SK Hynix Starts Prepping for Next Semiconductor Boom with $11B Memory Fab | Hacker News Reader

SK Hynix Starts Prepping for Next Semiconductor Boom with $11B Memory Fab

anandtech.com

62 pointsrbanffy12 comments

What is interesting is that microselectronics seems to have improved almost every aspect of our lives except the ability to drive down the cost of semiconductor fab machines.

They were saying that a single wafer etching machine could cost $100M. Yes, one machine. I assume this is why people aren't jumping on the bandwagon, the initial investment is enormous. I also suspect that the people who build these machines can't simply ramp up production of something so specialised.

But they have, we wouldn’t be able to design, build or operate those 100M machines without plenty of cutting edge micro-electronics at every step of the way.

When you look at the details of how these machines are made, the price tag almost seems low in the end.

Yes, stepper machines do cost hundreds of millions but then one realizes they handle more than a hundred wafers per hour. At the end the machines do seem cheap.

Down side is the cost of entry if you want to be your own fab is too much. Do not feasible for a start up.

A startup has many options to get chips made. But if the startup was itself a fab then it needs to think outside the box. For example smaller wafers yield less ICs, so more $/IC. But, you also don’t need a $100M stepper, which helps.

There are companies working on cheaper, low volume, solutions and some things exist. For example: https://cdn.aff.yokogawa.com/9/400/details/minimal-fab-one-s...

It’s the kind of thing that if you need it, you know about it. But it doesn’t make national news because it’s not billions of dollars.

> But several years down the road demand for PCs and other client devices will rise once again, and so will demand for memory. Which is why SK Hynix is already preparing for this with its Fab M15X expansion plan.

It's great seeing the market react to the shortages. Of course these things take time but the way the semiconductor supply-side organizes such huge amounts of capital and large scale development projects is quite fascinating.

Hopefully it will pay off in a decade and we won't have to worry about shortages for some time.

The fab game is just inherently boom and bust. The economics of a fab mean that it's built at massive cost but the unit economics are incredibly efficient so fabs then run at 100% utilisation regardless of demand until EOL. The result is that investing in a new fab isn't possible until demand is eclipsing current supply such that it can warrant the capital outlay. Boom and bust like clockwork.

I keep reading that semiconductors are on the cusp of a downturn and that the earnings multiples for eg AMD and Intel stock are too high as a result.

In contrast, there seems to be a crazy amount of investment going into fabs and improving chip processes.

These two things don't sound like they can both be true to me. Out of the two, surely the folks pitching 100B into building out their fabs have a better insight into demand than financial analysts.

Is there any way that this investment would make sense with a downturn in chip demand (/ is there some wall st malarkey going on / some other explanation that would make everyone's behavior congruent)?

My guess is you're looking at cause and effect here. Profits are going to drop because all that extra capacity is going to come online soon.

Only to then skyrocket as chips continue to be used in ever greater numbers/tech. War would increase demand. Switching to EVs would increase demand. Etc.

But, oddly, it’s hard to look at market valuations over the last three decades and conclude things are low priced.

  > War would increase demand.

But retail demand would tank, no?

Reply on news.ycombinator.com