[scardycat]

Just in time manufacturing contributes a lot to this shortage. We rely so much on integrated supply chain logistics, the pandemic clearly brought in to focus the "optimizations" of JIT manufacturing. The lack of local supply pools as buffer is so short sighted.

[makomk]

Not using just in time manufacturing might help with short-term disruptions, but we're approaching the two year point on the pandemic and the associated global disruptions and it's just not realistic to keep enough stock in for that long. The current generation of stuff like CPUs and GPUs hasn't even been released and in production for that long...

[kennywinker]

My understanding is that the 1st order disruption is long-over. The factories and ports are all back online, etc. two years out all that’s left is the ripple effects of the original ~6 month disturbance. It’s just that those ripples are amplified by the way the supply chain operates (just in time).

[brewdad]

So we are at the hoarding toilet paper stage of the crisis? Everyone is ordering as much as they can, so effectively no one actually gets their order filled.

[makomk]

The first-order disruption definitely isn't long over - important manufacturing locations like Vietnam and Malaysia were in lockdown until as recently a week or two ago, and I don't think things are back to operating normally there yet Parts of China might still actually be under lockdown right now.

[kennywinker]

Is lockdown really effecting manufacturing in those places? Because when we had "lockdown" where I live that meant no fun, not no work - but of course everywhere has done it differently

[ren_engineer]

bullwhip effect, this stuff has been known for decades. Our entire global economy depends on ideal conditions to function properly

https://en.wikipedia.org/wiki/Bullwhip_effect

[ericbarrett]

I agree, but if a manufacturer has the means (physical and budgetary) to store a significant supply of components, they can order larger quantities with longer lead times, so they would still be more resilient in extended shortage situations.

[genericone]

Extended shortage situations account for fires/earthquakes/tsunamis/singularities, short events with long-term but predictable downstream effects. Extended shortage planning doesn't account for 2 year-long political-struggles, which requires something more like war-time planning, spending, and the inevitable waste from stockpiling.

[vkou]

This is great for that particular supplier, but an order for a larger quantity will just squeeze the inventory of every other supplier even more.

The problem is that we are trying to squeeze blood out of a rock. Demand exceeds supply, and it takes years to scale supply up. It doesn't matter if we are using JIT or not, the suppliers for the bottlenecks have all been working at 100% since essentially the start of the pandemic.

[dwild]

Well to be fair, it's not 2 years of stock that would be needed, as there was still some production that was happening during theses 2 years.

[mrtksn]

I keep seeing that argument every single time and keep wondering, isn’t there reason to implement JIT? Something like cost benefits outweigh the resilience risks?

It’s not like people start dying on the streets, simply things are getting expensive and hard to get during disturbances.

Are there calculations demonstrating that JIT the risks of optimization are more harmful than helpful in the long run?

[Spooky23]

JIT is a boogeyman in some cases. Ultimately it doesn’t make sense to stockpile parts most of the time for financial and other reasons. If Intel had a warehouse of 1Gb chips, they would have a disincentive to invest in 10Gb, for example.

The dysfunction is the accounting games that companies are incentivized to keep stuff off of their books. Every company wants to look like a software company and avoid stuff like inventory. Some of this is absurd - many companies don’t “own” property, for example, they lease through entities that are sometimes only nominally separate.

Sometimes companies will outsource processes and fulfillment to layers of other entities, each of which do the same thing. Disruptions cascade - I had one issue last year where a supplier’s supplier had issues getting boxes, and a two week delay there delayed downstream fulfillment by 6 weeks. All because the people who ship the end product couldn’t deliver, and the “principal” outsourced the actual management of the process to a third party. If the principal controlled it, they would have gotten it done, as they were punished severely by the contract penalties. They bet on everything working out ok and lost.

[laumars]

Ironically the biggest two IT companies in the world, Apple and Amazon, are ones that go out of their way to control the entire supply chain.

[Naga]

The problem JIT is solving is that raw materials inventory ties up capital (it costs money), and having warehouses of materials is money sitting that could be used for different purposes (capital expenditures, salaries, dividends, etc). Then, if for some reason the raw materials aren't worthwhile anymore (spoilage, tech change, etc), you've wasted money on something that needs to be disposed of. In theory, the goal would be to have no inventory at all.

JIT is a firm level decision, not a society level. It's up to each firm to decide how much of something they are buying.

[gumby]

JIT assumes a steady state, so squeezes all the buffering out of the system. But buffers are useful.

> It’s not like people start dying on the streets

Well, that’s a good example: there was a huge rise in need for masks and other protective gear about 20 months ago and the supply chain couldn’t handle it. That was why people were encouraged to use makeshift cloth masks, to leave the surgical masks and respirators for medical personnel who had the greatest exposure.

Eliminating buffering cuts cost and can lead to lower prices as well. But at best it merely pushes the buffer elsewhere.

Another way of looking at it: leaving seatbelts out of cars would save money and really, most cars are not involved in accidents so are they really needed? Pass the savings on to the customer!

[mrtksn]

I know the benefits of having a buffer but do we need buffers everywhere? As in your example, wouldn't be much better to have a buffer in the hospitals(or maybe some kind of regional emergency organisations) as a precaution for an outbreak instead of advocating for buffers across the textile industry?

I like the way you put it, JIT simply pushes the buffer elsewhere. However, this seems like a very good thing to have because those who cannot afford running out of something can do a buffer themselves instead of blindly everyone keeps buffering.

[tablespoon]

> I know the benefits of having a buffer but do we need buffers everywhere? As in your example, wouldn't be much better to have a buffer in the hospitals(or maybe some kind of regional emergency organisations) as a precaution for an outbreak instead of advocating for buffers across the textile industry?

The problem with that specific kind of stockpile buffer is that it can become quickly depleted. No mask stockpile would have been sufficient for the COVID pandemic.

From my layman's perspective, you need stockpiles and excess production capacity to weather a supply shock. It's sort of like backup power in a data center: you have UPS batteries (stockpiled power) to fill the gap until the generators (extra production capacity) can come online.

[mhb]

You also need the ability to raise prices when there are shortages in order to encourage buffering. Unfortunately, the people who like anti-price gouging laws appear to prefer shortages and misallocation of goods.

[tablespoon]

> You also need the ability to raise prices when there are shortages in order to encourage buffering.

No, but I can see how someone would some to that conclusion by thinking narrowly in terms of pop free market dogma.

> Unfortunately, the people who like anti-price gouging laws appear to prefer shortages and misallocation of goods.

Price gouging is actually a worse misallocation of goods. It's still a shortage, but it just doesn't hit rich people as hard. If you have food stockpiles to barely feed everyone through winter, it's not a proper allocation to let the market price food so richer people can feast and some poor people starve to death.

Price gouging introduces a lot of (especially short term) inefficiencies as greedy parasites make profit-seeking decisions based on their greed and not social need.

Markets work very well in some contexts, but it's a mistake to think they work best in all contexts. Crisis shortages are not one of the contexts where they work well.

[bsder]

> Well, that’s a good example: there was a huge rise in need for masks and other protective gear about 20 months ago and the supply chain couldn’t handle it.

I'm going to bump you just a little on this because your point is mostly right.

In the US, there were manufacturers ready to add extra shifts for producing masks, but nobody would cut them the check.

The point of JIT is also so that when something goes wrong in inventory you don't get saddled with a bunch of waste. The problem is that means that you need quick, accurate decision making for when the inventory situation does go wrong.

In this instance, all the players who could cut a check for masks were all paralyzed by their systems for various reasons (bidding/disclosure requirements, political ideology, etc.).

[aidenn0]

I don't think anybody actually did the math on this though. It's more like "I get a fat bonus if I decrease inventory without a significant effect to sales over the next 3 quarters" so as long as the near-term risk is worth it, it gets done.

[edit]

There's also the "If I don't implement JIT, I'll get replaced with someone who will" effect in a lot of cases too.

[Guest42]

That’s the impression I get, not so much JIT but rather limiting supply budgets as much as possible. A lot of companies tout Toyota as the pinnacle of manufacturing without realizing that they are not Toyota and there was more to Toyota’s success than JIT.

[unilynx]

Well only a decade ago Moore's law ensured that keeping any amount of stock on CPUs/GPU/memory would cost you a lot of money, as that would depreciate fast. JIT was a good idea until it wasn't

[whatshisface]

The Moore's law argument does not apply to low performance embedded chips which are whatever the engineers want them to be, but must be exactly what the engineers wanted them to be. The chips themselves were often the same between different models of the product, but even though demand for them was a very predictable function of the number of products made, and stable over time (pursuant to the stability of demand for the product), everyone's inventory consisted of whatever was in the box that was being carried from the loading dock to the pick and place machine.

[q-big]

> Well only a decade ago Moore's law ensured that keeping any amount of stock on CPUs/GPU/memory would cost you a lot of money, as that would depreciate fast.

This clearly holded for PC components, but rather not for, say, microcontrollers for conservatively developed products with a much longer service life and/or duration of sell.

Consider, for example, a stock of microcontrollers for an industrial machinery that

* will be used for 20 years by the customers,

* will be sold for the next 10 years,

* after these 10 years, for the remaining lifetime of the machine, the customer will still be able to buy spare parts.

[tablespoon]

> I keep seeing that argument every single time and keep wondering, isn’t there reason to implement JIT? Something like cost benefits outweigh the resilience risks?

Just think about evolution: species specialize for short(er) term benefits, but then conditions change and a lot of those specialists go extinct.

JIT is specializing for an extremely reliable and undisrupted supply chain.

[lisper]

It's a value judgement. You can optimize for the normal case at the cost of disruptions when the unusual happens, or you can optimize for the unusual case at the cost of inefficiency when things are normal. But you can't do both.

[cannabis_sam]

If computer network technology had been built on the same shortsighted business logic we would never have had the internet..

Imagine a link-layer protocol without any buffers, no retransmissions or any kind of resilience, built on the most optimistic estimates; that’s essentially what the “business world” has built with their irrational philosophy based around “the market”..

[culpable_pickle]

https://en.wikipedia.org/wiki/User_Datagram_Protocol would like a word

[dv_dt]

While there are users for UDP, I would guess that TCP accounts for more traffic by choice. Of course there are sometimes problems associated with too much buffer too.

[sudosysgen]

UDP is almost always used with buffers, error correction codes, and other measures to deal with imperfection

[bryarsdj]

Yes, but I think UDP is still a good counter example for this analogy. Buffers can be detrimental to udp traffic for real-time use cases (eg VOIP). Here your trade off is quality (in voice this is jitter) versus latency. Typically you want to make this choice of buffer size at the receiving end only, and keep buffers elsewhere as small as possible - routers which store and forward, network card buffers,OS buffers etc, can add up to lots of bloat.

So for VOIP "JIT" is a good thing and your "Inventory" levels need to be tuned at the receiver.

[cannabis_sam]

Ouch, thanks for the pushback!

But isn’t it called tcp/ip for a reason?

[MandieD]

Do you use a VPN to connect your corporate laptop to your employer’s network from home? Do you use Wireguard, or its easier-to-use derivative, Tailescale, to connect any of your devices? Do you use any form of VOIP (Voice over IP)?

(Edited because I forgot the big one) Have you ever used a name and not an IP address to connect to something on the internet?

Congratulations! You’re a UDP user!

[cannabis_sam]

Yes! I absolutely use UDP!

I once built a prototype network stack on top of a udp library that simulated a physical network layer..

Doesn’t change the fact that tcp/ip was the network stack that enabled the internet.

But I was trying to sidestep that tangent by mentioning link layer protocols in my second paragraph.

[dv_dt]

It’s not clear to me that any of the proposed causes are singly driving the shortages. The closest analogy in my mind is that our global production and logistics system is suffering from multiple shifting bottlenecks with secondary induced inefficiencies. Those new inefficiencies include JIT buffer exhaustion, related slowdowns, capacity reduction, priority inversions and resource contention.

[slg]

I don't know much about how supply chains function so I am wondering how is just in time manufacturing still to blame when the chip shortage has been ongoing for over a year? Wouldn't everyone have burned through their inventory at this point?

[captainmuon]

The idea is not that they should have had a stock for multiple years of demand. Rather, if you have a warehouse and are able to build a bit of stock, you can ramp up much faster.

The actual closings due to lockdown were only a couple of weeks in most industries. What really hurt imho was that orders were cancelled, production capability was scaled back, but demand came back faster than expected.

[slg]

It makes sense why it would ease the initial shock from shutdowns, but how would it help with the cancelled orders and scaled back production? It seems like it would just stretch out the shock over a longer period of time. Would that actually stop the disruptions because it seems like they would eventually burn through all the inventory eventually because of that?

[dannyw]

It takes two years to ramp up more fab capacity.