TSMC warns over deepening slump in chipmaking sector | Hacker News Reader

TSMC warns over deepening slump in chipmaking sector

ft.com

109 pointsmfiguiere135 comments

TSMC just warned the world of a 10% decline in chip revenues. Given that chips today are critical component of every phone, TV, laptop, desktop PC, server, vehicle, dishwasher, and pretty much everything else, TSMC's warning suggests we're going to experience a significant global economic contraction soon.

Hopefully I'm wrong.

This isn't necessarily representative of the economics of downstream sectors, because it's probably caused by the Bullwhip Effect[1] like the recent semiconductor shortage.

"The bullwhip effect refers to a scenario in which small changes in demand at the retail end of the supply chain become amplified when moving up the supply chain from the retail end to the manufacturing end."

[1] https://www.investopedia.com/bullwhip-effect-definition-5499...

That would imply there is a surplus somewhere, which there's not. There's a reduction in orders from a reduction in demand regardless of the state of the buffer.

Their is a surplus of labor.

It's still the same effect. Supply doesn't have to overshoot. Computer chips aren't just stamped out instantly; a complex chip with many hundreds of process steps can take months to manufacture. The latency of design/revision cycle means manufacturers are always doing everything they can not to overproduce.

The chip shortage lasted longer than you'd expect because everyone from end users to intermediaries has been hoarding them.

It's like the covid toilet paper shortage, even if you keep producing a thing at the same rate, stockpiling creates shortages which stimulate even more stockpiling.

When that eventually reverses, expect a temporary glut.

> a complex chip with many hundreds of process steps can take months to manufacture.

"Three months, 700 steps" for the GlobalFoundries Fab in Malta, N.Y.

https://www.seattletimes.com/business/technology/three-month...

Very possible that this is a factor, it's very hard to sort out. I see execs and investors are nervous. Is it the bullwhip effect or am I seeing a real increase or decrease? You'll know looking back in a few years!

I certainly hope you're right. Keep in mind: TSMC is entrenched in a lot of markets characterized by short lead times (say, < 1 month), and a 10% drop is a lot more than anyone -- including the company -- was expecting. Sometime ago I recall seeing charts showing that changes in TSMC's annual revenues are highly correlated with measures of global economic activity such as changes in Apple's global revenues and changes in global GDP.

Could you elaborate on what you’re referring to? I’m certainly not an expert, but my understanding was that the vast, vast majority of TSMC’s revenue comes from production of completely custom wafers on newer nodes, and I’m hard pressed to think of something in the industry with _longer_ lead times.

I was referring to lead time on orders for existing products, i.e., those for which TSMC already has manufacturing capacity online, like current iPhone chips. I wasn't referring to the time it takes to go from design to manufacturing!

In addition to the bullwhip effect, and the unwinding of stockpiling, there's probably some real reduction of demand, and that's probably mostly from China's failure to bounce back to pre-covid GDP.

During the shortage lead time for many semiconductor products went through the roof, so it would make sense that there has been at least some stockpiling. Probably you're right that that doesn't affect top tier products which are the bulk of the revenue, because those are the ones that got preferential access to fab capacity. Whether the stockpiling affected a big enough fraction of revenue to cause this effect is not something I have access to the data to know, so like you I am merely hoping that this is the explanation.

Both explanations are probably correctly.

This is a bullwhip effect, but that does cause a real build up in inventories and contraction in demand during the down part of the cycle.

And this will be concurrent with the higher rate environment and contraction in homebuilding and other sectors caused by Fed policy.

And so the TSMC warning is likely is forecasting a brewing recession, precipitated in part due to bullwhip effects.

I don't think that's true. If my BOM costs go down by 10%, I don't need to charge any less money for my product.

If Apple could build an iPhone for 1 cent, why wouldn't they still charge $1000 for it? They can say it's how much iOS costs or whatever. (And hey, use the extra profit to hire more tax-paying iOS developers, making iOS even more valuable.)

I think the point is why has revenue dropped 10% for TSMC if everything downstream is going gangbusters? Doesn't it point to a drop in demand for chips, and therefore the retail devices they go inside?

If downstream things have gone to shit already, why the prediction that a crisis will be in the _future_?

TSMC's revenue grew too much during the pandemic. It's just normal it should contract when things go back to normal.

Potentially this is a tipping point for the things downstream. Honestly I’m just thinking out loud here.

> Doesn't it point to a drop in demand for chips

Yes, but in isolation it isn't evidence of an economic downturn in general. It is evidence of the whiplash of consumer demand shifting during COVID from services to goods and back again along with the cost of retooling the industrial plant to match that demand. There's a range of ripple effects across industries that will be ongoing for some time.

During COVID, with chip shortages, everything downstream bought as many chips as they could get. Now they have stockpiles to work through.

> If Apple could build an iPhone for 1 cent, why wouldn't they still charge $1000 for it?

Cause in healthy capitalism, a competitor could step in and sell it for $999 and Apple would lose market share quickly.

> Cause in healthy capitalism, a competitor could step in and sell it for $999 and Apple would lose market share quickly.

No they can't. You're neglecting the various moats that an entity like Apple enjoys with the iPhone.

First you have to be able to actually build a similar or superior product. You can't do it. Elite of elite tech engineering companies like Samsung can just barely keep up generation after generation. The investment required to build your $999 competing phone is comically enormous. You won't even be able to get access to enough chips to do it, much less have the manufacturing capacity to deliver at any meaningful scale given the components.

Do you have tens of billions of USD in cash ready?

You need to build a brand that consumers desire, such that they'll even want to spend $899 or $999 instead of picking up the iPhone at $1000.

Do you have another block of tens of billions of USD in cash ready, for marketing, brand building, advertising?

You need an extraordinary logistics system globally.

Do you have another block of tens of billions of USD in cash ready to build, deploy, refine your global logistics network?

Apple has ~$114 billion in annual operating income, most of it (including its ecosystem) is riding on the back of the iPhone. If it could be done, somebody would take that giant pot of gold from them. Even if a competitor could take 1/4 or 1/2 of that pot of gold, they'd move on it in a heartbeat. They can't do it, not even remotely close.

And then time. It'll take you a decade to get there, absolute best case scenario.

What Apple is reaping, is decades in the making, and required hundreds of billions of dollars in capital to be deployed over that span of time. Good luck.

> You're neglecting the various moats that an entity like Apple enjoys with the iPhone.

A world in which Apple could build iPhones for 1 cent would be a world in which those moats did not exist.

Why? How does the cost of microprocessors and metals decreasing make programming easier?

The cost of iPhones includes the cost of programming them.

But you're saying that moats wouldn't exist if the phones cost 1 cent? Why? I'm claiming that the cost of the iPhone in that world is from the included software, and the included software is their "moat". Cheap parts doesn't give you a "blue bubble" on iMessage. That's what the end user is paying for.

> you're saying that moats wouldn't exist if the phones cost 1 cent? Why?

Because if iPhones cost 1 cent to produce, the cost of everything that goes into producing them would have to be trivial. Which means any competitor could easily do the same as Apple is doing.

Of course that's not the case in our actual world; in our actual world, it would cost a huge amount of money for any competitor to do the same as Apple is doing. But that's my point: that very fact is why it costs Apple a lot more than 1 cent to produce an iPhone. Apple has expended huge costs on developing the product and keeping it ahead of competitors, and those costs amount to a lot more than 1 cent per iPHone. The high price they charge for iPhones is how they recoup those costs.

In other words, in the hypothetical I was responding to, Apple could not have done what they actually have done to make the iPhone a high end preferred product. Which means that all the moats that exist in our actual world because Apple did all those things to make the iPhone a high end preferred product, could not exist in that hypothetical world.

Does that mean the hypothetical is basically useless in understanding Apple's position in our actual world? Yes. That was part of my point.

That's an interesting take. I jumped to a Pixel for the 10x periscope camera, presumably offered because they were willing to take less profit on it than Apple, who couldn't find a way to charge more for it within the suite of features they were offering last fall.

The assumption is that the lowered revenue means fewer chips, rather than cheaper, less profitable chips. I don't know whether that is true or not. Might be a bit of both.

Typically, lower demand means lower prices - at least in the short term. I believe that TSMC is saying that demand is decreasing quickly and a supply glut will need to clear until it gets better. Just my two cents.

>> If Apple could build an iPhone for 1 cent ...

... then Samsung can also make a phone for 1 cent, and they might try to steal share by selling for $400, and some consumers would switch, etc.

In economics the closer you get your price to be MR=MC the better off your profit. If they did as you said more than likely they would be leaving money on the table. This is usually a fairly common intro economics problem as it can seem unintuitive to lower the price to make more profit. Now iPhones do have somewhat of a 'monopoly' effect. But only to a point. In a monopoly you want MC=MD. https://openstax.org/books/principles-economics-3e/pages/9-2...

Apple falls in the middle of those two axioms. If Apple could pull of an iPhone for 1 cent there would be a flood of entrants into the market doing the same thing as making the things would be cheap and easy. At first some would pay the premium (pay for the name) but after awhile that premium would erode and it would seem odd to pay when you can get the same thing for wildly less money.

Respectively, plenty of companies can build a smartphone for far, FAR less than the sale price of an iPhone. E.g. a $50-$130 Samsung Galaxy A03.

It's Apples install base, app ecosystem, and social moat (iMessage) allows continued high gross margins on their hardware and locks competition away from their profit center. The sale price has very little to do with the marginal cost of production and a ton to do with a naturally occuring software monopoly.

That Samsung uses the 12nm node, so you could say it falls somewhere between iPhone 7 (16nm) and iPhone (10nm). Those are worth about $100 used. A lot of the cost of manufacturing cost of cellphone parts comes from upfront capital expense, which is paid off by new models. This allows a strategy of selling cheaper, lower-end models using that old equipment.

Apple also pursues this strategy, but they don’t push it nearly as far as Samsung. The cheapest iPhone, the SE 3 at $430, has a 5 nm chip in it, which they only released two years ago. Apple just isn’t willing to release a parts-bin device, because they have a high quality bar. This is part of a moneymaking strategy, but my point is you can’t compare the A03 price to any current iPhone model.

Apple’s hardware margin is 35%, that is a reasonable estimate of the premium Apple is charging over a totally generic phone.

Very nicely said. They fall somewhere between MR=MC and MR=MD because of the type of product the are. I am surprised it is as low as 35%. But I will accept that. A company as big as Apple should know where its 'sweet spot' is for pricing. Too high and people walk, too low and you are leaving money on the table. It took me awhile to wrap my head around the concept. But the math does check out. Most intro econ classes teach it with 'the pizza parlor' and simple demand/supply curves. What the slope of those curves are depends on your competition and your market status (aka reputation). Apple has squarely tried to make themselves look like a premium product. All the way from its advertising to the box the thing comes in to the store you buy it in. That puts them more on the monopoly side as monopolies will do that to increase MR=MD.

The are not totally a monopoly. They may be that in some peoples influence. But mine for example my parents literally would not be able to tell the difference between a generic sub 50 dollar phone and a 1500 one. They do not use any of the extra features and it would be meaningless to them. For some people it is about the status of that logo on the back. For others it is the tech stack and integration. For others it is just something to text and phone with.

Apple has moats around the mobile space far greater than the moats in, say, cars or washing machines.

Many companies built stock over last years and stoped buying new semiconductors until stock is almost depleted. I am sure it will be fine, no contraction soon. Only semiconductor companies will suffer.

In addition to inventories like other commenters mention, remember this is a capital intensive business. When chip makers collectively add more capacity than is needed, prices drop (ie revenue drops).

It could also be "normalizing" demand as people shift spending from goods to services.

The possibility space is much larger than global recession.

I hope you're right. Keep in mind: TSMC is entrenched in a lot of markets, and a 10% drop is larger than anyone -- including the company -- was expecting.

Sometime ago I recall seeing that changes in TSMC's revenues are pretty highly correlated with measures of global economic activity such as changes in Apple's revenues and changes in global GDP.

> Given that [...] we're going to experience a significant global economic contraction soon.

What's with all the doomslinging in modern discourse? No, no, that's not now it works at all. There was a chip shortage just a year ago! Did that plunge us into recession? No, clearly it didn't. Then it was Inflation that was going to kill us all. And it didn't either. Now all of a sudden we're swinging the other way with chips and have a glut, and... now the doom will happen, I guess you're hoping?

No. This is a cyclic industry and it's cycling like it always has. A 10% YoY revenue drop isn't anything weird. I found a chart here: look at the bottom YoY numbers and see all the times they were in the red. Remember the unmitigated economic disaster of 2016? Me neither.

https://www.macrotrends.net/stocks/charts/TSM/taiwan-semicon...

At this point I just stopped listening to pretty much anyone regarding the economy. It has become plain as day to me that the whole field is kaput and that no one can really predict anything in this incredibly complex, multivariate chaotic system.

> What's with all the doomslinging in modern discourse?

Because public trust in the ability of governments to manage a crisis has gone completely fucking downhill, worldwide:

- the US couldn't even manage to beat a gang of cave-living Islamist terrorists using kalashnikovs in 20 years, and had to flee the country in shame

- we went through three years of a worldwide pandemic, then panicked governments acted like it was all over due to pressure by the far right - and yet, despite mounting evidence of millions of cases of long covid and almost all nation leaders silently insisting on covid tests prior to meetings, for the citizens it's a free for all for the virus, at the risk of disabled people and other at-risk groups. Governments didn't do anything to meaningfully curb the onslaught of propaganda as well.

- compared to prior environmental threats such as CFC coolants or lead, no one is doing anything meaningful (i.e. something that actually costs Big Money some wealth) against climate change. Instead, again, climate change denial runs rampant.

- we've got Russia invading Ukraine in a land-grab and it took us a year of brutal slaughter, rape and terrorism to send a handful of fucking tanks, and fighter jets are still many months away. I'm not advocating to glass the Kremlin, but NATO could do way way WAY more to stop this shit.

- China has been stealing Western IP for decades and used obvious price dumping to destroy our economies, and yet our politicians and companies keep kneeling before Xi despite the constant Chinese saber rattling against Taiwan

- housing prices and cost of living have been exploding for years and no one cares, despite tent cities of homeless in every major warm US city and people defecating on the streets because they don't have any safe place.

- on top of all of that (literal) shit, Nazis and other authoritarian politicians are back worldwide, to a large degree financed by the Kremlin, and guess what, politicians aren't taking care about that either, instead they let their parties be taken over by them or refuse to fight and go under in elections.

- ETA: And in the US, people will probably have the choice come next election between someone who should have retired from politics decades ago on the Democrat side and on the Republican side either someone who has incited a putsch, stole top-secret documents and hid them in a fucking toilet or someone who thinks it's a good idea to take on Disney of all companies or to fight everyone who doesn't think like him as "woke" and wants to eradicate them. JFC America what are you doing?

So yes, it's completely rational for people to assume doom when politicians don't give us any reason to hope that the situation will improve. It makes more sense to prepare for the worst case because it's increasingly more likely.

This is probably the most depressing comment that I've ever upvoted.

Nothing but the truth though. For me the question is not if we get a serious left-wing terrorist uprising out of all the shit happening at the moment, it is when. Personal guess it's going to be either UK or German's environmental activists or Israel, with a serious leaning to the latter given that everyone over there gets a ton of military mandatory experience.

> if we get a serious left-wing terrorist uprising out of all the shit happening at the moment, it is when.

What do you mean by left-wing terrorist uprising? Isn't left-wing ideology kind of relative to each country e.g. Icelandic left-wing/right-wing is not like Chinese Japanese left-wing/right-wing.

What would be the goal of these terrorist?

> What do you mean by left-wing terrorist uprising?

A lot of the issues I stated - e.g. the rise of the authoritarian/far right, climate change, wealth disparity - are usually associated with the (far) left, or at least the left is the one pushing for reforms.

Generally, people don't become terrorists... if they don't have the feeling they have no other choice left. Israel is the closest to that tipping point, with soldiers threatening to hand in their papers (which is very unusual as it puts the country at a serious risk should any of its enemies attack), but climate change protestors in the UK and Germany have recently escalated their methods (and so has public response, including a lorry driver running over a protestor blocking a street here in Germany), and god only knows what will happen in the US should Trump gain a second term or, worse, DeSantis.

That is why I'm relatively certain at least some people will choose violence - the question is how many will follow them.

Very astutely put!

> That is why I'm relatively certain at least some people will choose violence - the question is how many will follow them

Judging by history, long periods of wealth disparity, inequality, hunger, authoritarianism, etc, are inevitably met with brutal uprisings: e.g the French/Russian/Chinese Revolution.

As for how many will follow such a revolution: any kind of right(or right policies disguised as left as we have/had in Russia/China) politics, continued over a long period of time, will eventually have the effect of creating a small privileged ruler class out of touch with the rest of society. Thus, it's no wonder that usually EVERYBODY joins such upheavals to dethrone the rich/powerful.

> NATO could do way way WAY more

Where in NATO charter are they to interfere for a non-NATO country?

We could have delivered tanks from our storages the very second the first Russian crossed the border, we could have delivered fighter jets, we could have done a lot without directly involving us into the war.

Instead we got Biden being afraid of Republicans (most importantly the Russian asset that was the 45th President), our German Olaf Scholz being afraid of his own party and its close ties to Russia and the far-right and far-left and their even closer ties, and French Macron and Turkey's Erdogan both trying to act like they're kingmakers. Oh and we got Ukraine to agree not to use the pittances we send to hurt the Russians on their own soil.

All while hundreds of thousands of civilians and troops died in Ukraine. What a disgrace - our hesitation cost real human lives for nothing.

USA provided munitions before WW1 and Germans took out the Lusitania after warning Americans not to set sail on that boat.

https://archive.archaeology.org/0901/trenches/lusitania.html

Providing munitions to non-NATO country is meddling and could be considered an act of war. Where is USA Congress to vote on this war?

> Providing munitions to non-NATO country is meddling and could be considered an act of war. Where is USA Congress to vote on this war?

There is absolutely no relevant precedence for the supplying of weapons to a country at war to be considered an 'act of war'. It just isn't a thing.

The US Congress has repeatedly voted to approve those arms sales and transfers.

Under Article 4, which was invoked by Russia's invasion of Ukraine, the NAC can formally lead a joint response to any threat to any member (prior to any war). This led to a bunch of logistic support and prepositioning for closer NATO nations. It also led a joint decision for military material support for Ukraine, under the theory that it was cheaper support Ukraine today than fight Russia later.

Well, inflation is an actual danger (see Turkey) and Federal Reserve deserves credit for handling inflation well.

I agree semiconductor is an industry well known to be cyclic and prophecies of doom are ridiculous.

> Federal Reserve deserves credit for handling inflation well.

That is a stretch. They jacked rates on Trump, lowered rates for Biden, and in my opinion were forced to raise rates when the Ukraine money/power grab ramped up.

"Forced" or not, they still raised rates. They could have done otherwise, which would have been disastrous.

They followed war events in a reactionary manner. It would be cheaper and transparent to set rates algorithmically. The Fed can stay in business by continuing to inadequately regulate banks like SVB.

Interest rates were already increasing when Trump was elected. They jacked rates to cool the economy, as they should have done.

Rates were already decreasing when COVID hit, and they dropped rates to zero when the economy shut down, as they should have done.

Correlation does not necessarily imply causation, as I'm sure you know.

You could have simply looked up a long-term chart of the FED funds rate to verify your political belief, but you chose not to. Why?

https://www.newyorkfed.org/markets/reference-rates/effr

February 2020 discount rate was 2.25. March 2020 it was 0.25. Was that the biggest drop ever? Just two weeks to slow the spread, ha!

https://fred.stlouisfed.org/series/INTDSRUSM193N

I can't tell if you're being disingenuous or if you're just not very bright.

That was a full year before Trump left office though. Are you confused about dates?

Inventories are stacking up after the COVID supply chain crisis, so automotive and durable goods manufacturers are slowing down on input orders, even though demand for their products remains high. There's definitely a PC/smartphone demand slump (both sectors down around 10% YoY for Q2 2023), so add those two data points and factor in potential supply increases down the road due to CHIPS Act effects, and you're looking at decreased demand and increased price pressures for an uncertain amount of time. Given just how contradictory the broader market data is regarding the "soft landing vs minicession" debate, TSMC's specific revenue issues shouldn't be assumed to be dispositive.

New car inventories are increasing as well. Anecdotally, the US auto industry seems to be preparing for a slowdown.

https://www.coxautoinc.com/market-insights/new-vehicle-inven...

This is a 10% decline from the large increase over the baseline trend that we saw post-covid when people had tons of disposable income. It's just a return to baseline and no big deal.

Most consumer electronics don't need ICs built on cutting edge process nodes. Most of the global population of humans have more pressing needs than acquiring the latest gewgaws that can upstage last years models.

10% decline over the previous year, which was a boom year due to shortages after covid.

See this article back from distant Jan 23 for example

"Timeline: causes of the global semiconductor shortage" https://supplychaindigital.com/top10/timeline-causes-of-the-...

We have been seeing a significant global contraction for the last year lol. It’s still going.

We just all riding on the post covid backlog and over order.

>we're going to experience a significant global economic contraction soon

1. YoY 10% expected decline from the largest increase in recent history. i.e 2022 was an anomaly.

2. No it is not BULL WHIP effect. Has relatively little to do with it. The problem was with commodity chips, one should read into UMC and SMIC.

3. Nvidia, for the rest of the year is still capacity limited.

4. Samsung Foundry took some market to fill their capacity gap in other sectors. Although I doubt that is even 3% of the expected TSMC revenue decline.

5. In all likely hood it just means their Customer are reluctant to pre-book much capacity and would want the flexibility to increase order when needed. ( Usual Strategy )

6. It is likely China's volume is much worst than everyone expected.

This boom/bust is really obvious with regards to computer prices.

I'm seeing 2TB NVMe PCIe 4.0 SSDs with 7000+ MBps under $100 these days. Almost everything in the computer-world is on a significant sale.

You made me check and I notice the same in my area, great deals indeed, thanks!

Makes me wonder why I bother with anything in the cloud at all!

We have to counter with even more resource hungry and inefficient software of course

I'm doing my part!

I have my Windows installer raring to go

Real talk: ever since I put Windows and Linux on separate SSDs, rather than simply separate partitions, they haven't killed each others' bootloaders once. Best $100 I ever spent.

My second tip for solving incompetent software with redundant hardware: TOSLink -> 1Mii Bluetooth Transmitter to avoid the catastrophic linux bluetooth stack. It's brilliant. TOSLink = no buzzing from inadequate shielding next to digital lines (or "AI audio" noise floor massacre or popping or...) and no software bluetooth = OS can't screw up pairing and that OS switching can't screw up the pairing. I did this in a moment of rage after the Nth time dumping hours into yet another linux bluetooth issue only to find that someone had submitted a patch years ago but project management hadn't reviewed the patch for a year and then banned the person who submitted the patch for complaining that nobody was reviewing his patches >.<

Any specifics about the second paragraph anecdote? It seems like a serious scandal.

Hmm doesn't look like I wrote it in my journal but yeah, it was pretty gross. The reason given for the ban was some swearing in the message, but to my intuition the profanity was pretty tame, well-deserved, and ultimately just an excuse.

I do remember the bug: Bose headphones have a low quality mode and a high quality mode (low quality frees bandwidth for microphone and maybe latency), and the linux bluetooth stack defaulted to the low bandwidth mode so aggressively that even if you changed the script to set high bandwidth mode it would immediately set back.

I have no idea if the bug is fixed because this happened in 2020 and I haven't thought about desktop bluetooth since 2020 because TOSLink->1Mii Bluetooth Transmitter solved all my problems with it.

The sorry state of Bluetooth on Linux is well known at that point. Android had to roll their own stack because the standard BlueZ on Linux is a fucking pain.

I have windows on a separate SSD, and grub still stops working whenever I update my motherboard firmware...

Condolences. I wish I had a second trick for you but I don't.

At least I can fix it from Windows, because it's just some uefi parameter that gets borked. I have the command in a text file on my desktop.

I don't really know how to trash the SSD in my software.

Flash wear says hello.

Run something like Chia.

(Obviously not an endorsement.)

I don’t really know what software you write, but trashing flash from software is a thing.

https://www.tomshardware.com/news/flash-memory-wear-killing-...

Cache videos on disk aggressively

It sounds like we need to port more software to Electron

except for H100s from NVIDIA. Waiting for them to go down in price

They are in a separate bubble right now.

That's entirely consistently with Moore's Law in NAND chips. I honestly think they were gouging us by not dropping prices earlier.

I'm holding out for $100-200 8TB versions before I fill out the empty slot in my laptop and max out the RAM. Having all my in-use video, samples, and Kontakt libraries on-board would make life a lot easier.

Recently found a 4TB Samsung for ~200 USD

8TB is new enough that it's still at premium pricing, but I expect it to fall like 4TB did. I'd hate to get a 4TB today when what I really needed was 8, only to have it drop within a few months. They're already down quite a bit.

Have you used any of these sub $100 devices? I just took a look, and all of the names are vendors I have no experience with what so ever. All of the vendors I am familiar are above that price point. So, my immediate concern would be these cheap ones are made from chips rejected by the name brands, and their performance would be rejected by me after wasting the time/money on them.

I was looking at NewEgg and not Alibaba/ebay/amazon

Lots of <100 2TB ssds on Samsung / Intel / Micron / etc. products.

https://slickdeals.net/newsearch.php?q=2tb&searcharea=deals&...

> Intel

Note: Intel is literally obsolete in terms of Flash. Intel sold all of their SSDs to SK Hynix, who is now selling those SSD-designs under the name of "Solidigm"

So if you want a "modern Intel" SSD, look for Solidigm. Intel got out of this race in December 2021, so no "new" SSD (2022 or 2023 era) is from "Intel" branding, only "Solidigm".

--------

I'm more of a Crucial / Micron fan myself, but Samsung, Crucial, and WD (aka: Sandisk was bought out by WD) are all 2TB for under $100.

But yes, I'm talking about the big-names, with high-performance (7000MBps+ suitable for PS5). The good stuff. GPUs, CPUs, HDDs, and Motherboards have also come down in price. Its entirely across the board though the SSD market is where its most obvious.

I have an ADATA 2TB in a PS5 and my work machine. These aren't fly by nights.

They can be had for sub $100. Even the much lauded Samsung 980 Pro is dipping close enough to the same.

Assembling a new home PC, picked up a 2tb Samsung 980 Pro with a heatsink recently for $129 (I think it was $113 without the heatsink at that moment).

DDR5 ram is finally starting to get price reasonable as well.

I got a 2TB 980 pro with heat sink from Bestbuy last week for $99. I was astounded. No brainer upgrade for the PS5

be sure to upgrade the firmware, 980 pros are known to die after a while.

dunno why people keep paying this premium for them, there's other better options for less money, like HP FX900 or Team Cardea these days. And those companies have a track record of reliability and not... whatever Samsung is doing for the last 10 years.

Like honestly these days Samsung rates below stuff like Sabrent Rocket for me. You're better off picking a random other product from a big-name brand with equivalent specs as long as it's not a Samsung... the "SSD monty hall problem" if you will. Simply by switching, you improve your odds.

I picked up a 2TB 980Pro and installed it into an external USB-C chassis from SSK. I now have extremely fast external storage to edit from on the go without filling my internal system drive with crap media footage. At ~$100, that enough to have one per project and avoids the "which drive" question. It's now, get the ProjectX drive, or get the ProjectY drive. It's the happiest I've been as a mobile editor.

A 2TB Western Digital SN850x, a near top of the line SSD, is $99 at Best Buy.

https://www.bestbuy.com/site/wd-black-sn850x-2tb-internal-ss...

I noticed that too. What does it mean? That demand is sinking or that costs are going down?

Computer parts are always high-margin. They're made out of purified sand after all, the raw materials of chips is excessively cheap.

The ultimate price that consumers pay is very much related to demand and less about costs. The economics of chips is very interesting: its very expensive to design and develop a chip. But making additional copies of chips (or mass producing any design, really) is extremely cheap.

-------------

So yes, computer parts will go up in price mostly in reaction to demand. And they'll go down in price mostly in reaction to demand.

The long-term investment into computers is not guaranteed however. It takes $Billions to design a modern chip, which is the hard part in practice. To run a computer company means to sell enough chips to fund the next generation. But for any given week or day of the year, the price of those chips can be whatever anyone feels like... really.

> Computer parts are always high-margin. They're made out of purified sand after all, the raw materials of chips is excessively cheap.

The purification and especially the lithography (aka making actual chips out of a silicon disk) is what is expensive. The machines are pricey as fuck - a single line from ASML alone costs 200 million dollars, and that's just the machines for the lithography, not including the machinery that makes the ultra pure monocrystals, that slices them, or that cuts and packages them, or the cleanrooms and the remaining support infrastructure. Or the extremely specialized staff you need to actually operate them.

And it's a ton of money involved with very long lead times - say you're TSMC, you get an order for a batch of chips, and even if you had capacity starting tomorrow, the 700 steps to the final product take three months [1]. That's a massive amount of capital to be stuck for months, and errors can waste all of that.

[1] https://www.seattletimes.com/business/technology/three-month...

They are ( NAND maker ) selling at a lost.

Got a link/model number? I’m always down for a new drive.

There's lots and lots of 2TB drives under $100 available. Just picking one that I know of:

https://www.westerndigital.com/products/internal-drives/wd-b...

Meanwhile, Apple charges hundreds of dollars at purchase time for 256GB more storage. And it's non-upgradable on top of that.

Sometimes I feel they are just testing our limits. Same with iPhone price increases. I guess they are going to continue until the net profits decrease.

Well, if they continue on that path the EU will mandate upgradeable storage next.

I get why they do it though, my 2022 MB Air is serious competition against my work 2019 16" MBP. Far superior battery life, similar display quality (WTF that notch though), feels just as snappy with IntelliJ, 200+ Chrome tabs and Docker, and all of that despite just 16 GB RAM. No way to do that unless you got fast storage directly adjacent to the CPU so users don't notice swapping.

The SSD isn't all that fast with the Air, though.

https://archive.li/43GrJ

A year+ ago, the same folks were talking about the incoming switch from under supply to over supply, and the predictable outcome that we see today. None of those involved are surprised by this on the macro level, but when they put their short-term glasses on for the purpose of quarter-to-quarter reporting, they say stuff like "deeper slump than expected". Obviously at the end of the day, these folks are sleeping just fine, knowing and having expected this with >12m of warning.

Good hopefully the prices will slide back towards pre pandemic levels making it more affordable for people

Houses too …

i'd be happy with a car

I wouldn't mind a Nvidia RTX 3090.

I'd prefer that Mazda RTX 3090

..and a 4090 for the 1000 EUR it should cost, not the 1900 EUR..

houses need to reverse decades

The last few years in particular have seen a dramatic rise in housing prices[1]. Reversing even a few years would dramatically increase affordability for many people.

[1]: https://fred.stlouisfed.org/series/MSPUS

Yea, but it would probably put many people underwater on loans.

You don't heal 50 years of bad urban planning and bad housing policy in a short time.

The last three years make the run up to 2008 look like the salt flats. There’s a long term issue but an immediate crisis.

This countries middle class will turn into single issue voters if housing prices ever go down. Politicians are too scared of them to let it happen.

If anything, houses are going to stagnate until inflation catch up. No politician will want to sit in office while overseeing a nation-wide collapse of property value, this would tear the fabric of society too much. People can stomach making a stagnating investment, but they can't stomach making an investment that loses value, especially not when the investment is touted as a no-brainer and basic life goal investment by everyone (the American Dream).

Reagan did, and he was somewhat popular.

Seriously, preventing future politicians from doing this is part of why the central banks are independent in G7 countries.

How can this be happening with an AI revolution needing an almost infinite supply of high end graphics cards?

Near as I can tell, because of the nature of the chip business, there is a limit on how fast they can pivot production towards a market segment that is seeing increasing demand (in this case, the AI chip business)

So even assuming that there theoretically are enough purchasers of AI chips available to offset their other lost demand, it will take a year or two before they can fully serve these new customers and collect this revenue.

Yeah I was thinking something along the lines of, we have factories setup to produce AMD CPUs on 4nm for 6 months, another setup for Apple etc. and if say Apple reduces its demand TSMC can’t magically turn the AMD or Apple (etc.) setups into Nvidia GPUs that quickly. Also NVidia probably benefits from some scarcity so it might not be worth their while to order more?

They said AI chips are just 6% of the production output. They said the AI demand is not significant enough to offset the cyclical downturn in semis demand.

Do they report the % of revenue from AI / data center chips vs all others? Or is there a way to derive it? Surprising to hear the slump given what we're seeing in AI-driven GPU demand. I guess that's just not as meaningful a portion of the TSMC business as I would've thought.

To what degree does this also reflect the restrictions on China receiving higher-end chips?

It's probably the main cause.

Chinese companies use to make up 30% of TSMC's revenue.

Reply on news.ycombinator.com