Y Combinator’s Demo Day just wrapped up, and the startups that came out of the Winter 2026 batch are absolutely unhinged in the best way possible. We’re talking lunar hotels, AI-powered cattle herding drones, and one company literally trying to rewrite history with a video game. Investors were clearly feeling the vibes too, valuations this quarter are sitting around $30 million on average, roughly double what seed-stage startups typically command.

We reached out to a bunch of venture capital investors to find out which startups had them most excited, and the consensus is clear: the founders in this batch are not playing it safe. A company had to get flagged as a standout by at least two different VC investors to make the cut, and some of these are already crushing it financially before even getting out of the accelerator.

Beyond Reach Labs is tackling satellite solar arrays that are basically the size of a dining table when they launch, but unfold to the size of a football field once they hit orbit. The startup claims it can boost available power tenfold while cutting costs by 88 percent. They’ve already got a flight lined up for 2027 and $325 million in letters of intent from major space companies.

Then there’s GRU Space, which is going absolutely cosmic with plans to build a luxury hotel on the Moon by 2032. Founder Skyler Chan, a recent Berkeley grad who previously worked at Tesla and on NASA-funded projects, claims his startup has developed a “moon factory” that can turn lunar soil into structural bricks. The company’s already secured $500 million in letters of intent and somehow got the Trump family to make a reservation.

On the security side, Hex Security is building AI agents that act as penetration testers, constantly probing for vulnerabilities in company infrastructure. They hit $1 million in run-rate revenue in just eight weeks, which apparently had VCs “fighting” to invest.

For the agriculture crowd, GrazeMate is using autonomous drones to herd and monitor cattle on massive ranches, replacing the expensive helicopter and motorbike methods ranchers currently rely on. The founder dropped out of college mid-robotics degree after growing up on a 6,000-head cattle station in Australia.

Byteport’s tackling file transfer speeds with DART, a protocol that apparently transfers large files ten times faster than TCP, and up to 1,500 times faster on reliable connections. Luel is building a data marketplace where people can submit everyday activities like ironing to train multimodal AI models, and they’re already generating nearly $2 million in annual recurring revenue.

Pax Historia created an alternative-history strategy game powered by AI that lets you ask questions like “What if Rome never fell?” with 35,000 daily users already playing. And Stilta is helping IP lawyers automate patent disputes that can cost up to $4 million each by using AI agents to search and analyze patents across databases.

What’s wild is that a couple of these startups have already raised at $100 million valuations, though to be fair, they’re bringing in run-rate revenue of at least $1 million. The real takeaway here is that the bar for what investors consider “promising” keeps getting higher, and these founders are definitely clearing it.