Memory Makers Plan to Cut NAND Flash Production | Hacker News Reader

Memory Makers Plan to Cut NAND Flash Production

anandtech.com

64 pointsdeafcalculus36 comments

> three major manufacturers of NAND memory — Intel, Micron, and SK Hynix — have announced that they will be taking measures to address the oversupply

Isn't that illegal? To fix the market by colluding and price fixing?

How is this different from the act that Mercedes, BMW and VW are getting fined for for colluding to not release emissions tech?

it's illegal if and only if it can be proven that they coordinated their production cut.

They all respond to the same changing market conditions -- they can't just continue increasing their production when the price continues its downward spiral. Further, the largest producer Samsung hasn't joined the foray yet according to the article. Samsung is a big variable in this -- the company's expansion created the memory glut that drove the whole industry to record loss in 2015-2016.

The EU's antitrust authority claims that the automakers "coordinated" strategies to limit competition. This is probably not the case here.

It is possible that Samsung will not do so. It's the same problem that would have affected them had they actually colluded: each player has incentive to wait until others have cut production/raised price and maintain their own. Handy way to snap up market share. Every one loves to scream "price fixing!", but it's not easy to do and almost always collapses.

>This is probably not the case here.

Why do we think that? Hasn't these companies and/or industry been caught and fined for illegally coordinating their efforts before?

Well, one of the reasons is that that's what we've expected all along.

When Samsung announced a major expansion in early 2010's, many analysts predicted the price glut in 2015-16 that followed. That memory price glut (and a record loss by all industry participants) in turn discouraged capex by memory makers that resulted in the 2017-18 boom we just witnessed. So now, the windfall during the last boom likewise encouraged higher capex and production increase that is now driving down the price and their bottom line again -- the industry nearly completes a full cycle and is at another trough marked by decrease in capex and production cut. There is really nothing too conspiratorial about this business cycle.

Sure, I think every memory makers at one point got caught price-fixing, but you can't just say they are colluding again just because they are all cutting production under unfavorable market condition (ie, lower demand/falling price).

Another important contributing factor is how well the manufacturers do at getting each generation of NAND to actually work. Samsung's 48L 3D NAND was a relative failure around the same time that their competitors were trying to get their first 3D NANDs working to overcome the limitations of planar NAND. That kicked off a serious shortage, so when almost everybody came out with successful, high-yield 64L 3D NAND around the same time, prices had a long way to fall. The current moves are attempts to ensure that the transition to 96L NAND doesn't continue the price drops.

Price fixing in the NAND market has been rife for years now. All the same companies that have previously been busted for price fixing DRAM, LCD panels, then DRAM again, are involved in NAND. There is no rational reason why SSD storage should be as expensive as mechanical storage, let alone come at a premium. No rare materials are involved. No exotic manufacturing processes are needed. It's just a stupid regular grid of NAND cells. They're used in parallel, so it's hard to make them slow, not fast. By every logical measure, we should be able to get 10TB SSDs for $100. Instead, we see mechanical hard drive manufacturers stuffing high powered intricately machined motors into helium-charged cases with platters coated with rubidium and other rare exotic materials now talking about adding lasers into the mix and somehow THAT is the cheap option for capacity.

It's really pretty absurd how far they're pushing their price-fixing, but I suppose it's been made clear to them that this time they won't be busted for it. South Koreas president closed the investigation into it shortly before being jailed for corruption, and I haven't heard anything about re-opening it.

The depreciation on the equipment used to make NAND would not come close to supporting the prices you describe. In a market with a small number of players, public disclosures of capacity adjustments are sufficient to "collude". But 10 TB for $100 would require a something like a 10x reduction in pricing. None of the memory manufacturers have margins close to that good.

10TB for $100 is barely enough to cover the raw silicon wafers.

We could fix part of the problem (and plenty of other problems besides!) by working on best-of-class open-source FTL (Flash-Translation-Layer) and GC (Garbage Collection) for raw NAND or other solid-state storage. This would essentially enable us to open up the "storage" market to essentially any sort of solid-state memory, irrespective of physical implementation; such that cornering that market by intentionally limiting production of NAND Flash memory alone would likely become infeasible.

If you read it, it's mostly about phasing out costly older technology and slowing down the creation of a new fab.

The chinese are slowly entering the NAND business so there's not going to be few enough players to really make collusion possible.

http://www.businesskorea.co.kr/news/articleView.html?idxno=3...

Their defense would probably be that once they cut their legacy nodes there is a lot of incentive here for the individual players to defect from their announced participation in the market consensus.

The modern way of doing price fixing is through corporate M&A and that's been rubber stamped by every U.S. administration for >40 years.

> Isn't that illegal?

Yeah, how dare these companies adjust their production to respond to supply/demand imbalances? /s

I cringe at the vision of a world where that would be illegal.

I think you misunderstood my concern. I'm not talking about Intel adjusting its production to meet its demand. I'm talking about a collection of competetitors getting together and saying that we should limit supply so that the prices stay where they can make a good profit. How could that be healthy competition?

If all cement manufacturers suddenly got together and said they'll keep their prices at 40% more than others what happens to the construction industry and the consumers looking to build houses?

The "getting together" is entirely your own supposition. The NAND manufacturers have not made any joint announcements or otherwise admitted to any joint action. They have responded similarly to market conditions that affect all of them equally, which is to be expected whether or not they are coordinating. These announcements were completely predictable; we knew last summer that something like this would have to happen this year unless several companies experienced serious yield issues with their 96-layer 3D NAND.

>The "getting together" is entirely your own supposition.

A not unreasonable one considering two of the three (Hynix and Micron) are convicted price fixers.

But they won’t be leaving the market, right? Therefore, they expect to make some profit at the expected higher price, and would stand to gain the most if all others cut production, except for them.

This is a classic prisoners’ dilemma, where everyone should defect. If they don’t, it must be due to coordination, either ex- or implicitly. And because anti-trust is civil law, it criminal, the standard of poof is “preponderance of evidence”, i. e. >50%.

If none or only one manufacturer cut production/restricted expansion, then somebody would be leaving the market. That's how the DRAM market ended up collapsing down to just three players: When margins are too low, the smaller players can no longer afford the investment required to keep pace with technology improvements. A short-term desperate play for market share in conditions where you can't turn a serious profit just means you only get to play the game for one more round unless other business units will cover your loss.

This is not an efficient market. Large changes in supply have lead times on the order of years, and demand cannot be accurately predicted that far out. Months-long oversupply and shortage conditions are unavoidable. The costs of entry get significantly higher with each generation. A two generation old NAND fab cannot be operated profitably even after all the construction and R&D costs have been recouped.

Questions: 1. Can someone explain why this is not a clear cut of a collusion? Didn't the same "cartel" get busted for this behavior a while ago? 2. Why aren't they under tight regulation since they have a history of doing this?

because there is no proof of collusion. Absolutely no need for any further regulation.

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

They've pled guilty in the original DRAM price fixing collusion and they just got sued last year for the same charge, it's ongoing.

I can only guess that laws are more lenient in Asia and they can only be charged in US after the fact.

Edit: removed my opinion about getting sued again as that's off the topic.

Wow. Looks like all five manufacturers pled guilty in the original US case. Given that the suit from last year is ongoing, this seems to be a real problem in this industry.

> They've pled guilty in the original DRAM price fixing collusion and they just got sued last year for the same charge, it's ongoing.

Now you are making stuff up. We are in 2018, not 2006. There is no regulatory investigation by any gov't antitrust agency. China's NDRC announced that they had "massive evidence" of price fixing by Micron, Samsung and SK Hynix, in 2018. It turned out they had nothing, so instead they tried to blackmail Samsung and SK Hynix to drop all pending and future patent lawsuits against China's domestic companies in exchange for dropping a multibillion dollar fine and antitrust violation that they couldn't prove.

You clearly haven't read the class-action lawsuit by Hagens Berman in which the law firm alleges that the memory companies "coordinated" by signaling each other via PR. Do you understand why the class lawsuit is not going anywhere?

Please provide sources to the NDRC case, that's why I asked in the first place about why this isn't collusion. My understanding is limited to the original 2006 case where these companies were in fact charged and fined by Department of Justice, that's not a made up story and I provided the source to that.

Telling me "no proof of collusion" without any source is the same as making stuff up.

Technically, we're in 2019 but that's off the topic.

Google "china ndrc dram price fixing -digitimes"

If you are citing Wiki, you probably know very little about the older 2006 case. You seem to make haste assumption on what, how, and who are involed this time around without any facts or understanding claims made by different parties.

> Telling me "no proof of collusion" without any source is the same as making stuff up.

You are the one making accusation against the DRAM makers of price-fixing, it's your obligation to prove your allegations. The burden of proof is not on me to prove the negative (which is a logical fallacy).

I am highly suspicious of China's motivation for this. It happened at roughly the same time as news broke that the chinese stole Micron's technology. China is not exactly known for good or fair practices, and it would not be the first time they screwed over a foreign competitor and stole their tech to make a competitor.

Does this mean Intel's recently released high capacity SSD line (1 TB at 130€) will become more expensive?

It's hard to say exactly what the effect will be on Intel's consumer product line, because it exists primarily to use up leftovers that were discarded from their enterprise/datacenter SSD products. If their yields for QLC NAND get too good, supplies of their consumer QLC drives may largely dry up when most of their QLC makes the cut for enterprise drives.

Those drives are QLC (Quad-Layer Cell) which allows higher density at the cost of long term reliability. Higher density means more bits per wafer, which leads to lower cost per bit.

This also means speeds of 80MB/s once you fill up the SLC cache which is 2GB or so.

I would have preferred to see this happening to the dram cartel, who actually did fix prices.

Micron and SK Hynix were/are apart of the DRAM cartel.

Well they were loosing a lot of money over the falling NAND prices

All at once? Could be an accident but could be collusion

Buy Samsung NAND, it's good and they keep making it.

They were the first to cut back supply. I'm guessing you work for them?

https://www.techspot.com/news/79303-micron-follows-samsung-r...

https://www.bloomberg.com/news/articles/2019-01-30/samsung-p...

Reply on news.ycombinator.com