Hacker News Reader

Comment by madaxe_again | Hacker News Reader

You wouldn’t believe how fragile some of our supply chains are, and how many single points of failure we have.

For a while a decade and a bit ago, there was only one operational, legal tantalum smelter on the planet - the others had been destroyed by the Indian Ocean tsunami or shut down over their use of conflict minerals and slavery.

I think there are now three - all on the ring of fire.

Then you’ve got stuff like cobalt, which we get at the whim of the DRC, a notoriously stable and functional state.

And of course, gallium, graphite, and rare earths, which pretty much all come from China. They could, at will, bring the global electronics supply chain to its knees.

I guess risk is everywhere, really, in our fragile systems.

View on HN

I vaguely remember a story about an accidental single point of failure. If I recall correctly, during the 1970s or 1980 there was a critical part of the wafer production process (slicing the crystal into wafers, perhaps?) which relied on highly specialized tooling.

As it turns out, a bunch of manufacturers had independently discovered that by far the best source of them was a small company in Japan - unbeknownst to them essentially just one guy and his workshop. So everyone sourced it from him. Which was fine, until he took a holiday for the first time in a decade... Suddenly both original-source and second-source factories for some important chips were unavailable!

I'd love to know if anyone has more details, because I haven't had any luck with finding any reference to it. It's a story I read probably a decade ago, so who knows how many details I got wrong.

Sounds like DISCO to me, although details differ.

One advantage of these types of things is how easy they are to stockpile. Aside from the nominal storage costs, none of these things degrade over time:

https://www.dla.mil/Strategic-Materials/

Stockpiles and other forms of resiliency (such as redundant facilities) cost a lot of money because by definition you cannot run them at nominal capacity or it's just a ton of expensive stuff that sits around.

Capitalism is fundamentally incapable of accounting for resiliency as it optimizes to cut such "dead costs".

Toyota, who is widely recognized as the pioneer of JIT manufacturing, famously began stockpiling microchips after the Fukushima disaster disrupted their supply chains. It might not make sense to stockpile raw materials, but stockpiling things with comparatively low storage costs w.r.t. the item cost sometimes makes sense from a capitalist perspective. This assumes the company runs with a long-term perspective in mind, which often isn't the case in western companies that are publicly traded or owned/operated by investment firms. In fairness, governments also frequently fail at accounting for resiliency, even when that's their explicit goal, so perhaps the incapability to account for resiliency goes beyond the economic system.

> In fairness, governments also frequently fail at accounting for resiliency, even when that's their explicit goal, so perhaps the incapability to account for resiliency goes beyond the economic system.

That's because governments are run by the same braindead efficiency-first mindset these days.

It used to be easy to campaign on resiliency back when the USSR still existed - no matter what, all NATO countries had really large armies, months worth of stockpiles of everything from arms and munitions over medical supplies to food. Education was heavily invested in because the countries needed well educated adults to compete with the external threat. Companies kept stockpiles and had detailed disaster recovery plans and drills, not just for natural disasters but also for "how the fuck do we keep production running when the Red Army is marching towards the border" and "how the fuck do we transition from building cars and motorcycles to tanks and fighter planes when the Great War breaks out".

But once the USSR fell and the "end of history" was announced [1], all of that resiliency went down the drain - it was expensive after all and neoliberalism / "starve the beast" ideologues demanded that the "inefficiency" be cut to lower the tax burden on the corporations and the rich... and here we are, our education systems in shambles, companies got screwed left and right when covid hit and supply chains broke, and our societies in shambles as well as the people didn't have much reserves left after decades of wage stagnation and companies didn't either so they laid people off.

[1] https://en.wikipedia.org/wiki/The_End_of_History_and_the_Las...

What a lot of people (and companies) missed is that Toyota doesn't optimize till there's no slack left, and instead properly run Kanban let's you establish a capacity and slack "reserve" in the system.

Some people seem to run companies towards absolutely no slack at all, and are baffled by the idea of keeping some.

> the incapability to account for resiliency goes beyond the economic system

Reminds me of how bad pollution was in the GDR (communist Germany). You think capitalism is bad for the environment? Communism was worse.

You'll find similar levels of ecologic devastation in the GDR as you will find in the Silicon Valley - the greatest agglomeration of Superfund sites in the US, originating from all the silicon production before a lot of it moved to Japan, Taiwan and South Korea.

Also, for fucks sake, look at the Western German Ruhrpott and the Ewigkeitslasten there. Or the open pit brown-coal mines - both West and East Germany have these, and they're a menace no matter where. All resource extraction is incredibly devastating, no matter if under communism, capitalism or anything in-between.

In the 80s, the GDR had unfiltered or less well filtered coal power stations, more widespread use of toxic chemicals such as PCBs, less safe nuclear power stations, and the standard car type had a two-stroke engine. Did you see the blackened house facades in the East in the early 1990s? It really was worse than West Germany. Doesn't mean that everything was fine in West Germany.

Open-pit mining holes and other permanent problems from mining like the need to pump out water to prevent flooding aren't really preventable except by not mining. It's mostly HOW things were done in the East that was worse. They had fewer spare resources to do things "nicely" and the state controlled both industry and the media (even the population via Stasi), so problems could be swept under the rug more easily.

I don’t think that’s a fundamental inability; we just choose not to due to fashion in business thought.

But there’s nothing fundamentally that prevents pricing in risk of shortages to shareholders and appropriately stockpiling to mitigate that risk.

> But there’s nothing fundamentally that prevents pricing in risk of shortages to shareholders and appropriately stockpiling to mitigate that risk.

There is. Shareholders are just as likely to fall into the short-term gains trap.

IDK, Berkshire Hathaway sure has a big stockpile of cash, and they're emblematic of capitalism. The cash in this model would be considered a "dead cost", as it's "wasted" by not being immediately invested in equities.

I agree with you, in general, that the short-term incentives of our system encourage "optimization" away from stockpiling inputs, and that this is a worrying problem.

But I find it instructive to focus on and give credit to the long-term thinkers who buck those trends – and who have historically profited and persisted by doing so, by the way.

Half of the world's neon was coming from Ukraine, a quarter each from Odessa and Mariupol.

This is also critical for semiconductor manufacturing.

https://www.theregister.com/2022/03/11/ukraine_neon_supplies...

Everyone seems to forget that Silicon Valley and its datacenters are in an earthquake zone.

A typical multinational social media operator has datacenters around the globe. The question is rather whether these can really function independently should the central ones go offline.

I would be more worried about Northern Virginia. I doubt the world at large would notice much problem if SFBA dropped off the network, but they sure would if the same happened to NoVA.

AFAIK till quite recently, if not still, us-east-1 was SPOF for various components of AWS. Your already deployed things would continue to work, but APIs (and most importantly, Console) wouldn't, IIRC.

Reply on news.ycombinator.com