A big hurdle to fixing the chip shortage: substrates (opens in new tab)

Bear in mind that actual productive output, or how much a factory cost is quite possibly not public knowledge. But area is (just look at google maps). If you're an insider with some of these ballpark area ratios in your head, you can easily come to some conclusions.

I've likewise heard people talk about building particle accelerators in terms of cost-per-meter. In particular: ballparking the cost of a synchrotron light source as a million USD per meter.

The hardware designer said: “It is rumored that you are a great programmer. How many lines of code do you write per year?”

Master Foo replied with a question: “How many square inches of silicon do you lay out per year?”

“Why...we hardware designers never measure our work in that way,” the man said.

“And why not?” Master Foo inquired.

“If we did so,” the hardware designer replied, “we would be tempted to design chips so large that they cannot be fabricated - and, if they were fabricated, their overwhelming complexity would make it be impossible to generate proper test vectors for them.”

Master Foo smiled, and bowed to the hardware designer.

In that moment, the hardware designer achieved enlightenment."

- Master Foo and the Hardware Designer: http://catb.org/esr/writings/unix-koans/index.html

The work unit is completed wafers that later get binned into different device quality levels.

Makes sense when your biz is to blast plasma over mm2 of silicon

In a sense, buying the future substrate production is just a future substrate contract.

Do you know how building a power plant is financed today? Basically someone buys 10 years of gas futures, sells 10 years of equivalent electricity futures, and uses the profit (since electricity is more expensive than gas) to build the power plant that will turn the bought gas into the sold electricity.

It also takes over three years to build the plant itself. We aren't even talking about the people you'd even need to employ, design, and run the place (highly specialized labor).

This is such a huge burdensome cost. It's not even a guarantee that that such an investment would even make sense once 5-years pass.

I do think there should be more fabs and having them located in one area of the world is already playing out to be a geopolitical nightmare. It's obviously going to become a national security issue for nearly every industrialized country on earth over the next decade if not sooner.

On top of that, can the Banana grower even get all of the items needed to grow the same sized crop as last year? All of the fertilizers, etc.? Those companies have been shut down during this time too. Oh, you're not the only company who needs those exact same supplies and the Potato farmer down the road is trying to purchase it too? After you've grimaced that the price has risen due to short supply and you had to source the items from 3x as many suppliers as you used to, now, you're relieved that your Bananas are almost done growing. You call the trucking company to deliver them to the distributor, who you learn is short staffed and prices have risen in order to attract truckers who are now in short supply, and factor in the gas price increases. Every single segment of the economy that the Banana grower relied on is going through the same thing, at the same time.

Making chips is several orders of magnitude harder/more complex than growing Bananas. Think about how many different companies products must be purchased to make a wafer. How many different highly specialized chemicals are used in addition to solid materials? If any single one of them is having supply or labor issues...

One car example, you should ensure you have 4 times as many tires available as you do cars, because an imbalance there becomes either a constraint (too few tires) or a waste (too many tires). But it's not JUST making those numbers match, you have to forecast, how many cars and I going to build, are my supplies for the cars ok, look at normal fluctuations, and then also look at potential supply chain FAILURES and plan for them. Maybe it's easy to get brake pads but the lead time on tires is 4 months. Well then, I don't need to stock as many brake pads in case of a disruption, but I should definitely have enough tires to carry me through a potential disruption (maybe not 4 months worth, maybe I stock 2 and slow down car production, or something).

All these companies kept their shelves bare, and when some companies slowed orders, they didn't stop to think that the manufacturers they bought from would look for other customers to take up capacity. They acted like TSMC would sit there by the phone waiting for some customer to come back. No, they're in the business of selling chips, not in the business of servicing DumbCorp. No one planned for what might happen if they stopped ordering parts, they just expected the capacity to be there.

I'm bringing 6 new sites online this year, and a 7th early in the new year. I just bought a bunch of network equipment for not just our existing sites, but the 7 to come. I looked at supply chain issues, and said to my CFO, this inventory is going to sit here for 6 months, but we get a larger quantity discount, and we'll get the new sites online faster and not worry about getting this equipment if we buy now. We COULD have saved that money now, but the advantages of having it when we need it is greater than having that money in the bank for 6 more months.

JIT means planning ahead, not just micromanaging inventory. Too many MBAs don't get that.

One immediate cause was that some manufacturers reduced their chip orders month in advance since they expected less demand but then the demand didn't decrease and they had to scramble, increasing general demand. Another immediate cause was that Covid has made some production lines shutdown or become unreliable.

But the reason that now a year or more on, we have problems, seems require more explaining. Larger causes I've heard; The chip supply chain is extremely long, long enough so you get a whipsaw effect or anti-whipsaw effect. A whipsaw effect means by the time demand catches up with supply, it goes beyond it and manufacturers lose money. But what seems to be happening is an anti-whipsaw effect - suppliers are avoiding any panic measures to catch up with demand and so lose money but this means the catching up is a long process.

This is something of a product of a highly capital intensive economy pushing manufacturers to only produce, invest and consume in a narrow range. Limiting investment in capital equipment is a way to make the investment you make is 100% utilized. Limiting consumption to the lowest priced items is a way to make sure all of your production is profitable. It seems like a new phenomena. Instead of a "deflationary spiral" these factors pushed to their extreme could result in a "decreasing production spiral" where physical continually declines despite demand. We're not there yet, of course but it seems like an interesting economic phenomena.

Then COVID triggered a collapse of nearly all JIT supply chains (not just medical). Try getting ANY industrial or consumer product - the supply chains are all broken right now. Some of this is China intentionally fxcking us. Some of it is over dependence on outsourcing. Most is due to inherent instability of JIT as a fundamental flaw that's long been ignored.

Many supply chains have both short and long feedback loops to themselves. For example ALL the equipment required to increase semiconductor capacity always require the very same semiconductor supply chain capacity (chip making requires chips being made).

https://twitter.com/_mg_/status/1388321639524229120?s=21

The guy who runs my local bike shop is saying they still won't have enough bikes next year.

Demand for gaming PCs and consoles have nothing to do with inefficient software. Demand for automobiles has nothing to do with inefficient software. Demand for servers, networking equipment, and office PCs to support a suddenly-working-from-home labor force has nothing to do with inefficient software. Windows 11 not supporting 5-year-old hardware, crypto booms, and pathetically short support windows for smartphones have nothing to do with inefficient hardware.

My 6-year-old laptop and 5-year-old desktop handle Electron apps just fine. It would be nice if the software was more efficient, but practically nobody is rushing out to replace a computer that was any good to begin with because Discord and Skype take up an extra 50 MB.

Besides, making software more efficient is a long-term solution to a short-term problem.

Please follow the site guidelines, which include:

"Eschew flamebait. Avoid unrelated controversies and generic tangents."

https://news.ycombinator.com/newsguidelines.html

We detached this subthread from https://news.ycombinator.com/item?id=28448249 and marked it off topic.

You have, and are probably using right now, a voice controlled computer that can convert anything to anything else, by just asking.

Is it really worth typing this every time an American site uses American language? We all get it, Americans with their imperial units, ha ha.

Americans, myself included, generally acknowledge that the metric system is better but we are already familiar with US units. There's nothing really wrong with US units either in terms of accurate measurement. In an ideal world we would use metric, but transitioning would come at monumental costs - never mind the re-training of an entire population - just so we can make unit conversions a tiny bit easier.

I get it. You're not familiar with US units like we are so you have to do conversions to make sense of the measurements - probably with a calculator since the ratios aren't simple, whole numbers. But remember Americans have to do the same whenever we read almost any foreign publication. Difference is I never see HN or Reddit comments complaining about metric units.

"The metre was originally defined in 1793 as one ten-millionth of the distance from the equator to the North Pole along a great circle"

That factor of ten is because we have ten fingers. There's no cosmic significance to ten or ten million. Or to the size of a line of longitude on our particular planet.

And a meter is almost exactly a yard, or three feet anyway.