conclusion is the same:

Firms can't lay out capex the way that governments can. Firms are held accountable for ROI, their shareholders have little patience for delayed gratification, and their ability to spend is very small relative to that of governments.


US sources of capital tend to have much higher time preference. Chinese sources of capital (i.e government entities) are willing to play the long game and accept below-market returns on capital for an extended period of time when it's building a strategically important industry.


A state capital system with a comparable gdp is very efficient at quickly allocating resources to sectors

China is that


Making chips in US makes little sense except to get the coveted tarif relief for the very few remaining highly tarriffed goods.

And if China decides to put sanctions on your fabs, you will just get a $10B microelectronics theme park on your hands because you will have nowhere to sell your chips.

Do not count the last point lightly. I'm fully serious saying that US will likely have nowhere to sell those chips to.

And that itself is an even bigger problem for US than it falling behind in semi competition by itself.


> And if China decides to put sanctions on your fabs, you will just get a $10B microelectronics theme park on your hands because you will have nowhere to sell your chips.

This is already happening. Except that the sanctions came from the US, not from China. The US forced TSMC not to sell to Huawei and other Chinese companies the US put on the entity list. Chinese companies that aren't yet sanctioned, are becoming reluctant to buy chips from foreign suppliers.


Do you think the US can't retaliate by throwing sanctions on consumer goods back? Board assembly is automated. It can move back here.


china biggest trade partner is europe and they are grooming Africa and other countries, so they with almost a modern economy will care little of sanctions. Maybe Us will get the kickback having to deal (aka subsidie)with unhappy lobbies


> they are grooming Africa and other countries

What's up with this language, being an economic partner is now "grooming"?

My country was literally coup'd by the US in order to maintain its access to my nation's natural resources through US/British companies, how is that for "grooming"?


> And if China decides to put sanctions on your fabs, you will just get a $10B microelectronics theme park on your hands because you will have nowhere to sell your chips.

China could also decide that they won't sell their more advanced chips to the US (or any chips for that matter if it got serious).


Ironically, this is exactly what's going on right now... Except it's the US denying advanced chips to China. The US forced TSMC and Samsung not to sell to Huawei for example.


It also denied a big chunk of chip market to the automotive industry, including its own.


Drought in Taiwan is a problem as well.

"Taiwan's chip industry under threat as drought turns critical

TAIPEI -- Taiwan's tech manufacturers fear their output is under threat from the island's worst drought in decades, risking more turmoil for global supply chains already strained by shortages of semiconductors and other key components.

Taiwan's government will on Thursday further tighten water use in several cities that are home to a cluster of important manufacturers. Plants in Taoyuan, Taichung, Hsinchu and Miaoli will have to cut consumption by up to 11%, on top of a 7% cut requested last month."

https://asia.nikkei.com/Business/Tech/Semiconductors/Taiwan-...


Given how many times in the past how electronic tech production has been impacted by natural disasters, from a simple few minute power loss down to earthquakes or tsnami's.

It is somewhat worrying that with the shortfall as it is, any climate impacting event as been in the past would be way more impacting noe and with the way the climate is, such things are more likely year upon year.

That with current trends and for good reason to expand capacity more globally, with the time involved to do such things, we may well see over capacity headlines 5 years down the line and with that thought. If history is to be learned from, the good times are ahead, until then - we might see tech longevity support and tech repair become more of a thing.


Good thing we aren’t doing anything to destabilize a climate system that’s not causing us too much trouble as of now.


You mean, like manufacturing electronics? It takes about 1MWh to manufacture a laptop, the same energy needed to drive a Tesla for 3000 miles.

https://www.networkworld.com/article/2229029/computer-factor...


> the same energy needed to drive a Tesla

who are the memeticists driving this analagy into our collective consciousness? why is this popularizing? ive been seeing this crop up more and more frequently.

it feels a bit like begging the question, except the arguer also holds TSLA


It's just a handy reference for energy usage, and I assume this figure is more readily available due to the car being battery powered. It's a bit harder to work out for an ICE-powered car, as you'd have to calculate the energy required to move the car based on aerodynamics and weight. I have no interest in TSLA stock, nor do I think Tesla is anything like the sustainable future of transportation. I mostly picked Telsa because it's Hacker News.

As you insist though, it's about the energy required to cycle 100,000 miles.


You can do exactly the same back-of-the-envelope calculation for ICE engines as you did for the Tesla, you just need to know the average fuel consumption per mile and the energy density of gasoline. No need to worry about areodynamics.


I was using that number to represent just the energy required to motivate the car, and assuming that an EV has near-perfect energy translation from battery to movement (and therefore providing an accurate enough representation of that energy). It gets a little complicated if you want to consider the fuel input, as you start getting into wells-to-wheels calculations. Depending on your power grid energy mix, it also makes things quite a bit worse for EVs.


Actually you already have something like this. MPGe is used across EV and ICE vehicles alike.

Fuel is far more energy dense for ICE engines are far far less efficient.

I don't get this comparison point as well, electronics is dwarfed compared to other far more energy intensive industries.


You can, but that wasn't the point I was making. My only point was that it takes around 300Wh/mile to move a car about the size and shape of a Tesla, regardless of how you power it. I won't use the example of Tesla next time, as it seems to confuse matters.

MPGe also touches on the source of the energy you are using to propel your car, and depending on your power grid energy mix, can produce quite widely different results for EVs. Once you also take into account transmission and conversion losses, it is not as good as you'd think.

Electronics is one of the most energy intensive industries in the world, especially if you do the calulcation per weight of final product.


You can avoid a lot of flying and driving by using zoom on that laptop.


Now take a look at all your other laptop-like gadgets (the infotainment in your car, your smartphone(s), TV(s), PC, gaming rig, etc etc), and tell me how much flying and driving they prevent.


Yeah but flying is more fun.

- Sent from the Seattle Airport


So probably 10MWh to build a Tesla, before even driving it?


Nearer 25MWh, and that's just for the battery pack.

https://www.sciencedirect.com/science/article/abs/pii/S00078...


Also, just for fun, the embodied energy of a 1960's Fiat 500 (assuming its entire 500kg weight is made of steel, which is probably a good enough approximation) is 3MWh, or three laptops.

So what is more sustainable?


An electric bike?


You're not wrong. However, every time I think about that, I just tell myself to pedal harder ;)


just for reference 1kg of aluminum raw material,not alloy, not processed, not packed, not transported - just mined is over 40KWh.

In that regard even the raw materials for the car are orders of magnitude over.


Indeed. The embodied energy of the items we use every day is so under-appreciated, probably because we have outsourced the production of almost all of them to countries that have no problem burning oil and coal in the process.

Really, if we actually wanted to make motoring sustainable, we would all be driving a 60's era Fiat 500 or Honda C50 moped. They cost very little to make, in every way, are very fuel effecient (even with their very old engine designs), and discourage long journeys.

That won't work though, because there's too much money to be made from EVs.


Is there anything wrong with that? If people want EVs that's better than people wanting modern gas guzzlers.


Yes. People are not being given a choice (at least, here in the UK). Petrol cars are being legislated out of existence. Ideally, I want a reliable, cost-effective family car without any extra gadgets, and I don't want an EV. However, those cars are just not made any more.

Also, I doubt the people that want a gas guzzler will be among the first to switch to an EV.


I have a hybrid (so it can be charged home, for 3000km we used only 50L of gasoline) - the EV drive is amazing, (if not spooky) - silent start, no noise, no vibrations (of course no exhaust), so it's great for city drive. For longer haul - there is the gasoline. For added benefit the total power of both engines engaged is quite impressive, along with acceleration.

Until there is a massive network that can reliably charge on the go, gasoline is there to stay. Also flip note: I know Norway loves EV (due to tons of tax breaks), yet EV do suck at lower temps (say below -10C)


I drive more than 3000 miles per year and I've got this laptop for 4 years now.


Did you account for how much energy to create the Tesla as well?


No, I was just using it as a easily-understood reference of comparative energy usage. Really, it's about the same value for any other car with the same aerodynamics and weight as the Tesla.

As you point out, it gets a lot more complicated when you start adding up how much energy it takes to make a Tesla (or any other EV).

Really, if we actually wanted to make motoring sustainable, we would all be driving 60's era Fiat 500s or Honda C50 mopeds. They cost very little to make, in every way, are very fuel effecient (even with their very old engine designs), and discourage long journeys.

That won't work though, because there's too much money to be made from EVs.


It’s all interconnected and inter causal in a gigantic mess.


What about energy expended manufacturing the Tesla?


I really regret using the example of a Tesla, as it conflates too many issues. I meant that it takes around 300Wh to move a car about the weight and shape of a Tesla, one mile.

It is a very interesting calculation to look at the energy cost of manufacturing any EV, and how long it takes to "pay it back" compared to an ICE-powered car, and under various sources of power for the electrical grid used to charge it. However, that was supposed to be beyond the scope of this conversation.


Since we're trying to use human-understandable units for measuring energy inputs, It's also the same amount of energy that is necessary to boil 600 pots of soup.


Not much use unless you specify the size of the pot ;)

By my calculation, 1MWh is enough to boil 600 17-liter pots, which isn't very human-relatable.


We're talking about the full energy budget of common household objects, like laptops, so you need to take into account the inefficiency of your gas stove, as well as the full energy cost of extracting and transporting the natural gas to your apartment.


Well then, let's see your working. Gas is really quite efficient in all ways.


That's almost as much soup as I eat in a year.


also lots of purified water


While the demand is cyclic and volatile in this industry (compared to software), the huge demand and influx of capital is very exciting for those who care about the advancement of computing.

Note that there are only major shortages of leading edge nodes, not basic electronics on older nodes that people aren't as excited about using. So if some critical safety system needs a basic sensor, there is plenty of 65nm capacity available. That's why Global Foundries isn't experiencing the same growth and bookings as the leading edge fabs are.


> Note that there are only major shortages of leading edge nodes

Is this true? Do automotive companies use leading edge nodes, because they're experiencing a shortfall too.


Probably "a" leading node (<= 22nm), but I'd also be interested to see specific orders and shipment numbers in 2021 vehicles. Products like the Intel Atom are still built in the same fabs as desktop/server silicon, and these are presumably going into automobiles.


And people scream at nvidia trying to segment the market a bit and/or price gouging. Meanwhile nowhere near all launch day orders (7 months ago) 3080 orders have been delivered, and at this rate I suspect there will be new gen out before all the outstanding 3080 orders have been fulfilled.


I believe (something looks like permanent) ETH crash is enough to fix GPU demand including miners selling their GPUs as used. By introducing PoS or gov's regulation?


An ETH crash will happen, but I doubt it will be permanent.


The last BTC crash sent a lot of used equipment to the market, even though it wasn't permanent.


The supply chain disruption last year made me think of the beer distribution game [0], and the Bullwhip [1] effect it exposes.

I am very far removed from the semi-conductor industry and the supply chain they feed. But given that lots of manufacturers slowed production because of COVID concerns and expectation of lower sales due to the expected recession only to find that demand was stable or even up as people adopted to remote working/learning, are we seeing now the peak of the Bullwhip effect?

In other words, as manufacturers now try to catch up, how possible is it that manufacturers panic and over-order only to find that people got what they wanted and are not going to buy as much as expected this year? If that is the case, increased demand is not sustained growth.

[0] https://en.wikipedia.org/wiki/Beer_distribution_game [1] https://en.wikipedia.org/wiki/Bullwhip_effect


I'm having a tough time buying STM32 MCUs. When I started my most recent designs about 6 months ago, the SMT-assembly service I used, and retailers like Digikey and Mouser had ample stock. Now, I'm unable to start production due to being unable to find MCUs. The whole STM32 line is affected - Only a handful of variants are available, which may or may not be the ones a given design can use.


I started a new project last week and I chose my STM32 by sorting by stock High -> Low. A year ago it wasn't even a consideration (I make low volume stuff).

I sent a big complicated board to manufacturing a month ago. I was making changes up to the last minute based on availability. One of the parts went out of stock and I provided an alternate part number. A week ago my assembler contacted me and said that went out of stock as well. I'm a bit pissed off that they didn't order it when they confirmed the change and there was plenty of stock, but I'll cut them some slack because they're in Houston. Now there's no stock of the part anywhere in the world except a single Aliexpress store. I can't overstate how strange it is to see such widespread availability problems. Pretty soon I'll have to design things with nothing but discrete transistors.


Surprising. ST runs their own fabs.

People must be stockpiling, and ST is replenishing the most profitable stuff first.


What assembly service do you use?


JLCPCB. I'm curious if there are others that have stock; may have to switch if JLC and the retailers are all out. I posted on the ST forums; the replies speculated it would be months for this to sort out.


30% above supply at what price? (shrieks in economist)


The free market has a bit of a problem here: A new state of the art fab can cost many $Billions and takes a good 5 years to go into production. Not a lot of companies have this kind of cash on hand or the ability to wait that long to achieve profitability. Also much of the current fab capability (TSMC, Samsung for example) is located in some geopolitically unstable places (Taiwan, South Korea). There's lots of risk to the global economy here.


I think the point is more that there's an obvious remedy to demand being above supply: raise prices until demand drops.

But for some reason, this doesn't happen as much as you'd expect.


Isnt supply price-inelastic in the short term, so "30% above" is the same quantity at all price levels?


Demand is not inelastic so the 30% drops if the supply becomes more expensive as orders will be cancelled.


A lot of this is people now over-ordering since they can get stuck by a shortfall. As soon as the crunch passes, there will be a big drop in demand while people burn down their inventory.

Normally, planning is 18-26 weeks out and nobody really worries.

Now, planning is 52+ weeks out, and people are stockpiling.


This is known as the Bullwhip effect: https://en.wikipedia.org/wiki/Bullwhip_effect


Could people have anticipated this and speculated to make a bunch of money (with the side effect of smoothing out the demand)? That is, bought a bunch of ... semiconductors of some interchangeable, reliably-demanded sort (I dunno, RAM? Generic Cortex-A77 cores?) a year ago (or whenever production was plentiful), stockpiled them, and sold them for profit today?


Absolutely.

Companies with stock are making bank--there aren't many of them and they're draining. Companies with their own fabs are making bank because they can prioritize their production runs by profitability.

The problem is that inventory has both risk and weird tax consequences so nobody wants to hold it. The risk is simply that nobody ever wants the part so you lose the money you spent. The weird tax consequences are how you "write down" inventory--the problem is that the IRS lumps a lot of inventory into the same boat: bolts are sort of treated the same as semiconductors even though they have very different life cycles.

The real lesson in all this is that there is no longer a surplus in worldwide fab capacity. The "fabless" semiconductor business plans relied on this surplus to beat up the "fab owning" companies for years.

The fabs all just figured out that "Fabs are the walled garden" and the fabless guys just got a reminder that "Your profits are dependent on some else's platform and they can take it away."

It will be interesting to see what the reckoning is after this all shakes out.


That's the big question:

Is this a crunch, or has persistent demand for semiconductors been created by recent changes in the way that people live and work?


That all sounds great, but with semiconductors you always have to worry about double ordering, stockpiling, etc.


> Some market observers believe that demand for semiconductors currently exceeds supply by up to 30% […]

Then raise the price until it’s too expensive for the extra 30%. Problem solved.


Rather problem transferred down the line.

Sure, we can do without those 30%, but wouldn't it be better to find a way to satisfy demand rather?



And Intel other Us chip companies are trying to get money from Government... there is something very wrong in the CEOs minds.


Good news for semiconductor companies. It's time to expand production!!


This isn't a real criticism, because I understand what they are trying to say but....

"Demand exceeds supply by X%", often a labour related headline, always makes me make the Marge Simpson noise. The framing is a direct contradiction of the main supply & demand theories/concepts. Supply and demand (within theory) are curves not amounts. It grates my circa 2005 economics education.

That said... either definition of supply and demand are arbitrary. It is actually true that in many scenarios, the definition of supply and demand used here (demand is a number) is much more insightful and predictive of what happens in real life. Useful lesson that many/most abstractions and models are useful when they're useful, but we tend to get overly attached to them.

There's a babushka doll element to all this. Divergent economic perspectives defined this industry, and the difference between US policies from (eg) South Korea.

The Korean industrial policy is built on the "demand/supply is a number" concept. US industrial policy built on the "demand/supply is a curve" concept. South Korea subsidized the industry on the assumption that it's a massive growth industry. US policy of this era (circa 2000) expected declining US production as marginal profits declined. They also saw the process as natural and positive. Leading economies hand over cutting edge industries to the next tier economies, focusing on the new cutting edge industries. Next tier countries (ATT, south korea) can focus semiconductors, and offload their second tier industries. Textile manufacturing is at the bottom of this stack.

Both US and Korean policy makers were correct. Apple is a great example. The product design, software and other high ROI (because low/no capital cost) elements remained in the US. Apple's profits are great. Their jobs (except retail) are all high paying. The parts of Apple with high capital investment requirements, lower margins, lower ROI or lower paid jobs were outsourced. This is pretty much the future 1990s economists in the US wanted.

South Koreans were also right. The idea was that certain industries, like semiconductors are massive growth industries. A good place to establish a competitive lead. True. The also treated capital investment as a good thing, even though it lowers ROI by definition.

I think the US mistake, if it was a mistake, was considering manufacturing a dead end... or a terminal destination. The truth is there will be more things to manufacture in the future, and a cutting edge manufacturing industry is the way to get a lead on that. Pathe dependencies can be important.

It's an interesting story to pay attention to.. maybe a turnaround in some decades-dominant modes of thinking about industrial policy.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: