Moving Compute at the Speed of Demand with Mike Maniscalco of QumulusAI

So Mike, thanks for being here.

Thanks, Carrie.

It's great to be here, and I'm looking forward to the conversation.

Let's talk about you.

How did you get into this industry, and how did you end up with Cumulus AI?

Yeah.

It's not an industry I ever thought I'd be in, but I'm glad I'm here, [laughs] for sure.

My background has been I've got a computer science degree, so computer science by training, software entrepreneur for most of my career, um, founding and selling my own companies along the way.

So I, I took a path to this space through another company that started in the Bitcoin mining arena, and as they diversified pretty early in the mix from powered shell for Bitcoin to powered shell for HPC or high performance compute at the time to also spinning up a HPC as a service, which ultimately turned into a GPU as a service business, I got more and more involved in the infrastructure side of things.

And to be honest with you, after that experience, which was the ride of, of a lifetime, and I'm hoping this one rivals it, I stepped out about a year and a half of the infrastructure business thinking I wanted to go back up the stack into the software realm, but the timing was interesting.

The timing coincided with the release of the mixture of experts with the early kind of chain of, chain of thought reasoning models along the side of the release of some of the more capable, more modern models.

And as I started developing some of my own solutions and developing out some of my own tools and taking a bit of a break, [laughs] to be honest-

[laughs]

...

I watched these AI tools become really good really fast, and these are the agentic AI type coding tools that you've, you've seen in the market get even better.

And I took a step back, and what I was seeing w- was like nothing else I'd ever seen.

The, the speed that it was improving was like weekly and improving really, really well, and the output was getting better.

And I realized that all that improvement was driven by more consumption of tokens.

That combination of more reasoning and the combination of that with better models all meant more tokens.

And what powers the tokens? Well, it's the infrastructure.

So I, you know, I thought about it quite a bit [laughs] and spent a lot of time evaluating my opportunities.

And at the end of the day, it just made sense to come back to the infrastructure realm because it was what was powering everything.

And thankfully, I was in Atlanta visiting Georgia Tech where, where I went to, did my undergrad and met the Cumulus AI team that same week.

And I'm glad I did because I've loved the team, I loved the, their differentiated model, and I was already thinking about this.

So I got to know them over a matter of months, and that's how I got to know and, and join Cumulus.