Hacker News Reader

Comment by dhbradshaw | Hacker News Reader

As a chip supplier, I'd be feeling a lot of pressure right now. In the short term you can squeeze and the tail can wag the dog for a bit. But longer term this opens the door wide open for new competitors in the market and / or vertical integration.

View on HN

I don't think chip suppliers are playing games here, are they?

I think in general all chip manufacturers are at capacity and it comes down to getting in line. And I'm sure many chip manufacturers are trying to expand their capacity. I can't imagine they're sitting there thinking this will last forever.

I'm not sure the reasons for the shortage matter. If you have a dependency that's causing you major problems, you try to become less dependent on it.

> But longer term this opens the door wide open for new competitors in the market and / or vertical integration.

Not really. Chip fabrication requires enormous upstart cost - IIRC, TSMC plans with something like 10-20 billion $ - and a lot of time, to the tune of three years at least (https://pr.tsmc.com/english/news/2033). There are not many companies in the world who have that amount of cash lying around, and even the ones who do like Apple still resort to using TSMC.

Not to mention it's not just the machines from ASML and a host of other vendors plus the cost of building factory-sized ultra clean rooms, but you also need the expert staff trained to operate the entire setup and a lot of fine tuning of parameters which are closely guarded secrets...

> There are not many companies in the world who have that amount of cash lying around

Which is why establishing a chip fab to me seems like an obvious national goal. It's a booming market with costs of entry that few private companies can afford, why not make it a national capability? Resilience and profitability in one package.

Even for a nation, that is a lot of money - Germany's budget is about 500B €, a 10-20 B € investment would be a 2-4% of that.

IT doesn't have to be. You can be producing chips in a month if you are willing to pay. A clean room can be created, and the process of making of making chips one at a time in labor intensive processes is known. Just that nobody is willing to pay $50k each for their chips, and you need to get that price to pay for all the skilled labor to make a chip one at a time in a lab.

It takes years to setup a fab, but in the end it is a lot more cost effective, which is nobody is bothering to do the manual process even though we know it will work.

Maybe? It seems to me the overheads of establishing a fab process anywhere near the incumbents are astronomical. I would guess that even this bottleneck in supply would represent a minuscule fraction of that overhead. What are the actual numbers?

>> Maybe? It seems to me the overheads of establishing a fab process anywhere near the incumbents are astronomical.

It depends which parts are in short supply. A lot of automotive parts are still on 20+ year old nodes. 90nm to 130 or even 180nm. I would guess those are not the parts experiencing shortages though. So that leaves the fancy stuff. Bummer if a car can't be built because the infotainment SoC can't be obtained because it's competing for production with crypto mining GPUs.

> the overheads of establishing a fab process anywhere near the incumbents are astronomical.

That's a big opportunity in itself.

I'm sure someone's trying to spin up a fab right now, but that's inherently not a quick process.

Or cheap. Trying to game a short-term downtick in supply by spending 100 Bn USD only to come out the other side as not competitive when you're up-to-speed and your competitors have an abundance of supply and can undercut you by 20%.

It's not a winning strategy.

There are tons of n-th tier fabs who seen the shortage coming 4-5 years ago, and trying to move the ladder in the legacy processes.

Rushing to compete with them when they had a 4-5 years headstart is not a wise decision.

I've noticed that the shortage of chips is also strongly driven by speculators simply buying up everything in sight and trying to resell at multiples of the cost.

You think people are buying up Chips designed to go in to Toyota cars?

Surely they're beyond worthless to anyone but the manufacturer and Toyota themselves?

Parent was speaking of the chip shortage in general, which goes way past cars. Car manufacturers will do what they can to isolate themselves from the open market (and buy the numbers and have the political backing to do so), but demand for non-car parts affects allocation of production capacity to their demands too.

> designed to go in to Toyota cars?

Chips rarely are designed for a specific car, and a car will also contain quite a few that are not strictly specific to automotive.

That is a whole different level - consumer retail as added middlemen. Manufacturers who include multiple in one project don't buy just retail which is oriented at one-at-a-time.

Reply on news.ycombinator.com