Hi everyone. My name is Catherine Schwab. I'm an assistant managing editor here at Forbes

covering technology. And I'm so excited to be here today with Andrew Feldman, the co-founder and

CEO of Cerebrus, which is an AI chip company. It's about eight years old and is planning an IPO for later this year,

which is exciting. So Andrew, thanks for joining us. Thank you for having. Let's talk about

chips and AI, which is really hot right now. Normally the first company that comes to

mine when you talk about chips and AI is Invidia, obviously.

Can you talk about how Cerebrus' chips and kind of approach to chips is different from

NVIDIA?

Sure.

Invidia has its heritage in graphics.

They pioneered a part called the GPU, which stands for

graphics processing unit. AI work is slightly different, and we began with a clean sheet

of paper and designed a chip and a system and software just for the purpose of AI.

And as a result, we're orders of magnitude faster and we use a tiny fraction of the power

per unit compute.

Okay, so in terms of practical speed, what does that mean for if I'm, well, I know you're not powering chat GPT per se

because that's open AI. But how much faster are we talking in kind of layman strengths?

Well, if you ask, say, deep seek, which we are serving right now, if you ask deep seek a question and you

ask the same question to open AI 01 mini one of their top models we will be

about 57 times faster to get you an answer so yeah it's a lot. And if you compare head to head with one of, say,

meta's models, Lama, 70 billion per meter, which is one of the most popular models,

will be 70 times faster. So it's head and shoulders faster. The difference for the user is between a second and 24 seconds to get an answer back.

...