Yeah. Welcome everybody. This is an example of it's good to have people, you know, who actually knew how to do stuff. I walked in this morning and it was like, man, that looks fantastic. And then thanks everybody for joining us today. I have a lot of material to talk about. And I wanted to start with this. So, Tenstorrent's been around for a while, been working really hard on AI, software, hardware, chips, systems, CPUs, and we are now a shipping company. And yay.
Oh, we got to, we got to applause. Makes me wonder if would we ever do it. We're doing it. And it's really hard. Like shipping silicon networks is really hard. It's really expensive. And the AI software stack, it looks so easy. Look at a PyTorch program. 600 lines of code, couple matrix multiplies. What could be hard about it? 500 people. Four years later, holy cats is it hard. But we're getting really good at it. So, we're shipping our Galaxy server, 32 chips in a box, scalable, scale up, scale out. I don't know what that means, but apparently that's two different things. For us, it's only one thing. It's Ethernet. We hook hardware together. We ship small servers. We have two kinds. We have a loud one that goes in a data center and a quiet one that goes in your living room. There's an example upstairs of our new one. It's great. We're building PCI Express cards that go in standard enterprise servers. Today we're going to talk a lot about our IP. Our Ascalon RISC-V CPU is shipping. Our Tensix IP is shipping and we open source a ton of stuff and we're really into the open-source ecosystem. We open sourced our software stack. We support the RISC-V environment top to bottom. It's been really amazing.
We have a, we have a really simple mission. I want to make AI and silicon cheaper. Everybody draws those graphs. Every year it gets more and more expensive. The license fees goes up. The tape out costs go up. The IP cost goes up. That's not the future I want to live in. Like literally that's not where I want to be. I want to make silicon cheaper. I want to make AI cheaper. I want to make IP cheaper and easier to use. So we have an open architecture. RISC-V, I don't pay anybody a license. We own our CPU. We support open ones. We contribute to the ecosystem. We open source our Tensix processor, right? We license our Tensix processor core. The software stack is open source. The architectural document and reference model are online. If somebody wants to copy it, bless you, go ahead, please do it. We have people using our software stack for their own applications. It's really cool. But we license that CPU IP. Our tensor software is open sourced. We have low-level kernel environment, runtime environment, big rich model environment, and now we're building a great compiler on top of MLIR. Again, software is all being developed open source. You can see it.
My favorite stories is I was talking to an investor and I said, I want to give you an update on our software stack. He said, you don't have to. We're in there every day. We can, we know exactly what works and doesn't work. Our plan is published and that's an amazing difference in clarity. Like you know what you're getting. You want to change it, go ahead. You know what the progress is. There's no secret stuff. We build high performance AI which I call normal memory and Ethernet. You don't need really expensive memory and Ethernet. I don't know what's going on with Ultra Ethernet and quasi Ethernet and it's marketing, it's expensive, it's incompatible. It's wrong. So, we're going to make really straightforward hardware to build and we're starting to build chiplets. We're taping out three chiplets the end of this year, early next year. We invested a lot of time and money to build chiplets. The first step is hard, but we did it and you're going to hear about some of that today. And then once you have chiplets and an ecosystem around it, it makes it easier to build new products because you can iterate on the part you want to. You can build a new AI processor, a new CPU, a new memory subsystem, but you can reuse the other subsystems. And I think that's really important.
So you're going to hear from these four guys today. We're going to talk about our high performance RISC-V CPU, our chiplet ecosystem, our automotive and robotic strategy, and then what we see coming forward for the future of this kind of ecosystem. So, thanks a lot for joining us today. We have a lot to show you, and we're really excited about it. It's really fun to ship products, hardware, software, IP, and next year we'll start to ship chiplets. It's pretty cool. And then we want to partner with anybody who wants to do business with me, call me and we'll get together. All right. Thank you very much.
