Tara Tarakiyee – Why Has Hardware Infrastructure Diverged From Open Software? #FOSSBack

This talk explores the divergence between open software and hardware, highlighting the barriers that limit access to chip design and manufacturing. The speaker discusses the monopolization of hardware tools by a few corporations, which restricts innovation and access to technology. They delve into the infrastructure needed for open hardware, proposing the formation of new institutional frameworks that operate as commons to support collaborative hardware development. Key points include the community's need for more equitable access to resources, the suggestion for sustainable funding models like public bonds, and the importance of governance that balances technical, public, and commercial interests. The speaker also acknowledges the underlying complexities of hardware creation, including ethical concerns related to material sourcing and ecological sustainability.

[music] >> Thank you, everyone. Some of you might know me from like my day-to-day work with open source software and sovereign tech agency, but I just wanted to say that today I'm here more on the side quest. This has been part of a fellowship I was doing under the SOAM called called SOAM refund our digital future. Um and the goal for me was to explore an area

where I'm not um I've always been interested in, I thought was really important, but haven't had the chance to dig into before. Um another thing you might not know about me is I'm a computer engineer by training. I actually studied embedded systems, but I never worked in the field. Um and I think after that you'll see also why like what are sort of the barriers that stop

lots of people from actually pursuing a career in in in in hardware and silicon. where where do we start? Uh every public AI project, every sort of climate sensor effort, every medical lab device has this hidden barrier. Like the tools to design the hardware and the factories to manufacture it are controlled by a small number of people. And today I want to attempt to sort of explain

why I think this divergence happened, like why software became open and a commons and hardware didn't. And what's sort of like the structural reasons behind it. And it's also more of a speculative exercise, so I invite you all to also speculate with me, ask me questions if you think I'm I'm I made a mistake in my analysis, please let me know because that's how I want to

I'm sorry, I'm having Yeah, it should work now. Um yeah, so uh let's start. So, this is a a chip under a microscope. As you can it's an incredible feat of engineering like um those electrons like often run in some of those newer systems run closer than a like a packed U8 on rush hour. everything we talk about that's digital whether um software where it runs on

your computer, on someone else's computer everything ultimately I think needs to run on one of those things. So, in a visceral way it represents the physical manifestation like everything we work on. And for something that I think that's important I don't feel like we talk enough about how those things are produced and how um yeah, sort of the structural barriers there. And this is what I want

to explore more here. So, the elephant in the room, open hardware does exist. You can design boards, you can share schematics, you can build with open components. But the infrastructure underneath, so the tools to design the chips, the factories to manufacture them is all proprietary. I think there's sort of like some what I call like the silicon boundary. Like this is where like all the openness stops

and like underneath that you're forced to use someone else's uh chips made by someone else. So, if you look at the design tools or EDA 74% of these are controlled by three companies. So, Siemens, Cadence, and Synopsys. if you want to have a seat to use one of those tools, it's somewhere between half a million to a million dollar. Like it's it's a lot of money. So,

actually like in even I've heard stories of like many uh people who work on designing chips, they would use uh whatever open tools that exist up to a point. And when they need to go into production, they would have one seat that they would all share to actually turn the things into production ready. the inside of a fab looks like. It's a again, it's very it's a

very uh expensive place to be in. Um and for the machines inside those uh fabs or places where people make chips one company essentially has a monopoly on the most advanced technologies, what are called lithography tools. it costs like some of the most recent ones cost up to a 380 million just for one machine. So, if you compare that to the budget of an average research company,

that's like seven to eight times the entire budget. so, hardware infrastructure essentially functions as a private utility with no public alternative. Um and you can have the best ideas, the best designs, you can be thinking of the most urgent social needs, but if you can't access this utility, you don't get to build what what what you want to build. let's go back to this uh divergence that

I'm talking about. So, this is something we're familiar with. We software we have open protocols, open tooling at the basis. We have compilers and runtimes and operating systems that are open. And then on top of that we build all these open products and services or even closed ones, it doesn't matter. They all use the same infrastructure. And everyone gets to participate in in this at every level.

But if you look at hardware sure, the research, the design ideas might exist, you can do them. But once you want to move to production, first you hit one gate, which is the design tooling and all the verification tool chains. And then even if you get past that, then you have um the PDKs, so the uh production development kits. So, that's what the whole that that's like

the whole chain that turns like everything from a design to an actual something that could be produced. And then you have access to foundries. Actually, if you don't already have an established relationship to a chip foundry, chances are you as a newcomer of getting a chance to even uh make your design chips is you'll have to wait a very long time and pay a lot Um so,

essentially participation is blocked by all of these different gates. So, in software we learn to share the parts that no one wants to reinvent twice, but in hardware for some reason we never did. so, what are the things like the theoretical chips that we could design? So, if we think about like custom like all systems on a chip for climate purposes. If you think about purpose-built medical

silicon. If you think about uh sensors in agriculture. If you think about processes that are resilient for disasters, which we continue to have more of day by day. No all of these are you can't do them on the you can't do them on the silicon level. You can use open PCBs, you can use Arduino, you can open hardware products do exist, but ultimately you are locked into

the drawbacks of using these uh systems and you can't design um you you're locked into like consumer grade electronics and you're you're you're you have this barrier where you can't actually build what you really need at the level that you really want. And I think that's really like just a cap on innovation. So, the tragedy is not expensive chips, it's the chips we never get to build,

I think. so, if you if you want to look a bit closer about sort of what the barriers are that I see here. So, first of all, you have this capital intensity. So, these tools and machines that we talked about the uh what ASML has or what the EDA tools have those didn't come overnight. Like those have thousands of engineering hours put into them. And that's something

that's beyond the capacity of a volunteer community or what a grant cycle can sustain. Um on another level, um there's an institutional vacuum. So, there's no permanent operator or there's no career path for maintainers that want to pursue this on the long term. So, more often than not you end up with knowledge dissipating between different funding rounds that tried to address this issue before. And then finally,

you have this revenue dynamic. So, which often heavily favors the commercial revenue versus the social benefits of hardware. obviously the since we're at an open source conference, the question stands can licensing fix any of these problems? The answer is no. Um better licensing can't finance a tool chain that requires a thousands of years of engineering years. Um if we want to really address the the issue of

closed hardware. If you really really want to address hardware infrastructure on that scale, we require a sustained coordinated investment at scale where community governance and licensing alone just aren't able to address. Um and I mentioned existing efforts and I want to point to that are important in this context. So, people talk a lot about RISC-V. So, it opens up sort of the instruction set architecture. Really important

work. Um but you still like if you want to use RISC-V, you're these days you're mostly dependent on proprietary EDA and restricted foundry access. So, while the instruction set is open, the infrastructure is still a bit closed. And another initiative I want to point to is OpenRoad, that was DARPA-funded project to open source a whole production development PDK tool chain for chip design. And it did manage

to achieve real progress towards reducing the barriers. But it's a grant-funded project and obviously grants run out at some point. And it didn't really solve that sort of last mile problem. So, how do we get this research into something that can actually be used by people everywhere. so I think the thing that's actually missing is not the knowledge, it's not the Well, it's it's an institution. It's

a new kind of institution that I want to speculate about with you all here. I call it the open hardware infrastructure works. And the way I see it, it could be a permanent institution employing thousands of engineers building and operating this hardware infrastructure as a as a commons. Um cuz right now we compete on who can afford the tools, not who can build the best tools. And

I think that's obviously something that we need to fix. Dutch uh dike, so it's water infrastructure. Those things are big, they require a lot of effort to build. And obviously you can see the parallels here of like big infrastructure project, too big for person to handle, too big for a small village to handle. But if everyone gets together, uh we can build something big like this. And

we've been doing that for hundreds of years. and there's other precedents, I think. I won't go through them one by one, but I it's important to note institutions like CERN, uh companies like the Autobahn GmbH, uh Airbus as well. Um NASA and the ESA are other examples of like big things that we decided are big enough that they're worth having their own institution for. a little uh

teaser for later. So water boards were mostly funded using bonds, like public bonds. so this is where I think could like a project like this, which is infrastructure, it should be funded like infrastructure. uh I'm not going to go through some detail. I'll share a link at the end, but the gist of it is instead of um having to wait until all those smaller build up to

to have uh have some momentum, which might never happen, we want to give this uh project a bit of a runway. And that's why like government secured long-term bonds gives you this capital that you can use to invest to build those essential building blocks. And once it's up and running, it can switch the the revenue mechanism from extracting rent from licenses and tools to actually providing services

that are useful. And that way it can fund itself, it can pay back those loans, it can pay back those bonds, I mean. this is essentially also how uh those um water board bonds that have been working for hundreds of years work. I also thought a bit about governance. I think something like this should have more of a tripartite model of governance where you have the public

funders on one side that can help prevent it from being captured. You have the public interest represented to make sure that it stays on course, it's remains on that social mission, not too driven by commercial interest. And then you also have need to have a technical community on board to make sure that it also stays on the technical mission. So this model would help sort of prevent

any one interest from being overrepresented or anyone from capturing the governance of and I'm going to pause here because that was the end of the speculative thing. Um I want to um preface that I still still think there's a fundamental problem at the layer of this that I haven't addressed through this fellowship is that even open hardware still runs through mines and fabs. There's incredible harms that

comes from those practices, extractive practices hidden underneath all those layers. And I haven't had the chance to explore through this. I don't think this institution is also going to solve this. But if you're interested in talking about this, >> [snorts] >> I'm having a workshop at Cables of Resistance to talk about more this area of it, the mineral extraction, the labor exploitation, the e-waste. Um I'm welcome

people like it's not going to require any prior knowledge, but I welcome people to come there and join me in this workshop and help us help me think through some of that aspect of it. If you're interested in the presentation, you want to see it, you can also go to ohiw.teraki.com where I have the presentation essentially in a mini site form. Uh thought it would be better

than giving handouts or something like that. and that's it for my presentation today. So if Do we have time for maybe Yeah, absolutely. So round of applause for Taraki. >> [applause] >> And according to my watch, we've got even 15 minutes of Q&A. So let's have a discussion going. Okay, thank you very much for this talk. Um I guess you are not really honest because uh there's

I mean you're talking about the high end, but on the on the low end, there's so much going on in terms of open hardware. I mean we're talking about Skywater 130 GF180. E-fabless is even alive again after wafer space. Um there's there's tons of EDA software available. And OpenROAD, by the way, is a very good project in my opinion. So it doesn't deserve being called, let's say,

a failure. I You didn't say so, but >> I didn't want to imply that. Just to be clear. I I'm a fan of it, that's why I mentioned it. I I think you didn't say so, but it's just, uh there's actually a live community available. So there's a conference in summer in Ljubljana, the Free Silicon Conference, which deals exactly with that problem. There are multiple foundries around.

One of them is, let's say, 70 km east of here in Frankfurt an der Oder, who offer services where you can, let's say, I don't know exactly the price, but like it's a few K, affordable, and there's funding available by the government. So there's a exciting uh development in the last 5 And I'm one of the the tool maintainers of one of the tools, KLayout, if you

know it, maybe. it's worth explaining that there is such a such a ecosystem available. No, definitely. And I This is actually like maybe if we're also being honest, like one of the hardest parts about doing this is because I didn't want to skip or go over like all the important open hardware things that are existing. I think they're incredibly important, incredibly valuable because also like otherwise like

we would be starting from way behind if we didn't have those efforts. I think those efforts should be the core focus of an institution like this, of supporting them, creating an ecosystem for them. Um but the key thing uh the key point I was trying to make with that slide is we have been having those efforts for a while, but if you look at the scale of

the problem, if you look at how far commercial hardware or commercial hardware production is concentrated and and proprietary, we need something that's we need all those efforts to to be more and and and on a bigger scale, and we need something to anchor all of this this work and make it more permanent. If that if that uh answers your question. Um nowadays, it's look like um people

don't even cross their minds about that because you look there's this feeling that uh the materials it's already established feud. Where do you think like uh it would be most beneficial for societies if uh governments or specific um not by nations, but for a specific so not monopolistic way, um would it be a bit more beneficial for society itself uh the worth to invest in on hardware?

Sorry, can you repeat the last part? Um where do you think it would be more public interest? Where would it be more valuable for societies to be make this big commitment and investment? Um I mean I don't know if I'm being too television, but for me like I think considering um I mean the climate disaster that we're in, like uh that would be where like my focus

is. And I think that's my my showing in some of the examples I brought up. I I'm not a techno-solutionist. I don't think technology will solve everything. But I think we rely a lot on technology. And if we want to have a sustainable future, if we want to have a burning future for our planet, then the way we make our technology should should reflect that. And um

the way we make our hardware should reflect that. So and there could be certain applications where yeah, um if if that was the goal like if the goal wasn't to make like the most efficient fastest chips, the goal was to make the most ecologically sustainable chips, then um that's where I see like the most benefit of something like this happening. But you can also extend this also

to other um areas of concern as well. Like I think the fact is that I think there's too much focus on the commercial interest, on the efficiency, on the speed, on making things smaller, rather than on looking at the big picture, looking at the public benefit, looking at the public interest. Um I was wondering, speaking of public interest and sustainability there, the first thing that came to

my mind was the lack of supply chain knowledge in the actual hardware already, like just just for an example, maybe most everyone in hardware uses solder. But yeah, I don't know if you know the company called Naga that has actually a supply chain for their solder and they couldn't dwindle like down where all the little stuff came but that that's what they're there for, you know, that

is the purpose of that solder to have supply chain information. So, the idea with open hardware you know, is always a noble feat but that that is kind of problems that are of the discussion when we want something open and sustainable. But there will be these little walls. How can we cross them or even if we want to cross them? Like how can we have it at

least to a point in including these efforts but at the same time acknowledging that it will also not be fully sustainable ecologically friendly. Like you can't I think at this point like what Naga has already shown is you can come from the good side with all these efforts and at some point you have to face defeat there will be mining and there will be and then how

do you weigh that against the public interest if you want to be on the good side? So, that's my question. Yeah, and I don't know if I can speak directly to this but I think that's where like maybe the governance issue comes in the governance part of this comes into play cuz I think ultimately it's not only technical problems, there are social problems as well, there's people

and that's why like I didn't suggest like a board of just like technical community or a board of just funders. I think having the public interest represented in a way and I don't want to go deep into how to do that exactly cuz I think there's multiple ways to do that that could but I think having that represented in in in a fair way would help us

come to grips with that. On a more fundamental level, that's also why I'm doing this workshop is because and again like this is me being totally [snorts] like maybe honest and vulnerable here. I don't know if there's an ethical way to make hardware. I don't know if there's a simple way to I don't know what that means for me or for the world or for everyone and

that's part of the questions why I want to explore here is is there is there a path towards or if this is a harm we're willing to accept, how much harm are we willing to accept and to whom and what that means just make that a bit more explicit maybe. I'm just I'm thinking like that's also like fundamental to how we treat hardware in our lives like

you almost never see chips the way like you saw in like the first slide like we always hide them behind like these fancy nice clean looking things. But you look at them they're actually complex and brilliant and all of that but and I think that's why like for example like a home like I have like my tower desktop with like a side screen. I have I used

to have on my phone like one of those covers that show like the chips on it. Cuz I I love that. I love looking at that and but I think the downside also like not looking at the chips is also that we're hiding all the harms that come from the and hiding all uncomfortable truths maybe. So, um and I I think like that's not uh good for

society in general. Thank you. First a comment. I think what you're saying about making things smaller and faster, there's another side benefit to that is usually makes them lower power. you know, for the same amount of work we can get a lot less environmental effect from having things smaller. But my main point was the I get the impression that is that there's a chasm between doing wafer

sharing with open source software and uh moving to sort of low niche production like a medical chip where you might need a thousand or ten thousand or something like that and the the masks are incredibly expensive and the times are so long. So, I'm wondering if there's attempt to maybe first am I wrong and secondly is there a way to attack that particular problem? I get the

impression in China it's relatively easy to make lower end chips in small quantities. They they seem to be doing it all the time so many fabulous companies. And certainly I think I think again like this was also like a speculative exercise. It invites us to think about the whole thing by imagining what like this big billion dollar institution. Now, I don't have 30 billion dollars and I

don't know if any of you do but probably like the right answer like the realistic answer the realistic policy choice would be somewhere in the middle or somewhere closer to building on the efforts that you've both explained. So, yeah I mean I would say like yeah exactly like but for me this isn't maybe about like trying to tackle like those small problems but looking at it more

what what does this mean for society and and using it as a way to investigate all those problems with with hardware and making it more visible at least I think. Please round of applause for Tara.