Key insights
- Normal Computing secured $50M in funding led by Samsung Catalyst to scale commercial partnerships and expand its team. The company focuses on computing where data resides, aiming to improve AI energy efficiency. While not directly impacting near-term US equity prices, successful development in this area could indirectly benefit semiconductor companies like Micron and Intel in the long run.
Normal Computing CEO Faris Sbahi joins Market Domination's Josh Lipton and Yahoo Finance Tech Editor Dan Howley to discuss the software company's place in the AI infrastructure buildout, especially after receiving backing from the Samsung Catalyst Fund (005930.KS).
And Ferris, you made news this week. You announced that you raised $50 million in a round led by Samsung Catalyst. What what does that mean for your company, Ferris, and how do you put that money to work?
It's a great question. I mean, so we're calling this an accelerator round. Um, so predominantly that's going to mean scaling our commercial partnerships, uh growing the team, um, and you know, like a lot of this for us has been about the partnership. uh so embedding ourselves uh even further than we have been in the strategic ecosystem. Uh so when you look at some of the investors who are participating, um, you know, from Samsung but also uh Micron Ventures, Celesta which was founded by Li Bhutan, the the current CEO of Intel, um and also Eric Schmidt's First Spark ventures, um you kind of get a sense of, you know, the the um level of embeddedness that we're um kind of going after here.
Fair's uh Dan here. I I want to ask you what separates kind of, you know, your design process from what, you know, other companies have because you mentioned 6, you know, obviously Google's got their own 6, you know, Broadcom's obviously there. you have different companies, we have uh Grock and their uh LPU, um, obviously coming from from TSMC most likely, but you know, what is what is it that separates your chip design from others?
It's very different kind of design that we're targeting. Thank you for the question, Dan. Uh so really there are two pieces. One is that we're computing where the data lives um instead of moving it back and forth between the memory and the processor. Uh this is an area which people are calling processing and memory that's sort of emerging as a very important area for solving um the energy problem, other kinds of efficiency uh problems. The second is that, you know, we're seeing that AI workloads like video generation, um are already noisy and approximate in a sense. and so our hardware works with that noise instead of burning energy to fight it. Um so we're kind of not fighting the physics but leaning into it.
You know, the efficiencies are definitely improving, um, when we think about where conventional hardware is headed. But I think, you know, we're still seeing that as projected even with um, where the, like how quickly the hardware is improving, uh we are running up against a huge potential supply shortfall. Uh, so right now it's projected that we're going to need, if we keep going the way that we're going, um 134 gigawatts of power by 2030, um which means that we're going to have like almost a 50 million uh sorry, 50 uh gigawatt gap. Um and so kind of what we're demonstrating here is that uh there's uh an alternative solution to acquiring more energy um and finding, you know, creative ways to build nuclear power plants and otherwise, and that actually we can drive down um a couple orders of magnitude