Discussion in 'Article Discussion' started by bit-tech, 24 Apr 2018.
Why would any company want to increase production and lower profits? SK Hynix, Samsung, Micron,Toshiba - all perfectly happy with the current state of the market.
Just came here to say much the same thing.
...so they can make more money?
Simple example: Company has three Widget Factories, each producing 100 Widgets a day. They're selling all 300; market demand is high. Profit, accordingly, is high; let's say 60%. At that, it's making 60p profit on each Widget so it's making £180 profit each day.
Company builds an additional Widget Factory, also producing 100 Widgets a day. It's still selling all 400 of its newly-increased output, because market demand is still high. Profit, however, has dropped to 50% (either because the Widgets now cost more to make, the increased supply has dropped average selling price, or both). Thing is, it's now making £200 profit each day because it's selling more Widgets.
So, why wouldn't Company increase production, providing the drop in profit is covered by the increase in volume?
Because in this instance, it's not just company A producing the widgets - Add company B, C and D (and those new Chinese fabs supposedly catering for the low end markets), apply rapid ramp up in competition, everybody loses.
I'm not going to explicitly claim that the big boys in the industry are actively colluding to strangle supply (though they probably are), but they are at least in a position of happily waiting to see who moves first...
Overly simplified economics. The new widget factory has a construction cost, taxes, and employee cost. Also there is no guarantee the market demand for Widgets does not vanish before the new factory is even done with construction making their whole investment a waste. Then there are the variables of competition Wakka mentioned too. If any Widget making company was truly confident that demand would remain high indefinitely and had sufficient financial leverage they would be announcing new factories.
All of that said, I miss the days of cheap RAM.
SK Hynix has three fabs (Icheon and Cheongju in South Korea and Wuxi in China); adding a fourth, even assuming it immediately hits the ground running, is only a 33% capacity uptick (and the chances of it equalling the output of its current facilities on day one are between slim and none). Samsung has... 20, with two additional plants under construction. Assuming everybody's plants are equal (which they aren't) and that Samsung and SK Hynix are the only companies in the market (which they aren't), SK Hynix's new plant equates to a whopping 4.35% capacity uptick.
Don't think anybody's going to be sweating that.
The first two words out of my mouth
fingers were: "Simple example."
Which is why the new widgets cost more to make: "Profit, however, has dropped to 50% (either because the Widgets now cost more to make, the increased supply has dropped average selling price, or both)." C'mon, if you're going to pick an example to pieces, at least read it first.
You mean... Like the new factory SK Hynix has already announced and will soon open? Like Samsung's Fab 16.2 expansion? Its Cheonan facility? Intel-Micron's recently-expanded Utah facility? GlobalFoundries' Chengdu plant? Intel's Fab 42?
All the big players in the semi game are opening up new fabs, for the very reason I explained: when there's excess demand and constrained supply, you're leaving money on the table.
Because they first have to be reasonably certain that demand will last for years to come as it takes years to stamp a new fab out of the ground and if demand was to suddenly drop then the shiny new fab would be a multi billion write off.
Aye, but unless we stop using computers I'm not seeing demand for memory falling off any time soon - and neither are any of the companies named 'ere who are literally doing the very thing people in this 'ere thread seem to think nobody's doing...
(And there's always mothballs: remember that President Trump's "woo, I convinced Intel to build a new US fab" announcement was actually talking about 2012's Fab 42 which had been mothballed in 2014 following a dip in demand then pushed back into service when demand picked up.)
I just want to know where i can get one of these widgets, if they're only make 3-400 of them a day they must be worth having.
Neither do I, plus these manufacturers should logically be among the first to know when exactly DDR5 will lower demand for the old stuff...
That doesn't really matter fab-wise: a fab currently making DDR4 can make DDR5 on the same node without difficulty. It can even make DDR4 and DDR5 side-by-side, adjusting the mix as demand for the latter waxes and former wanes.
The reason for not building loads of factories is they are high capital cost and subsequently high ongoing expenditure to maintain. If the market goes South they will be liabilities not assets. You always need to think what the market will be looking like in 5 years, not 1 day....
It's not the first time there has been collusion and price fixing in the industry.
But that's an inefficient way to make money.
RAM is proving to be quite price elastic, it goes up and up yet demand stays relatively unchanged. For the manufacturers it's ideal, the profit margin goes up for little or no risk on their behalf. The small increases in production capacity they are making indicate to me that they want to increase supply but only to the point that it allows them to sell more at the already huge margins they are making now, rather than increase profit by volume of sales.
High volume, low margin drive the profits downs for the entire industry and a fall in demand leaves eveyone sat on lot's of stock, lot's of expensive factories and comparatively small cash reserves. For now at least the Koreans appear to have decided that's a mug's game. Of course Intel and GlobalFoundries might muck up the plan, I hope they muck up the plan.
But... But... But... They are opening new and expanded fabs.
It's always amusing how consumer price rises (RAM, NAND, HDDs, etc) are inevitably put down to "it must be price fixing!" rather than "people with deeper pockets than you want to buy those parts more than you do". RAM is in demand for mobile devices, which takes capacity (and thus production volume) away from the more niche DDR4 we want. HDDs are in huge demand for nearline datacentre storage (all that data 'in the cloud' actually needs to be stored somewhere). NAND is just plain in demand from everyone everywhere.
Large production volume only reduces margins if demand is already saturated, and on top of that reductions in sale price also increases demand (because markets that would not accept higher prices are now open). We're a long way from an oversupply situation, and the o ly reason we're in this mess is because demand is rising faster than fabs can physically be constructed and fitted out.
The 'We'll keep volumes low to keep our margins high, and have a big slice of a small pie!' tactic only works until one player decides to have a slightly smaller slice of a larger pie, at which point they now have a volume advantage over everyone else. Which is why everyone is racing to build fab as fast as they can rather than standing still.
I mean, sometimes it is, but I'm not sure how many different ways I can reply to variations on the theme of "they'll never open more fabs 'cos that's not how you make more money" beyond "they're all literally doing the very thing you're saying they'll never do, like, right now."
Samsung has 20 fabs. 20. Soon to be 22. They sure as heck didn't get to 20 fabs by building one then going "and now we stop, because that's how we make money mwa-ha-ha-haaaa!"
I do hope no cats were harmed in that mwa-ha-ha-haaaa!
And just to throw in another complication, Samsung, Hynix etc aren't necessarily the ones who limit how many fabs get built in the first place.
As semi conductor production taunts the limits of the laws of physics ever harder the sourcing of the tools which the fabs require becomes an absurd challenge by itself.
For example ASML (a Dutch company specializing in that kind of thing) is absolutely overwhelmed by demand.
Separate names with a comma.