Sphere Semi raises $20M to build AI-designed analog chips for defense and data centers

a vacation home, but better maybe you book a wander next to the nor life and take a couple laps around the old Nurburg Ring. But we have our next guest, Stephen. Welcome to the stream. How are you doing, Stephen? Good to see you. Good afternoon. How are you? Uh, give us the news. What's the fundraising announcement?

Let's ring the gong at the start of the segment, then we'll get into the details. Yes. So, big announcement. Um, we have, uh, just, uh, raised $12 million in our, uh, thank you for the gong. Um, uh, we've just raised $12 million, uh, led by Acme Capital and, uh, with a co-lead with Future Ventures.

And this adds on top of the already $8 million we've raised from Construct and Abstract and Generational Partners and Village Global and XFund. So that brings us to about $20 million raised and we're very excited for the for you know after this financing round for the firepower gets us to move forward.

Talk about the state of the business. Talk about the semiconductors uh and try and put it in in terms like relative to what else is going in the market. people have a high level understanding of like you know GPUs and CPUs but what are you doing specifically in semiconductors? Yeah. So that's a good way to put it.

CPUs, GPUs, FPGAAS, these are the things that people usually think about when they think about semiconductors. Uh they they tend to be the things that come top of mind and they are indeed twothirds of the of the chip industry. So the trillion dollar chip industry, they're they're twothirds of that.

But there's this other one-third which is uh analog and analog tends to be um overlooked but it is unbelievably important.

Um you have uh the way we are communicating today uh and the way you are you know beaming out your your your show to everyone who is watching that this is all done over you know analog digital signals processors and many many many many different you know bouncing signals across the planet and and I think it tends to be overlooked but it's unbelievably important it's unbelievably important in military warfare it's unbelievably important in communications and that's where we've really focused Now big picture sphere is a is a is a chip and product company.

Um the difference between us and everyone else on the planet is that um we at the base of what we do have an AI engine that allows us to from concept all the way to fabrication produce uh you know produce chips that are entirely AI designed. Um, philosophically we we think that chip design by humans is coming to an end.

It already has mostly ended in the world of digital. So CPUs and GPUs and FPGAs, they're not designed the layouts of those chips are not designed by hand. Um, you uh you actually have TSMC basically do the layout for all the logic gates and things like that for you. Um, but in analog it's still done by hand.

And we think that the the thing we are conquering today is we're putting that to an end and we are in doing so building a product company around that where everything is designed by AI which so is it correct to characterize you as a as a fabistless chip design company.

I come to you with a something that can be done in math digital signal processing. I have some stream of information that's coming in. I need to transform that with math. You're gonna use AI to design a chip and then you're gonna call an actual fab to go and make those chips and then you send me the completed chips.

Yeah. So, uh just a bit about what we do today.

We uh found an interesting niche in uh the defense sector as a good starting point where we companies like like Ander actually specifically Andrew and some others come to us and they say hey I have an electronic warfare system or I have a sigant system and I need uh an analog front end for it right I need a bunch of chips at the analog front end and uh these are you know pretty critical they come to us they give us a custom spec u and then we go ahead and we go on to our fab um you know and we say hey we're you know this is the design our AI came up with actually hundreds of possible designs and these are their performance characteristics etc etc we work between the customer and the fab and then we fab it out we package it we ship it off um we're actually you know actually the big thing we're you know we we've done over the last you know year and a half since starting the company is is is build up this capability around these small RF components and we're actually going to start uh building a joint venture with one of the big defense primes.

Um one of the big four, I guess, uh to allow what? Let's go. Yeah, we Let's give it up for the defense primes. I know I know we like Ander here, but they don't get enough love. They don't get enough love. Um I uh I I told the an executive at Northrup about the B2 bomber meme. Yeah. Where it feed me the three. Oh yeah.

Have they seen it? He really he thought it was really funny. Maybe good, maybe bad. I don't know. Like, well, it was designed for that. Yeah, that's actually intended purpose. Um, yeah. So, so, so, so what is difficult about scaling the business now? You're raising more money.

Uh, it doesn't sound like you're deploying that capital to build a fab. Are you just hiring more software engineers, AI scientists to build more efficient AI systems that can actually design better chips? Even are we beyond superhuman capabilities?

But there's still a frontier where we can continue to advance like what what is the longer term goal?

So we have this business that we've we we've really started that we're going to build a joint venture around in um in these these I guess these RF components which are it's it's a bit of a niche business and the few things we're going to be doing with this capital is uh one actually you know investing in expanding out that joint venture etc.

There's going to be some capital investment from us. We we do pay for fabrication, things like that. But also, we're going to be moving into what you might think of as the more high value area of the analog world, which are mixed signal chips.

Um, Broadcom and Marll have made a lot of money over the last uh, you know, since the AI boom started. And a lot of that is because you need to communicate between two GPUs these mixed signal components that Broadcom and and and Marll and all these guys make. Um, you know, and that's what we're going after next.

these mixed signal components or IP blocks and that's that's the next area we need to conquer. Um that's going to take quite a bit of hiring.

Um it's going to you know whereas our fabrication for the defense stuff was in uh more reasonably you know less expensive nodes some of the stuff is going to be in much lower nodes. So any fabrication we do is going to be much more expensive.

Um but in general that's the next area we want to conquer this mixed signal uh area which is the you know the core of what data centers are really using when they think about analog. Um it's a lot of hiring at some fabrication might be some test equipment.

Um though um we're not sure about that whether we want to borrow it or buy it. Uh and you know we're hiring um software engineers, AI engineers, chip designers, a lot of chip designers. Um and then obviously I need some I need you know some business operations people to really uh come and help deals get dinners.

Well, we are rooting for you. Thank you so much for coming on the show. Uh congrats on the round and we will talk to you soon. Have a good rest of your day. Thank you. All right. Byebye. Cheers Stephen. Up next we have robots in space. But first a startup idea. bulletproof silk chararmmuse American street wear.

Are you familiar with this at all? You can silk that's tough. They modified silk worms to make a material that's stronger than Kevlar. You can make a whole jacket out of it, which I think should be bulletproof. Maybe. I have no idea.

I feel like I've been hearing about this like make a bulletproof vest that just looks like normal clothes for like decades and no one's ever figured it out, but I have no idea if it's actually