Y Combinator Startup Podcast
00:09:02 · Harj Taggar, Jared Friedman, Diana, Garry Tan
[00:09:02] S3: you understand that? Like, kinda what... How are they growing so fast? [00:09:05] S2: So one reason is that a lot of their customers are these new defense tech startups that have sprung up and need metal to build all their all their stuff, and the existing suppliers that are these sort of, like, sleepy old businesses, mostly run by old people, just, like, can't keep up with the pace that the new defense tech startups want to build at. And it reminds me a bit of, like, when the web two dot o boom happened early in in the YC days. We would have these new startups, but then they would prefer to buy from new startups that could sort of, like, move at their speed and, like, work well with them. Like Stripe, for example, You could use a legacy credit card vendor, but, like, it's just, way better to work with Stripe. And so I feel like they're sort of becoming that for the whole defense tech ecosystem. [00:09:48] S1: Now the third trend is basically compute is a very heavy physical atoms process to get all these data centers live very quickly because a lot of the demand for AI that we've been talking has been skyrocketing. And there's a very interesting stat where GPUs from NVIDIA, let's say, like, a 100 GPU per hour is actually appreciating in cost, which is unusual. In the past, when you get a... A a one hundreds by now are sort of old. [00:10:18] S0: They're pretty old. Yeah.
00:10:18 · Garry Tan, Diana
[00:10:18] S0: They're pretty old. Yeah. [00:10:19] S1: The price is going up because there's just too much demand and not enough supply and compute. So there's a lot of startups that are now working on bringing data centers live, you have everything from the construction of the sites to the software to plan it, to actually doing the data center build out, to interest... Interesting solutions that have to do with how to power them and combination of energy, battery. So there's all these category of start ups. And even to the point of going down to the core compute silicons, there's a number of startups that are building new new silicon for an alternative to to to NVIDIA. There's this company that Tyler worked with called Lam Labs that's building new processors for compute. There's another one that I'm working on this batch called Bot that is trying to build basically new custom hardware architecture that's using ternary representation for models because what it turns out, which is a funny trend right now, if you look at all the NVIDIA architectures from a one hundreds to h one hundreds and now the b b three hundreds, each of these generations, they're actually going down in floating point precision in terms of what they were. They're going from f p thirty two sixteen eight, etcetera. And it turns out that the LLM architecture doesn't need the full precision floating [00:11:41] S0: point. Two is even somewhat usable. [00:11:44] S1: Right. So this is what bot is trying to do. And I think you have an interesting one that's doing the interconnect with with photonics.
00:11:44 · Diana, Jared Friedman
[00:11:44] S1: Right. So this is what bot is trying to do. And I think you have an interesting one that's doing the interconnect with with photonics. [00:11:51] S2: Yeah. There's a company called Dipole Labs in the current batch that is replacing the switches that are in data centers, which are essentially the, like, routing systems between different GPUs. If, GPU want... A, it wants to talk to GPU b, they talk to each other through this device. It's called the switch, and these switches right now are electronic. And so there's actually an issue, which is like the switches are not keeping up with the GPUs. The the speed of the GPUs keeps going up, and the switches are actually the bottleneck for many data centers and any different workloads. And so Dipole Labs is building the first fully optical switch where it's like all photons from GPU A all the way to GPU B. And so it will actually be much faster than the electronic switches that we use now.
00:11:51 · Jared Friedman, Diana, Garry Tan
[00:11:51] S2: Yeah. There's a company called Dipole Labs in the current batch that is replacing the switches that are in data centers, which are essentially the, like, routing systems between different GPUs. If, GPU want... A, it wants to talk to GPU b, they talk to each other through this device. It's called the switch, and these switches right now are electronic. And so there's actually an issue, which is like the switches are not keeping up with the GPUs. The the speed of the GPUs keeps going up, and the switches are actually the bottleneck for many data centers and any different workloads. And so Dipole Labs is building the first fully optical switch where it's like all photons from GPU A all the way to GPU B. And so it will actually be much faster than the electronic switches that we use now. [00:12:32] S1: Now the last one that's driving all this move to Atoms is this aspect where robotics is gonna happen. So there's a lot of companies building the stack around that, and everything from vertical robotics in specific industries to the infrastructure to deploy robots, to data selling to the new robotics labs. Because there's this moment that everyone in in the industry is feeling that we're gonna get to the chat GPT moment. It's not quite there yet, and I think we're figuring it out a new scaling law around it. So there's a lot of that. And we had Kwan here a couple episodes ago, and we're believers that's that's gonna happen. [00:13:18] S0: Mean, From robotics Pi. [00:13:20] S1: From Pi. Right? Half is AI and half is hardware.
00:13:20 · Diana, Harj Taggar, Garry Tan, Jared Friedman
[00:13:20] S1: From Pi. Right? Half is AI and half is hardware. [00:13:22] S3: I was hearing this... Reading this morning that even Astra is, like, a big leap forward for for robotics. Like, on... Like, I forget the benchmark, but there's a benchmark where, like, Fable was maybe at 10% and Asterisk showing, like, you can do, like, 60 to 70% of the tasks. [00:13:35] S0: So data, they just, you know, wake up in, you know, another couple weeks, and another breakthrough happens, and we're a little bit closer. [00:13:43] S2: It's been really cool for me to see the resurgence of heart attack because, you know, at YC, we've been funding a heart attack company since 2014. That's really when we we started. But it was, like, pretty hard to get these companies funded before. Like, I remember, like, pre pre this recent resurgence, we would, fund awesome stuff that we were super excited, like rockets and planes and chips and data centers and stuff like that. And then, like, VCs would just be like, oh, we only do b to b SaaS. And it's hard to bootstrap a company like this, so you really do need, like, downstream investors who can fund lots of... Who who who can who can fund, you know, a a full a a full capital build out. And so it's cool that, like, it seems like Silicon Valley, which historically, like, venture was set up to fund hard tech, but it, like, drifted away from it for a decade or two because it was so profitable to just fund SaaS companies. And so it's cool to have it coming back to its roots.
00:14:33 · Harj Taggar, Garry Tan
[00:14:33] S3: Yeah. On the point about like the the peer investors wanting to do hard tech again, it does seem like that. I've never seen that happen so quickly. I mean, it seemed like it happened pretty immediately when like SAS stocks were down earlier this year, Claude code was surging and that just became like the, I mean, I feel like even at demo day literally happened. I feel like that happened probably mid the winter batch at the start of this year. And it seemed by even demo day that investors were starting to be a lot more interested in hard tech companies and that's just extrapolated. It is worth knowing though on the other side, like since that a bunch of the SaaS stocks have actually recovered and are doing better than ever. Like Salesforce is like the prime example of that. I think Snowflake recently had like like two days ago had these like blowout earnings. And so it's possible that [00:15:20] S0: It all hits. [00:15:21] S3: Yeah. Maybe. I mean, that would be the green case. I mean, I still think we're seeing real stuff though. Like, clearly, the software that gets built in the future and what's valuable is different. Like, it just has to be. And so partly, it seems like what we're seeing with Salesforce is the classic, the system of record argument is actually playing out. [00:15:37] S0: The moats are intact for now.