AI Agents Will Cause a Massive CPU Shortage | Ivan Burazin
MTS
October 21, 2026
Ivan Burazin, co-founder and CEO of Daytona, discusses Daytona's pivot from developer environments to AI agent infrastructure, their recent $60 million Series B funding round at a $1 billion valuation, and why the surge in AI agents will trigger a massive CPU compute shortage.
Ivan Burazin
If for Muse to have 100 million users, it will spend 2.8 billion dollars of CPU.
SPEAKER_2
Just CPU.
Ivan Burazin
Just CPU compute. Now there's some optimization that can be done, whatever, just take it, it's very hand-wavy, but directionally correct. That's 100 million users. Meta now has 3 billion users, and that's just Meta. Now think of all the consumer apps in the world, the scale that they are at, and if they all become agents, how much compute is that? That compute doesn't exist in the world today, right? So that is directionally what we are building out, that CPU execution cloud for agents to get their job done.
SPEAKER_2
Hello, everyone. Welcome back to MTS. Today, we are joined live in the studio with Ivan Burazin. Ivan is the co-founder and CEO of Daytona, a company focused on standardizing and managing developer environments, and earlier today, they officially announced their 60 million Series B at a $1 billion valuation. Quite the numbers. Big congratulations to you. Welcome to MTS.
Ivan Burazin
Thank you so much. Thank you for having me.
SPEAKER_2
All right. Well, this is an exciting announcement. I know, fundraise days are quite the world wind. What did you guys announce today? What's the big plan? What's the big news?
Ivan Burazin
Sure. So Daytona is a infrastructure provider. So we provide computers for AI agents, popularly called sandboxes. You've probably heard a lot about sandboxes lately. There's a lot of work around to talk about sandboxes. And so, but the way that we think about the product that we're building is not necessarily purely a security product, rather a performance product. So security obviously is there, there's guardrails. Hopefully, it doesn't get out of the sandbox. We will try our best not to let the agent out of the sandbox. But if you think of agents as digital knowledge workers, the tool or the computer that we give them is essentially like the laptop you have right there. It is where it will run a browser, will use Excel, where it will write code, will run it. And can we do that for the agent at the speed and velocity it needs, and at the scale it needs to get the job done? Yeah.
SPEAKER_2
What's really interesting, I think you alluded to Sandbox, and I will say the hugging face incident was probably the first thing that got the mainstream public thinking about Sandboxes, unless you were a developer. Prior to that, you would never really think about what a Sandbox actually is. What's wrong with our current infrastructure? Why can't you just have all of our agents work on already how we use tech?
Ivan Burazin
Sure. And this is the question that we post people. So Daytona is actually three years old. And so made about a year and a half, we did pivot. So we were automating dev environments for humans, and then we switched over to serving infrastructure for agents. And this was a year and a half ago, agents as they are today were not a thing. There were a few people building it and they were looking for infrastructure and we would give them our old human infrastructure. And they're like, this sucks. And we're like, what's the difference, right? The same thing you're asking me, the same thing other people ask. And basically TLDR is everything from speed and performance and scale and security. It's actually quite different. So I'll just give you a quick example. One, historically, any of the virtual machines they can get on the hyperscalers will spin up in a few minutes, two or three minutes. Ours will spin up in 60 milliseconds. So you blink at 120, so it's like a half-year blink. That's how fast it spins up, right?
Sometimes you need that, sometimes you don't. But other things is you as a human or as a human can work on one computer at the same time. An agent can literally, if anyone has watched the Marvel movies, there's the idea of a multiverse. So every decision you make sort of branches off the entire world. And so we do that in an infrastructure layer. So you have your computer and if you have fork and road, you can actually replicate it into two computers, then four, then six, then continue until you get to your outcome. So there's a bunch of different things that are very specific that helps agents perform faster and obviously more securely.
SPEAKER_2
Yeah, I want to talk a little bit about agent infrastructure. I know it's the big hot topic about what we actually need. The thing is agents are early. I would say even the rise of personalized agents are only a few months old. Why do agents need their own infrastructure? What's wrong with current infrastructure?
Read the other 20 minutes of this episode here.
$15 gets you this episode and 99 more. Credits never expire. After checkout, one link brings you back here to read it.
Not what you expected? Full refund within 14 days.
Thousands of transcripts fetched by people building searchable podcast archives
Fetch transcripts with code
The same credits work from a script or an AI agent. The demo key returns a sample episode in full, no card needed:
curl -H "x-api-key: pt_demo" \
https://spoken.md/transcripts/1000651996090Markdown with the speakers named, for your notes, your knowledge base, or anything that makes HTTP calls.
From $0.10 per transcript. No subscription. Credits never expire. Prices exclude VAT, added at checkout for EU customers. Not what you expected? Email us within 14 days with 20 or fewer credits used and we refund the pack in full.
Using your own key:
curl -H "x-api-key: YOUR_KEY" \
https://spoken.md/transcripts/1000797440611