Democratizing Chip Design - Naja

Welcome to the NGI Zero podcast where we talk to the people who are building the next generation

internet.

I am Ronny Lam.

And I am Tessel Renzenbrink.

We're both from NLnet, a foundation which supports people who are working on free and

open source technologies.

For season two of this podcast we will be focusing on digital sovereignty through free

and open source technologies.

Our guest today is Christophe Alexandre.

Together with Noam Cohen they founded Kepler Tech.

We will be talking about their project Naja EDA, an open source Electronic Design Automation

tool.

Welcome Christophe and thank you for joining us.

Hello, I'm very happy to join you.

So Christophe, I was wondering, where do we meet semiconductors in our lives?

Well you know, basically pretty much everywhere.

Even if we don't often realize it.

Nowadays semiconductors or chips are the invisible force behind the modern world.

They're everywhere.

They power the devices we rely on every day like laptops, smartphones, cars, TVs, even

fridges in your kitchen.

And now of course with the rise of artificial intelligence they're more essential than

ever.

They provide the computing power that makes AI possible.

So yes, we may not think about them often, but without semiconductors today's digital

life simply wouldn't exist.

Yeah that shows very well how important it is in our lives.

So thinking about how important it is, it's interesting to see that during the COVID pandemic,

supply chain, the global supply chain was disrupted seriously and there were severe

chip shortages.

And next to that there's also the current geopolitical situation that has placed technological

sovereignty high on the agenda in Europe.

And there is a push for to increase the domestic production.

Can you explain why the semiconductor industry is so vulnerable to disruptions?

Yes, so due to its complexity the semiconductor industry is very global and very interconnected.

So it makes it sort of fragile in some sense.

And since the early 2000s most chip companies have adopted what is called the Fabless model.

So this means they focus only on designing chips while the actual manufacturing is outsourced

to very specialized foundries.

A good example which makes the news from time to time is TSMC in Taiwan.

They produce the most advanced chips like the ones that are using 3 nanometer technology.

But there is also a dependency here.

Those foundries, they rely on extremely sophisticated machines, photolithography systems, which

are almost entirely produced by just one company.

It's ASML in the Netherlands.

So there you've got a situation where just a few players in very specific locations are

responsible for the entire global chip supply.

If anything disrupts one link in that chain, be it a pandemic, a political crisis or just

a delay in shipping, it creates a ripple effect that you can feel all around the world.

So we saw this during COVID.

And now of course the geopolitical tensions are making countries, especially in Europe,

rethink how much they rely on external sources.

So that's why there's a push toward technological sovereignty and to boost domestic production.

Can you tell us something about the process of designing and manufacturing chips and specifically

about what Electronic Design Automation is?

Yes, sure.

So designing modern chips is incredibly complex.

We are talking about billions of transistors just packed in few square centimeters of silicon.

So it won't be a surprise to anybody that this design process heavily relies on advanced

software.

So that's where Electronic Design Automation or EDA comes in.

So EDA, it's a suite of tools and software that automates and manage every step of the

chip design process from the initial architecture all the way down to physical layout.

And you can think of this as a sort of digital twin of the chip.

Interestingly, this concept exists in EDA even long before this term digital twin became

popular in other industries.

Because designing a chip can cost hundreds of millions of dollars.

So accuracy here is really critical.

And any bug that makes it into the final design sent to the foundry can really kill the entire

investment.

That's why even around 55% of the time that is concentrated in a typical project in EDA

will go towards verification.

Because you need to check and recheck that everything works as intended before manufacturing

happens.

And the other point also in EDA, it is an industry that is dominated by three companies

which control about 85% of the market.

So this kind of monopoly situation tends to limit innovation in a way.

Because with so few players and little competition, the pace of progress in EDA tool is often

slower than what the industry might need.

Especially at a time when chip complexity is growing exponentially.

Okay, that's super interesting.

And also about this monopoly that you speak about.

So what then is the role of open source EDA tools in this story?

Yes, so for a long time the EDA world was not famous for its openness.

It's not like in other software sectors like AI today where open source tools and share

standards are everywhere.

EDA has traditionally been closed.

And not only were open source tools rare, but even the standards themselves weren't

always fully open or consistent.

And it's not uncommon to see different companies interpret the same standard in incompatible

ways and this slows progress and collaboration.

But now there's a push to customize hardware for specific workloads.

So you see big software players like Apple and all the GAFAM group that are investing

in their own chip designs.

And it's now not just the tech giants, but you also see startups and SMEs that are jumping

into.

So the open source ecosystem can be, we think, a very good answer to those new needs.

So the next big challenge is to evolve this open source ecosystem and to improve tool

integration, to handle more complex hierarchical design, and to be able to ultimately support

industrial grade chips at advanced nodes, which means the advanced foundry technologies.

That's where open source can really start to democratize chip design and really spark

the kind of innovation we've seen in other software fields.

Yeah, sounds good.

So can you explain what Naja is and how does it fit in the EDA field?

And why did you choose to make it open source?

Yes, so Naja is an open source EDA project we created to help build the next generation

of EDA tools.

So, especially in the area of physical synthesis, which is a critical and complex part of chip

design.

And our main goal is to provide industrial quality APIs that can serve as a foundation

for other open source tools.

So we've built Naja since day one with industry level performance in mind.

So we have the core of the APIs and the tools that are in C++ for speed and efficiency.

But we also provide a clean and easy to install Python interface.

So that means even people who aren't C++ experts, like students, researchers, or startups, can

use it and contribute.

And today you can literally install it with just a pip install command and get started.

And so why we made it open source?

So it's really because we believe openness can drive innovation.

And we think the semiconductor industry can't keep relying on closed and tools that are

in silos if we want to see real progress.

And so we want Naja to be a sort of catalyst for this change and to empower engineers and

researchers to build better tools faster and to help bring open source EDA to the same

level of maturity as other fields in software like AI, for instance.

So with your work with Naja, you obviously know a lot about the semiconductor industry.

Can you tell us something more about whether the industry welcomes innovation?

So in a way, I mean, it has to, because the semiconductor industry depends on constant

innovation to keep up with the technological progress.

So everybody knows about Moore's Law, and now it's even if it's slowing, but we're going

beyond, but there is this constant pressure about the evolving underlying technologies

and the complexity.

So innovation has to be there.

But at the same time, it's a very close and cautious ecosystem because designing and

manufacturing chips is incredibly complex and high risk.

So there is a long standing trust between EDA vendors, chip designers and foundries,

and this plays a big role in this industry.

So newcomers aren't exactly always welcomed with open arms.

You don't get in just by being innovative.

You need to bring solid proven solutions to very serious problems.

That said, we're starting to see shifts because the rise of software driven companies,

startups and SMEs entering the space is beginning to challenge the status quo.

So while change is slow, there is movement and for us, it's promising.

So how then can an open source EDA tools like Naja contribute to a more resilient

semiconductor industry?

Yes, so open source EDA can be a game changer when it comes to building a more resilient

and accessible chip ecosystem.

So first, it lowers the barrier to entry for startups, researchers that can prototype

chip design or tool developments without needing millions in licenses.

So this is already huge for innovation.

But it also enables more specialized tools tailored to new types of hardware instead

of relying on one size fits all solution from the big vendors.

But also it's not just about startups or SMEs because to build a long term pipeline of talents,

we need to empower students and researchers too.

And here open source tools gives universities and academia in general full access to the entire

chip design flow, letting students actually build and test real design.

There is a great example of this, which is a company or a project that is called Tiny Tapeout.

Tiny Tapeout uses open tools to teach newcomers how to design chips.

And it's really from idea to silicon.

And that kind of hands on experience is exactly what the industry needs to grow stronger and more

self reliant.

Yeah, it's really interesting.

You explained really well how open source can help make changes in chip production.

And that brings me to the question about the chips act.

So to increase its technological sovereignty, Europe wants to boost its domestic chip

production from 10% to 20% in 2030.

Can you tell us how this works?

So the EU chips act has been launched in 2023.

And it aims to double Europe's share of global chip production from 10% to 20% by 2030.

So it's a very important step in the development of the chip industry.

And it's also a very important step in the development of the global chip industry.

So the EU chips act is focused on increasing the chip production from 10% to 20% by 2030.

And so it focuses on three main areas.

Investing in research and skills, which is skills is a big subject there.

And also boosting manufacturing by building advanced fabs.

And of course, improving the supply chain resilience to avoid the kind of disruptions

we've seen during the COVID period.

And in this supply chain resilience, EDA plays a key role.

So I think it's really a much needed move.

But here, ambition alone isn't enough.

So the act needs a clear strategic roadmap.

But it must go beyond just factories.

To truly compete globally, we also need to invest in the full ecosystem.

And especially open source EDA to empower innovation from the ground up.

Yeah, so you mean not only just physical manufacturing, but also the skills to actually

design the chips.

And for that, you need the EDA tools.

Yeah, the tools, yes.

And also the skills, meaning engineers, that are trained,

students.

So yeah, a full ecosystem to be able to have a full chain that is addressed.

And because much has been closed sourced, a lot of people couldn't develop these skills.

And one of the reasons you want these open source EDA tools is that it's a lower barrier

for people to get these tools into their skills.

Yes, exactly.

It's about people that can be trained on tools.

And it's also tool builders that will have a lower entry barrier to start building their

own contribution, their part of the tools on top of something that is open source.

And in EDA specifically, it's really not the case until recently.

You have to build a full stack from ground to be able to start having a tool that is

useful for users.

So how can open source EDA tools contribute to Europe's push for more independence in

chip production?

And is there enough attention of policymakers for the potential of open source EDA tools?

Yes, so it's true that reaching 20% of domestic chip production by 2030 is an extremely ambitious

goal. And many now believe it may not be achievable.

There's some articles in newspapers that say it's going to be very difficult.

So it doesn't mean we should give up on it.

And of course, recent global events have made it even more clear that some technological

dependence matters.

So one strategic way to keep m strategic way oving towards that goal, we think it's by embracing open

source EDA tools, because these tools can play a key role in building a stronger and

more connected ecosystem, especially by bridging chip design with other fast moving fields

like machine learning and AI.

And they also make it easier for newcomers to get started by providing solid, accessible

foundations that don't require massive initial investment or deep legacy knowledge.

So in a way, open source can act like a sort of time machine. It can accelerate the creation of

a local semiconductor ecosystem by breaking down the traditional barriers and opening up innovation.

So trying to grow a new EDA ecosystem in Europe just by using the old closed and behind the gates

model just doesn't seem realistic anymore. And if we want to build something resilient, open

source needs to be a core part of the strategy.

And so in this matter, the policy makers have clearly identified

this item. And there is in the CHIPS Act, a dedicated call that there's consortiums

implying all the open source European community that have created consortiums to apply for this

call that is directly identifying open source EDA as a key component of this overall strategy.

That's really positive to hear that there's attention for that.

Yes.

So if the open source EDA to ecosystem is to help Europe become more technologically independent,

what kind of support does it need?

Yes. So if we really want open source EDA tools to contribute to Europe's technological sovereignty,

we need a long-term commitment from the EU and its member states.

And this is because developing EDA tools and developing a full flow isn't something that

can be rushed. The open source community needs time and stable financial support

to build up the tools and infrastructure properly.

But also beyond funding, we think that the most critical support needs also to come within the

European semiconductor industry itself. We need the local companies, startups,

and SMEs, but also the larger players to actively invest in using these emerging tools.

That means they need to dedicate both time and money

but also something equally valuable for developing such complex tools, which means direct feedback.

Because no matter how much effort goes into developing open source tools,

they won't reach industrial standards unless you have real users that are testing them,

pushing their limits, and helping improve them during the development process.

So it's really about creating a healthy, sustainable ecosystem that will be connected.

So developers, researchers, and industrial users are working side by side.

Of course, this won't happen overnight, but if we invest now,

we think we can build a kind of serious and production ready tools

that Europe needs for long-term independence in chip design.

Sounds good. Can you give us some insights into the open source EDA ecosystem?

Is it large? Is it active? Is it well organized?

The open source EDA ecosystem is still relatively small compared to other open source domains,

but it's nowadays very active. It's gaining structure and there is a momentum.

There is really a sense of purpose.

in this community happening right now.

You see regular communication through conferences,

meetings, and a lot of community calls.

Of course, initiatives like the EU Chips Act is helping to push for more coordination and define

long-term visions.

So what is exciting in this ecosystem is that it now spans a wide range of tools,

so from validation, synthesis, simulation, and you have now also analog and mixed signal design.

It's really not just isolated tools anymore. There's really a growing effort to build

full design flows and connect the pieces together.

As I said before, we are seeing the formation of consortiums and working groups that are

focused on sharing knowledge, avoiding duplication, and building towards industrial grade capabilities

together.

That sounds so good. There's so much that sounds like there's a lot of energy and a lot of

goodwill to get there together. That's really nice to hear.

So one final question would then be what has been the role of NGI Zero funding for the open

source EDA ecosystem?

Yeah, so NGI Zero funding has played a massive role in shaping the open source EDA ecosystem.

As we know it today, it has supported the growth of key existing projects.

So we can name Yosys, for instance, or Coriolis. So Yosys is a very important

both European projects and it also helped launch completely new projects. So like our own project,

Naja. And this kind of support has been really essential.

The need of open source EDA ecosystem is now widely recognized both in academic circles and

in the broader tech economy. But we wouldn't be where we are today without that early and

sustained investment from NGI Zero. And in the case of Naja specifically,

I can say it very clearly, it simply wouldn't exist without that funding.

It gave us the time, the stability, and the trust to turn our ideas into working tools and APIs.

And to do so in a way that contributes to the wider community.

So yeah, we can say it thanks a lot to NGI Zero help here.

Yeah, good to hear. But maybe one follow up question. You mentioned academia.

Do you see adoption already within academia? Are they using it in lectures?

So yes, open source EDA, I think there is an adoption movement that is already quite happening

in academia. And that is already entering the SME space. And so yeah, we see a direction that

if we follow up this direction, it could really be adopted by the wider industry.

So there are tools that are, let's say, really reaching the standard level today. So there is

the tool that is called Yosys, which is to summarize going from the human description,

it's called a synthesizer. So it's going to the human description to an interconnection of

standard gates. Then you have the place and route tools. So you have the most famous one,

which is OpenRoad. It's coming from the US. It's also widely used. And you have the equivalent

that is also used, which is Coriolis, which is a project based in France. You also have a project,

which is a physical design editor, which is called K-Layout. And it's also, yes, not only

used in academia, it's also used by industrial grade users. And one also very important point

here is that you have the full flow connection. If you see, for instance, the IHP from Germany,

IHP is a foundry that provides open source PDK, which is an open source technology.

And this means that with low investment, with all the tools I cited before, and also this open

source foundry, you can go from human description to the real layout by using now this full flow,

which covers the entire process. Great. Thank you. Yeah. And I also wanted to come back to what you

said about that the funding of NGi Zero helped the OpenEDA ecosystem, because this is exactly

the idea of the next generation internet, that you help fund stuff from the bottom up so that

things can start just by the enthusiasm and wish of the people that have the great ideas.

And that you water the ground a bit, but then it will flower really quickly and good,

because you just provide the opportunity to the people with the ideas. So I'm really happy to hear

that it really worked like that for the open

source EDA ecosystem.

So on that note, Christophe, thank you so much for this super interesting story about

definitely not only Naja, but the entire EDA ecosystem and telling us about how the

semiconductor industry works. I have really learned a lot today. So thank you very much.

Yeah. Thanks a lot for having me in this podcast. It was a super interesting questions and

I hope my answers were clear enough. It's a technical field, so it's not always easy to

explain to a broader assembly like this, but yeah, it's super cool.

No, it was very clear what you have to say. So thank you. Thank you very much.