Mesh Optical raises $50M to build US-manufactured optical interconnects for AI GPU clusters

store, on mobile, [music] on social, on marketplaces, and now with AI agents. And without further ado, let's bring in our next guest, Travis from Nash Optical. Travis, how you doing? Hey, nice to meet you guys for having me on.

Thanks for hopping on.

Big big big day.

Huge day. We

almost happened on uh February 17th. Uh we we launched on February 17th as well.

That's true. That's true. Add it to the list. How much did you raise? Break it down for us.

Uh yeah, 50 million total across our seed in series A.

Nice.

Congratulations.

First time on the show. Before we get into mesh, uh introduce yourself. what got you into this, all that good stuff.

Yeah. Yeah. I'm Travis Brashier, one of the co-founders and CEO here at Mesh Optical. And um I worked I I I I've been working on laser since I was in high school and um with the professor at UCSC, [laughter] uh Philip Luben, and then

Oh, interesting. So, in high school you were you were basically studying uh at UCSB. You're studying the blade.

Yep.

The laser beam study the beam. Starting studying the laser. That's right. Been been one with the laser since uh I was, you know, 15. Um and then I I went to SpaceX and started working on the space laser comm system there and got to the really fortunate opportunity to design and work with a really awesome team there on on the free space laser com terminal.

Cool. Uh so talk about

living the space laser dream. [laughter] Incredible stuff.

Uh yeah. So talk about talk about the actual product. uh you know how much of this is still R&D ready for commercialization then I want to talk about the implications uh like how this product actually rolls out into the into the world.

Yeah. Yeah. So our first product is um a pluggable transceiver that is used in all of the GPU clusters. So you just mentioned Jensen um like the Nvidia GPU clusters. For every one GPU, there's like four to five of these optical interconnects that help connect all the GPUs together through the networking through making sure that they can be like a coherent cluster. Um, and so the first product we're making is what we call like a linear pluggable optic. And it deletes a really expensive and costly like like power hungry thing that uh is called a rettimer or DSP. Um, and so it allows the people who deploy compute um to save a lot of power. Um and so that first product is just simply allowing those GPUs to connect over optical um and deleting some power power hungry chip inside.

And this is within a data center or in between different data centers.

This is within the data center. Yeah. So our first Yeah. Our first product is is yeah within a data center. They connect them with fiber optic cable on either end. Um and uh yeah. And so the

And these are these are already already in market. You're selling these.

Wow.

Wow. Well, ours ours are we're standing up the production line. So, we we care a lot about colllocating the engineering talent with the manufacturing. Um, one of the key things we learn at SpaceX is to to be like right where the stuff is built. So, behind me is the manufacturing line and um we'll start testing in the next month here uh some samples and then start scaling production uh right after that. and um trying to get to you know millions of units in 2027 is like really where the high volume starts to come into play.

Wow. Uh is this the kind of thing where it's really just a it's execution risk like can we build this at scale and if we can build them there's insane uh demand and it's and it's less demand risk.

Yeah. the demand we sort of think of as infinite because you know the more intelligence we have uh [laughter] the more intelligence we have it it just like is more compute and so whenever you see people deploying more and more compute they need more and more transceivers and so um as intelligence scales with compute so does transceiver demand. Mhm. Uh $50 million in the bank. Uh where are you based in the company? How big is the team? What does the next uh 12 to 18 months look like for you?

Yeah, we we're based in Los Angeles. We're in Gardina, California. Um of lasers. Thank you. Thank you. Los Angeles needed a win. [laughter]

Yeah. Yeah. Um yeah, a little outside Elsagundo, but uh yeah, we're going to bring Gardina onto the map as the the laser hub of of the world.

Um love it. And uh yeah, so we're about 15 people now and and growing quickly and need to move to a bigger building now.

Good sign.

Are there are there like two buyers for this? Like are you going to do deals with like every data center builder, every Neocloud, just the hyperscalers? Are you going to be a supply chain partner like just to Nvidia or just to AMD? Like what does your business look like in like at scale? Yeah, we we want to sell I mean my our long-term goal here is to be able to connect all all things with optical photons and um whether that's an interconnect in a data center or an interconnect in space whether that's free space or in a fiber um but to start out to answer your question specifically like the anyone who who deploys compute uh on the ground we will try to sell to um that could be a neocloud as you mentioned it could be a hyperscaler um we have a strategy that we want to follow to kind of follow our volume, match our volume to our potential like future customers and make sure we don't start all the way at like really high volume because obviously we need to scale our line. We care a lot about getting our manufacturing done here in the United States. Um, one thing about all these transceivers that are made today, they're all made overseas. Um, they're made mostly in China and Thailand. Um, and so no one has like really stood up these high precision uh what we what they're called flip chip die bonding uh or pick and place machines. Like when we bought the most high volume uh flip chip die bonder, they asked for your Chinese social credit number on it. Um because no one buys them here.

Wow.

Uh what do you think that we'll get uh any mesh products in space in the next five years? Uh yeah, [laughter]

that's awesome. Uh 2027, you thinking what I'm thinking?

Valentine's Day gifts. Here we go for that special person.

Lasers. Lasers for lasers for Valentine's.

I Crazier things have happened. We

We're big fans of their girlfriend to

Yeah. They like They like They keep saying they want some metal, you know. So get them get him some Get him some hardware.

I I was a I was a laser cat for Halloween once.

No way. [laughter] Oh, that's cool. Yeah. uh talk about uh like what is the shape of like actually working in the factory? Is this stuff risky or does it all happen in a clean room? How much of this is like a TSMC type fab? Like what is it like today?

Yeah, the the the the clean room is here. We actually took some pan we took some panels off to get the machines in. But yeah, it's it's a pretty like we take the approach of like question the requirements and delete the part of process if we need to. And you know, we start in like a semi- dirty environment to make sure like like we we want to we don't want to go overboard, right? You don't want to just go all in on a clean room when you don't have to. Um so actually back at SpaceX, we started making the space lasers in a tent. Um and so like there's actually a tent behind me um or in front of me, sorry. And so like you you you want to not you don't want to go too all in on like it has to be any cleaner. You you do tests end of line testing and and qualification to make sure that like if it starts to impact your yield then you would implement procedures and processes to keep it cleaner. This is like a pretty good clean room but not the best. And so we'll we'll see how our yield looks with with this one. And um we'll implement stronger uh strategies if we need to. But yeah, all of our semiconductor packaging is happening here. And you know, we we get some dyes from other foundaries that are all outside of Asia. Um, and then we bring them here and package them.

Last question for me. What was the biggest lesson you learned from working at SpaceX?

I mean, just what I said is like uh question the requirements and delete the part and process. It's like it's so simple and but it's so useful and like when we're designing something like you you want to try to know why each part is there. So in our design we've deleted quite a few parts and most people don't delete that part but it ends up work. Anytime you delete a part you delete a potential failure mode. Um and you want to be like a really reliable system you need to delete as many parts or processes as possible. It also helps you like assemble things faster and um so that that was one of the big lessons I learned.

Yeah, that feels like I don't know easy to say, really hard to do in practice. Like it feels like you have to experience it. Like I've heard that a million times and like I'm sure that there's things that I could delete like even my daily workflow that I haven't figured out how to do. That's a I love that.

Yeah, I is a company where like you learn learning things the hard way in the most extreme way. Most you ship a soft ship a piece of software, it doesn't work. It's like, okay, let's patch it. You ship a rocket, it has some like dependency that you didn't think was maybe that important and it blows up and you got to uh live with that

or the whole laser mesh doesn't work. So, [laughter]

let's hope that happen. that, you know, like putting a a kid in charge of that is is really a crazy thing that SpaceX does is they hand the baton to a really young person and put the whole company like bet the company on those people and um I think that forges things like within people to be able to deliver and then the delete the part and process like helps them understand the whole system really well.

I love it.

Well, we are not far from Gardinia.

Yeah. Come by. uh come by for your next appearance. I got a feeling you'll be back on the show this year. Uh

yeah, I would love to be there. You guys are welcome to come anytime as well.

I love it. Thank you so much.

I love lasers. I would love to.

Have a great rest of your day. We'll talk soon. Let me tell you about Plaid. Plaid powers the apps you use to spend, save, borrow, and invest. Securely connecting bank accounts to move money, fight fraud, and [music] improve lending now with AI. And you saw it at the opener. Jordy, get those juggle balls ready because it's time to tell you about Turbo Puffer Serverless factory and full tech search built from first principles and object storage. Fast, 10x cheaper and extremely scalable.

Sent me these these little puffers. They're incredible.

Very fast.

They're going to be hard to keep here at the studio.

Well, without further ado, we have Evan Spiegel from Snap in the TV Ultra Dome.