Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

If you are wondering what Sam Zeloof is up to these days: he founded Atomic Semi [1] with Jim Keller (yes, _that_ Jim Keller).

[1] https://atomicsemi.com/



Quite hard to grasp what they are really trying to achieve there. They are building what kind of tools, and for whom?


Think back to the 2021-2022 pandemic-triggered supply chain chaos, when the media widely reported that car companies, for example, were all having very painful experiences with parking lots full of $50k trucks they couldn't sell due to a lack of $5 chips.

At the time I saw some of Sam's videos (I think I originally found them here on HN), and I had a big question: If a hobbyist can build 1970's era chips for ~$10k in his garage, and Intel can build 2020's era chips for ~$10B in state-of-the-art factories, why doesn't someone start a startup building 1990's era chips in a ~$10M facility?

I'd assume Atomic Semi is that startup :)


Earlier this week I got a warning about supply chain issues delaying deliveries while building a ford transit van in the online build & price feature

Not sure what global chip availability is, but this event made me feel like the supply chain chip chaos is alive and well


Semi is the next big thing in distributed manufacturing you already have companies that offer what is essentially ride sharing for semiconductors they manufacture single IC/SoC with multiple blocks on them and you essentially pay for a % of the floor plan which can operate independently from the rest.

So whilst not quite garage ready we’re probably 2-5 years of you being able to order completely custom silicon form the likes of PCBway and have it shipped in the matter of days rather than months.


For very basic cmos nodes 24 turn is possible. But for most advanced node the lead time is months because that’s the minimum manufacturing time.

https://www.seattletimes.com/business/technology/three-month...


Some past discussion suggested there was an opportunity for very low volume production at much lower cost because the method doesn't use masks.


There’s no way it’s lower cost. However what it could be is guaranteed line capacity and delivery dates. And if you’re doing small orders from a large foundry you don’t get that. Those things have value too and someone might be willing to pay a bit extra to get them.

My main issue is semiconductor lines are fragile beasts that need constant tinkering. So if the system requires full time staff to keep running, that’s going to be really expensive.


If a mask set is 50 million and you only need ten chips, these guys would only need to come in under 5 million per chip to win on cost.


We don’t know what node they’re targeting so this is all speculation but if it’s anything below say 22nm then that’s finFET, and there’s nothing low cost about finFET at prototype volumes.

If it’s not finFET then a non production mask set is much less than $50 million in 2023.

Also, keep in mind maskless lithography has been used commercially for roughly ten years now. And it’s been in Universities for significantly longer.


If you only need 10 chips, you may just be able to use 50 or even 500 FPGAs.

500 VU19P will cost less than 5 million.


There's a large number of applications which cannot be done in an FPGA, for instance any complex GHz-clocked digital circuit, or many analog circuits.


Not if you want to put it somewhere small like a satellite or bunch of other things.


https://roboticsandautomationnews.com/2020/06/09/xilinx-laun...

Xilinx launches ‘industry’s first’ 20 nanometer space-grade FPGA chip


They're being very cagy, but it's quite likely they're working on "fast" maskless technology. Basically going from FPGA to chip in low enough volumes that maskless makes sense. It'd be a small, but lucrative, niche if they can get a good system at a decent node size.


My take was that a low and slow approach that did not involve the traditional dangerous chemicals was being developed. Currently TSMC etc processes require hydrofluoric acid and other things that can kill you.


You can get Hydrofluoric acid at the grocery store... it's used in "Wink" stain remover. It's strong enough to get the job done, but dilute enough to be... well, in the grocery store.

The stuff that you don't want to mess with is Silane, which Sam describes as "a toxic and explosive gas".[1]

[1] http://sam.zeloof.xyz/second-ic/


For anyone else interested in him, Sam has a Youtube channel, but hasn't posted in awhile. He started this stuff as a teenager with a garage that I think many of us would envy now, let alone as a teenager. I've seen him on HN once or twice too. There's a whole world of these home built processors channels and they're pretty impressive. I know some others are posted in other threads, but this isn't my niche so I'll leave it to them.

https://www.youtube.com/@SamZeloof


Looking at their careers page the typical salary is $100,000 – $170,000 for SF. Seems a little low for SF no? Is this an indication that they are not well funded or going to have issues finding top tier people?


Process development is closer to chemical engineering and that’s not bad when compared with their salary.


Its low if theyre paying 100k for people with a couple years of experience, but for entry level hardware jobs, 100k is about right.

Ive been looking for almost 5 months for an entry level hardware job (preferably in SF, but willing to relocate anywhere), and 100k is about what Im looking for based on what Ive seen.

Some go as low as 85, and some start at 120, but 100k is pretty standard.

The salaries are not what they were in 2021




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: