Y Combinator’s latest Demo Day revealed a shift toward deep technology startups, attracting significant attention from early-stage venture capitalists. Unlike previous cohorts, this batch emphasized advanced scientific and engineering solutions, with valuations remaining grounded despite ambitious projects.

Among the most notable startups is Atomarine, which aims to deploy nuclear-powered data centers on barges at sea. This approach leverages seawater for cooling and addresses power shortages and community resistance to land-based data centers. Atomarine plans a gas-powered pilot by 2028 and a transition to nuclear-powered vessels by 2032, having already secured over $4 billion in customer interest.

Dipole Labs focuses on improving data center efficiency by developing optical networking hardware that eliminates the energy-intensive conversion between light and electricity. This innovation targets the bottleneck in GPU clusters, promising faster and more energy-efficient data transfer critical for AI workloads.

Isengard is producing jet-powered strike and counter-drones domestically within allied countries at a fraction of traditional costs. Founded by defense veterans, the company has generated $10 million in revenue and is recognized for its strong market potential.

Lamb Labs is advancing AI hardware by creating custom inference chips with AI model weights hardcoded into silicon. This design reduces energy consumption and overcomes memory bandwidth limitations common in current AI chips.

Another startup collects real-world video data of human work to train robots, partnering with publicly traded companies and capturing diverse environments. This data-driven approach supports the growing experimentation in task automation and human-robot collaboration.

Nori Robotics offers an affordable humanoid robot capable of household chores like cleaning and folding clothes. Priced around $1,600, it presents a more accessible option compared to higher-cost humanoids, addressing a key question in consumer robotics.

Robotics startup Ramp is developing autonomous machines capable of heavy lifting, currently installing solar panels across the U.S. With $25 million in contracts through 2027, Ramp envisions its technology aiding future Mars colonization efforts.

Parasma explores using human brain cells to power computing, aiming to create more energy-efficient alternatives to conventional AI hardware.

Waddle Labs provides an API that enables natural language commands to generate robot control code via large language models. This innovation could accelerate robotics programming by allowing developers to set up and operate robots quickly without extensive coding.

This cohort’s focus on deep tech innovations reflects growing investor interest in startups tackling complex challenges in computing infrastructure, robotics, and defense technologies.