Operator: [Operator Instructions]. Your next question comes from Mark Lipacis, Jefferies.
Mark Lipacis: I just want to make sure I understood. I had the same question as the previous caller. Was it — was it 50% of the licenses over the last year were enabling AI? Is that — was that the right —
Karel Janac: This quarter.
Nicholas Hawkins: This quarter, yes, Q3.
Mark Lipacis: And of all the kind of AI design wins you have, I think, Charlie, you started to answer this, but I don’t know if you qualified it. Are — I appreciate that at the end of the day, this is going to come to the endpoints. But are the design wins that you have, are they for endpoint AI applications? Or are these mostly data centers? I’m just trying to understand like what you saw in the last quarter, what you’ve seen recently, to what extent is the AI workload or application manifesting in the big data center chips versus the maybe the smaller chips that would go into IoT devices or cars or something closer to the endpoint? Thank you.
Karel Janac: Yes. The early generative designs that we’ve seen are data center designs. There’s a lot of activity in the hyperscaler community for those designs, so those are going into the data center. The NeuReality one that we are allowed to talk about is going into data center, but there’s others. Those would be high-priced, low-volume chips. But we think that Generative AI is going to expand to the edge and it’s going to expand to the endpoints, and we’re gaining valuable experience working with our customers to be able to provide high-performance system IP that allow people to move incredible amounts of data in these designs. And one of the things that we did was we created a kind of an XL package for — because some of these generative AI things need buses or NoCs that are 20-bit, 2,048-bit wide because of the amount of data that has to be moved.
And so that again provides us the ability to slightly increase the ASP for these kinds of very demanding applications. But generative AI is evolving very, very quickly right now and people are actually starting to move on from the transformer architectures into some other types of implementations, and so we’re just following that and we’re providing the data transport for those projects.
Mark Lipacis: Got you. Okay. That’s very helpful. And then and I apologize if I’m asking you to repeat this, my line dropped. I think you started to talk about how the larger customers are starting to shift to your solution from internal. And I think this has been part of the Arteris story longer term anyway. But can you play that back? Is there — was there something about the current macro environment that gives you higher conviction or you’re seeing an acceleration of that trend? If you could play that back again, I’d appreciate it. Thank you.
Karel Janac: Yes. I mean if you look at the investment that it takes to do the next-generation system IP, it is tens and tens of millions of dollars, right? And so, the CFOs of those companies are basically asking, okay, do we want to make that kind of an investment for the number of projects that we do or do we want to stick with what we have for existing projects? But for the new technologies, do we want to go to someone like Arteris? And that’s starting to play out. We have I think a very nice penetration of some of the companies that have had, were 100% internal before. We added one. We talked about 3 of them last quarter, and there was kind of beachhead deals. We got one of the ones that we did in Q3 was fairly large. That’s for a company that was almost 100% internal.
And so that’s one of the tailwinds, right, is that people in this macro environment are more willing to outsource the system IP to proven commercial vendors such as Arteris. And we’re seeing evidence of that.
Operator: Thank you. That concludes our question-and-answer session. I will now turn the call over to the CEO for closing comments.
Karel Janac: Well, thank you for joining us on the third quarter Arteris earnings call, and we look forward to updating you for our future quarters. Thank you very much.
Operator: Thank you. Ladies and gentlemen, this concludes your conference call for today. We thank you for participating and ask that you please disconnect your lines.
