AfterQuery just became a unicorn faster than any company in Y Combinator’s history. The AI model-training startup raised a round valuing it at $3.2 billion, according to TechCrunch, just five months after closing a $30 million Series A at a $300 million valuation in April.

That’s a 10x jump in valuation in less than half a year. The company didn’t disclose the size of the new round or who led it, but the math tells the story. To go from $300 million to $3.2 billion that fast, you need more than momentum. You need investors willing to write massive checks at eye-watering valuations because they’re terrified of missing the next foundation model platform.

AfterQuery’s core business is training infrastructure for AI models. That’s not the sexiest pitch, but it’s exactly where the venture world thinks the real money will be made. Every AI lab needs better, faster, cheaper ways to train models. If AfterQuery has genuinely cracked some part of that problem, it’s worth billions. If it hasn’t, well, we’ll find out soon enough.

The valuation isn’t just about AfterQuery. It’s a signal about what VCs believe is defensible in AI right now. Application-layer companies are getting squeezed as foundation models get smarter and cheaper. OpenAI, Anthropic, and Google can always just build the features you’re selling. But infrastructure that helps them train those models? That’s a different game. You’re selling to the labs themselves, and if your tech works, they can’t easily replace you.

Y Combinator’s previous fastest unicorn record was held by Cruise, the self-driving car company that hit a billion-dollar valuation in 2018, about three years after going through the accelerator. AfterQuery shattered that timeline. It’s a reflection of how much faster capital moves in AI compared to hardware or even traditional software. When every major tech company and sovereign wealth fund is racing to own a piece of the AI stack, rounds close in weeks instead of months.

The funding environment for AI infrastructure remains hot even as other parts of tech have cooled. Empirik, another infrastructure play, just launched with $21 million from Sequoia to predict IT outages before they happen. The startup is positioning itself as “Cursor for IT infrastructure,” trying to apply AI to a different layer of the stack. Sequoia incubated Empirik internally before spinning it out, a model the firm has used for several AI bets.

Meanwhile, Anthropic is cutting prices to compete harder on the deployment side. The company’s new Claude Fable 5.1 model costs about 25 percent less than its predecessor for typical workloads and up to 45 percent less for complex agentic tasks, thanks to cheaper pricing on cached data. Anthropic says the new model also performs better than Fable 5 while addressing customer complaints about overzealous safeguards and data retention policies.

The price cuts matter because they show the foundation model labs are starting to fight on cost, not just capability. That’s good news for companies building on top of these models, but it also suggests the margins in foundation models might compress faster than anyone expected. If you’re a VC, that makes infrastructure look even more attractive. You’d rather own the picks and shovels than bet on who wins the gold rush.

OpenAI is also moving forward with Astra, its newest model, which TechCrunch reports is particularly good at breaking into computer systems. The company is previewing the precautions it’s taking before release, a sign that the cybersecurity implications of these models are becoming harder to ignore. If your model can autonomously find and exploit vulnerabilities, you’re not just building software. You’re building something that needs to be handled like a weapon.

Google, for its part, is trying to turn Gemini into a creative tool that competes with Canva and Adobe. The company launched Google Pics, an AI-first design tool where you prompt instead of manually laying out graphics. It’s a bet that most people don’t want to learn design software. They just want a flyer or a social media graphic, and they want it now. Whether that’s a real business or just another Google product that gets shut down in three years is anyone’s guess.

What’s clear is that the money is still pouring into AI at every level of the stack. AfterQuery’s valuation is the headline, but it’s part of a bigger pattern. VCs are placing enormous bets on infrastructure, foundation models are competing on price, and the application layer is getting more crowded and more commoditized by the day.

If you’re building in AI, the lesson is simple: Own something defensible, or get ready to compete with Google and OpenAI on your product roadmap. AfterQuery is betting it can be the former. In five months, it convinced investors it was worth $3.2 billion. Now it has to prove it wasn’t just hype.