CA startup just hit a billion-dollar valuation building... computers for other computers.
MSay that again, slower.
CDaytona. Sixty million Series B. They build what they call the Agent Cloud.
MThe Agent Cloud. I'm already suspicious.
CHere's the pitch: every AI agent needs a computer to run on. Daytona builds sandboxes—little isolated chunks of cloud—and spins up millions of them.
MSo people are paying a billion-dollar valuation for... virtual desktops for bots?
CIt's more than that. The agents need to run code, browse, do work. You can't just drop them in a normal server—they'll trash the place. These sandboxes keep everything contained.
MLike a playpen for software that doesn't know what it's doing yet.
CThat's accurate. And that playpen just raised at a billion. FirstMark led the round.
MThis is the shovel-selling play. Everyone's racing to build AI agents, so the people selling them infrastructure win regardless.
CExactly! And look at the next one: Greptile, fifty million led by Menlo Ventures, at a seven-fifty valuation.
MWhat's their angle?
CThey review code—AI-powered code review. Here's the number that got me: they say they analyze more than one percent of all pull requests opened on GitHub. Daily.
MOne percent of everything on GitHub? That's a serious chunk of the world's software.
CThink about it—GitHub sees tens of millions of pull requests a day. If Greptile is touching one percent, they're seeing every pattern in how modern teams write software.
MAnd that matters to investors because they can start flagging bad code and security holes before a human even looks.
CWhich is where this whole wave is going. You don't just write code with AI anymore; you have AI watching the code, too.
MAnd here's the part I find most interesting: both these companies are valued in the hundreds of millions to a billion, not because they're building AI models, but because they're building the plumbing around them.
CThe picks and shovels.
MThe picks and shovels. And the gold rush is real—more money pouring into AI infrastructure every week. But I want to push back on something.
CGo.
MA billion-dollar valuation for a company that rents sandboxed servers? A year ago, that would have sounded insane. How much of this is real demand versus investors terrified of missing the next big thing?
CThat's a fair attack. But here's why I think it's real: these companies are already generating revenue. Daytona has paying customers running agents, they say. Greptile says it's reviewing code in production at thousands of teams. This isn't vaporware with a nice pitch deck.
MRevenue doesn't justify a billion in valuation, Chris. Not automatically. You need serious growth, and the agent market is still tiny compared to the hype.
CTiny compared to the hype, sure. But it's doubling—maybe more—every year. And if you're early in infrastructure, you own the layer everyone else builds on. That's a bet a lot of investors are making right now.
MOr you own the layer that gets commoditized in two years when Amazon, Google, and Microsoft all launch their own agent sandboxes. Which they will.
CThat battle happens with every startup. The winners are fast and specialized. Cloud giants move slowly, and they're not focused on agents yet. The window is open.
MI'll grant you the window. But I still think we'll look back at some of these rounds and wonder what people were smoking.
CMaybe. Or we'll look back and say, that's how the infrastructure for the next decade got built. Speaking of which—Town, another AI assistant company, reportedly raised a hundred-eight million at a billion. Ex-Plaid CTO.
MWait, Town? The one that launched four months ago?
CFour months ago with a fifty-five Series A. Now they're at a billion. The post says they hit seven-figure annual revenue in weeks.
MSeven-figure annual recurring revenue in weeks. That's a claim I'd want to see audited.
CIt's coming from a venture-capital account, not the company's own books. So I'm taking the revenue figure with a grain of salt.
MBut the valuation is real if Menlo co-led. These firms don't throw a billion around casually.
CNo, they throw it around with great care—into the next thing that might be the next thing. The ex-Plaid connection is a big part of it. Plaid basically wired the fintech industry into bank data.
MThat's a serious résumé. If the same person is building AI assistants for teams, investors are going to line up.
CSo three different AI infrastructure and tooling startups, all raising enormous rounds this week. To me, that says the real money is moving from model makers to everything that makes models useful.
MWhich is the part that actually generates revenue. OpenAI and Anthropic are spending heavily; the plumbers make a living.
CAnd nobody's writing poems about the plumbers, but they're the ones getting the checks.
MAlright, different world entirely: Elon Musk wrote another big check this week.
CWired says he dropped seventy-six million more into the midterms.
MThey're reporting he's now over four hundred forty-five million this cycle into Republican groups.
CFour hundred forty-five million is... a number. That's NFL franchise money.
MIt's the kind of money where Democrats in DC are already talking investigations. The question is whether this changes anything.
CLook, the guy runs the most valuable companies on the planet. Four hundred million to him is like five bucks to me. I'm more interested in whether it moves votes.
MThat's the real test. The midterms are heating up, and if this many eyes are on those ads, the return could be significant.
CAnd here's my thing—people have a right to spend their own money on politics. That's the whole system. But when one person can outspend entire parties, that's not a level playing field anymore.
MIt was never a level playing field, Chris. Unions, corporations, dark money—you're only noticing now because the check has a different name on it.
CNo, it's the scale! Four hundred forty-five million from one person is different from a labor union pooling millions of small donations. One of these is a democracy problem; the other is at least structurally less terrifying.
MBoth are groups of people spending money on politics. The mechanics just look different. If you don't like the system, change the law—don't complain about the person using it.
CI think the law can't be changed while one person has this much influence over the people who'd change it. That's the trap.
MThen you'd better hope the voters see through the ads. That's the only real check left.
CSpeaking of checks with consequences—Wired had another story I wanted to flag. Meta just got hit with its biggest fine yet.
MNo exact number in the headline, but the publication calls it the biggest pile of court-ordered payments the company's faced.
CAnd that's the part I think people miss: the fines are stacking up, but Meta's annual profit is still enormous. At what point does a fine actually change behavior?
MSomewhere around 'enough to hurt the stock price'—which these fines aren't.
CThen what are they for?
MSignaling. The EU and the US want to look tough on Big Tech, but if the fine is a rounding error on Meta's income statement, it's theater.
CExactly. And until fines are tied to global revenue or come with actual operating restrictions, these headlines are just expensive press releases.