Y Combinator Demo Day's Buzziest Startups Lean Deep Tech, VCs Say

Atomarine, one of the highest-valued startups in the batch according to one VC, is building nuclear-powered data centers floating at sea. Co-founded by an MIT computer science and naval engineer and an MIT nuclear engineering PhD, it plans to place data centers on barges where seawater could provide near-free cooling. It plans a gas-powered pilot by 2028 and a transition to floating nuclear power ships in 2032. It claims more than $4 billion in customer interest through letters of intent.

Dipole Labs is building high-speed optical networking hardware for AI data centers. It says its optical switch skips the conversion of data from light to electricity and back again, allowing data to remain as light and go directly where needed. The company is targeting a problem in which GPU clusters waste compute time waiting for data to move between chips, a process that burns power and generates heat.

Isengard is developing locally producible, jet-powered strike and counter-drones. It aims to mass-produce them within allied countries at a fraction of what prime contractors charge in the U.S. Co-founded by a former Australian Army officer and a defense entrepreneur who previously scaled a Ukraine-focused drone startup to $60 million in revenue, Isengard is already generating $10 million in revenue, according to TechCrunch. Two investors said it drew one of the loftiest valuations in the YC batch.

Lamb Labs is building custom inference chips with hardcoded AI model weights. Co-founded by an Imperial College London AI PhD and an Oxford theoretical physicist, it calls the chips Model Processing Units, or MPUs, and says they eliminate memory-bandwidth bottlenecks that cause traditional AI chips to burn energy during inference.

Another startup partners with businesses to collect videos and data of humans doing work, then turns that material into training data for companies building robots. It says it is already working with publicly traded companies and has captured video data in more than 150 different environments.

Nori, launched six weeks ago, says it has almost $500,000 in sales for a humanoid robot designed to help clean and fold clothes. Users can operate the robot through a laptop app, and it is priced at around $1,600, compared with other humanoids such as Neo at around $20,000. One major question in robotics is whether an affordable at-home robot can actually stack a dishwasher and do everyday tasks.

Another startup is building autonomous robots that can lift heavy objects. Its founders want to build a city on Mars, and the company says its technology is already installing solar panels across the U.S. and that it has $25 million in contracts through 2027. It aims to automate construction for colonizing Mars.

One investor described the technology in this batch as feeling “like science fiction.” Even so, the consensus among VCs was that valuations were more grounded than in recent cohorts.

Editor's Summary

Y Combinator's latest Demo Day drew VC attention to deep tech startups, including floating nuclear data centers, optical networking for AI data centers, jet-powered drones, and custom inference chips. Investors told TechCrunch that valuations were more grounded than in recent cohorts, even as some companies reported contracts, revenue, or customer interest. The batch highlighted efforts to address AI compute power, defense production, robotics training data, and affordable home robots.