> I would also be cool to sell a single compute unit with the cool fireware for home users.
I seriously doubt that they're interested in commercially hitting that market, but they published enough software as open source that you can recreate chunks of of what they're doing on consumer hardware, ex. https://artemis.sh/2022/03/14/propolis-oxide-at-home-pt1.htm...
I think the main thing that would be cool is to get the secure hardware/fireware architecture. But of course this would be very niche and not a money maker.
Being demanding of your suppliers seems like a risky choice. Especially if you're small and your demand is small. But it does have swagger. Being candid about real dates and real requirements sounds great, refreshing even. But then "we need our suppliers to jump through hoops for us, to expedite, to get supply not otherwise allocated to us" sounds like you're asking suppliers to 'hustle' to meet their commitments to you. What if the problem is there's no cadmium available anywhere, world wide? To what extent does oxide care about how supply is produced and how they get their supplies, transitively? What if to secure the supply they need to offer more money, but instead you tell them to "hustle"?
But yeah sourcing middling sized batches of project components has got to be rough in the best of times, since it's probably mostly coming from overseas and mostly from Shenzen. That route gets more expensive with Covid and also US geopolitical tensions with China. Its an interesting time to be a hardware-startup constrained by how freely atoms move around the world. My suggestion is that you hire a few smart, attractive Shenzen biz types, give them some generous entertainment accounts, and watch your orders get filled with ease.
We definitely dig pretty deep into our suppliers' supply chains. For details on that, I strongly recommend the Twitter Space we recorded with Kate Hicks (one of the authors of this piece)[0] -- and in particular, how we engaged with Sanyo-Denki (our fan supplier) on one such challenge.
There are a few things that I'm seeing differently in the market than what is outlined in this post.
1) Caps and res with 20+ week lead time? I'm assuming they are not building their own boards, but are getting them built by a reputable manufacturer. From what we've seen, the manufacturers we've spoken with all have stock to supply ceps and res for our needs. Why are they not seeing the same?
2) As a start-up, how much hardware are they building now? and why? If you know you are going to need 2000 power regulators (one of the components we've had difficulty getting), find one that suits your need, and buy your needs for the next 12 months. We bought up every EEG chip we could get our hands on because without it, we have no business. Yeah, it cost a few thousand more than we wanted to spend at the time, but it kept us in business. It also kept these chips out of the hands of competing start-ups. :)
3) buying from exchanges is a real challenge. We actually had some breakout boards (I can't remember the supplier) and the chip didn't respond in the way the spec suggested it should. We're assuming it was counterfeit, and programmed around it for a time. We also bought a bunch of stock from one supplier that we're unsure of, so we made a small purchase, and tested it to make sure it was legit before buying more.
I know there are supply chain issues, I'm not suggesting there aren't, but I've found that with a little bit of planning ahead we were able (so far) to get around them and have pre-purchased stock to last us the next 9 months.
Supply chain is now a strategic advantage, not just a challenge. However, it looks like this is already turning around and chips are becoming available again.
It depends on what you mean by "building [our] own boards." We (obviously?) don't have our own PCB fab -- but we are absolutely designing our own boards (and therefore securing our own supplies). On caps and resistors (and other passives), I think the point was that we saw even these very plentiful parts develop supply chain issues in the most challenging times in 2021. Broadly, these passives haven't been an issue for us -- but the fact that they have been an issue at all is notable. And there were several times where we did exactly what you suggested, where we either went with a slightly more expensive part or secured a substantial amount of inventory to assure our own supply. We certainly agree about the supply chain being a strategic advantage!
In some of our prototyping, we were "manufacturing" as in placing our own components on the boards we ordered, and cooking them in a toaster oven converted for PCBs assembly.
We now get our next gen stuff manufactured by a local "manufacturer" who takes our PCBs (which come from China), and puts our components on them. We supply all the "non-passives", simply to make sure the stock is available. They've got all the passives for us.
This is our first intro to building hardware, we're a bunch of software engineers, but I've commented on how the "hardware is hard" we feel is a misnomer.
It's another point of failure, but as long as you're planning ahead, it seems manageable. My tune may change when we get into production on the scale of many thousands. But we're only at 250 units per batch atm.
Somehow I just learned about this company recently. I really enjoyed this post. I didn’t know much about supply chains before reading it and I don’t now I’d imagine but I know more than I did. It is written in a common sense and straightforward manner. I really identify with their values and I think they shine through here.
If I’m being entirely honest, I’m a software developer and I am not quite sure what they are selling. I wonder if their compute product compares to Amazon ECS, their storage to S3, and their networking to Amazon VPC. I am not entirely sure. I am new at my job and I would love to one day to convince my company to go with an ethical cloud provider but I believe that the agency required for such is still some long distance from my current position.
They are not a cloud provider, they sell servers. So you have to pay a couple million for a big rack full of servers. That product will offer some of the same things that you would do on a cloud.
If you were a cloud provider, you might want to buy this system rather then some Dell/HP machines.
How many do they need to sell make the company work.
I would also be cool to sell a single compute unit with the cool fireware for home users.