[HN Gopher] How to make it easier to build semiconductor factories
___________________________________________________________________
How to make it easier to build semiconductor factories
Author : steelstraw
Score : 55 points
Date : 2022-01-21 18:11 UTC (4 hours ago)
(HTM) web link (semiliterate.substack.com)
(TXT) w3m dump (semiliterate.substack.com)
| amelius wrote:
| Let's not forget about better open source development tools too.
| neilpanchal wrote:
| I was involved in writing software for a small Fab and there is a
| big gap in the MES software space. If you want to start a new
| Fab, you need a manufacturing execution system (MES) to process
| wafers. This whole MES industry is ripe for innovation. For small
| Fabs, it's not possible to afford MES-3000 from Applied Materials
| ~ multi-million $ license fees and it's a beast of a system. I've
| had to deal with this ancient dog of a MES written in Coldfusion.
| Had several conversations with coworkers - "Wonder what the
| market is like for modern SaaS MES". It's a complex problem if
| you want to auto load recipes and control equipment with
| SECS/GEM, implement PCS, and run a Fab in the manner like Intel.
| This is one of the problems with scaling the Fab. Think
| automotive factory scaling but with expensive precision equipment
| and garbage software ecosystem.
| uoaei wrote:
| SaaS is too unstable for semiconductor manufacturers to rely
| on. They would want a locally-installed, stable X.Y.Z
| implementation.
| octoberfranklin wrote:
| This. "Copy Exactly" is not compatible with SaaS:
|
| https://en.wikipedia.org/wiki/Copy_Exactly
| neilpanchal wrote:
| Well, it would be a typical enterprise installation.
| Presumably on-prem or some hybrid.
| nwiswell wrote:
| At that point how is it differentiated from the incumbent
| product?
|
| If you are running a serious fab you already have AMAT
| equipment anyway and there's probably good reasons to get
| the industry standard AMAT MES software.
|
| Realistically, this software will probably continue not
| to exist because there are not many target customers in
| this "mini-fab" segment, and they are uniformly poor.
| That's not a great outlook for sales. If your software is
| any good, you want to market to the same segment AMAT
| does and go for the multi-million $ licensing fees.
| uoaei wrote:
| Then what makes it a service? It would be unwieldy to
| bring in consultants... unless they can take over for the
| extant fab techs as soon as they arrive. Any downtime in
| the fab is millions lost.
| phpnode wrote:
| curious whether you've seen Oqton? https://www.oqton.com/
| dqpb wrote:
| Do you have any references where someone could learn more about
| this?
| nwiswell wrote:
| You can read the SEMI standards. All of this is thoroughly
| standardized to allow interoperability between equipment from
| different suppliers.
|
| SECS (SEMI Equipment Communications Standard) and GEM
| (Generic Model for Communications and Control of
| Manufacturing Equipment) will be of particular interest.
|
| https://www.semi.org/en/industry-groups/smart-
| manufacturing/...
| mjevans wrote:
| They might not be able to share more given NDAs and such.
| mjevans wrote:
| Some of the lead time to build fabs in the US is misleading.
| Offhand I know of at least one site that was 'under construction'
| for a long time, but intentionally very slow construction. The
| company wanted a fab there eventually, but wanted to lock in tax
| rates and other subsidies today. I guess the math came out ahead
| just making the project take forever to build.
| willyt wrote:
| 600 days to build a semiconductor plant seems quite quick! In my
| part of Britain just now it is not uncommon for it to take over
| 250 days to get permissions in place to build a 3 bedroom house.
| zbrozek wrote:
| I'm more than 600 days into trying to get permission to ask for
| permission to edit an existing home.
| destitude wrote:
| Environmental regulations were sited without clear specification
| of what exactly the issue is. There is strong precedence on why
| we need those regulations to prevent issues that the "future"
| then has to deal with. How many super fund sites are there in the
| USA today and how many are a result from companies being able to
| get away with lax environmental regulations at the time? China is
| already dealing with this by forcing heavily polluting industries
| to shut down, cut off power, unbreathable air in cities, etc..
| all because of lax environment regulations to begin with.
| octoberfranklin wrote:
| I agree, but the US has a very shortsighted attitude here.
|
| If you place regulatory burdens on some undesirable behavior
| (e.g. pollution) but do not place compensatory tariffs on the
| import of products resulting from _that exact same behavior_ ,
| you're not doing anything to reduce the undesirable behavior.
| You're simply encouraging people to do the undesirable things
| elsewhere. This not only fails to accomplish the goal, but also
| significantly weakens the country.
|
| The exact same problem totally undermines labor policy and
| unions (except those of government employees).
|
| There really ought to be some sort of automatic mechanism for
| "if X is illegal/taxed/regulated then importing the result of X
| happening overseas is forbidden/taxed/regulated". The US
| already has a specific, very efficient court that does this for
| one specific illegal act (patent infringement); it wouldn't be
| difficult to expand that scope:
|
| https://en.wikipedia.org/wiki/United_States_Court_of_Interna...
| fennecfoxen wrote:
| To be fair, for most types of pollution, moving it
| "elsewhere" solves your problem.
| AnthonyMouse wrote:
| The way the regulatory environment works in the US is something
| like this.
|
| Something bad happens, so new regulations are imposed. The
| regulators don't always know what they're doing and frequently
| make poor cost benefit decisions. Meanwhile, the _existing_
| regulations usually _already_ prohibited whatever the offender
| did, and the bad thing happened because they were breaking
| existing law, so they also get sued or prosecuted etc. under
| existing law anyway.
|
| Despite this being completely nuts, it kind of works to deter
| pollution. The new regulations are like extra punishment for
| screwing up. The company, and the entire industry, has to pay
| more in regulatory compliance forever because they did a bad
| thing. So they really don't want to do that.
|
| The problem is that the effect is cumulative and doesn't just
| impact the offender. Every time someone breaks the law in a
| high profile way, the compliance costs go up. They never go
| down. The new laws are useless because the bad thing has been
| prohibited for decades and the actual cause was companies
| violating existing laws. So the more time passes, the more
| regulations we get and the less competitive the US becomes.
|
| Somebody who knows what they're doing needs to go through the
| existing regulations and strip out the ones that are wasteful
| and redundant while leaving the ones that efficiently prohibit
| bad acts.
|
| But the political process is terrible at handling details like
| that. If someone proposes getting rid of an environmental
| regulation, it doesn't matter what it says. You can predict who
| will support it and who will oppose it. And it's the same
| people in support and opposition whether it's an actually
| necessary regulation with a moderate burden and a very large
| important reason for existing, or a useless web of red tape
| which is fully redundant with existing prohibitions but makes
| everything cost more for no benefit.
| specialist wrote:
| I'd like full life cycle accounting. Or something to that
| effect. Basically "you broke it, you bought."
|
| I have _no idea_ how the accounting and finance would work.
| Bonds, insurance, set asides...
|
| Putting cleanup on the balance sheet would then incentivize
| both prevention (harm reduction) and investing in new
| technologies. Like maybe some sci-fi stuff like new microbes to
| eat toxins, which could then be another biz opportunity.
___________________________________________________________________
(page generated 2022-01-21 23:01 UTC)