Y Combinator’s latest Demo Day featured a distinctive group of startups, with a notable emphasis on deep technology ventures. Early-stage investors highlighted several companies as the most promising in this batch, noting that while the ideas are ambitious, valuations remain more reasonable compared to previous cohorts.
One standout is Atomarine, which aims to deploy nuclear-powered data centers on floating barges. By leveraging seawater for cooling, the company addresses power shortages and local opposition to traditional data center construction. Atomarine plans a gas-powered pilot by 2028 and a transition to nuclear-powered vessels by 2032, already securing over $4 billion in potential customer interest.
Dipole Labs is developing energy-efficient optical networking hardware designed to improve data transfer speeds within AI data centers. Their optical switch technology eliminates the usual light-to-electricity conversion, reducing power consumption and heat generation, which is critical as GPU clusters become increasingly expensive.
Isengard focuses on producing jet-powered strike and counter-drones domestically within allied countries. With a leadership team experienced in defense and drone startups, Isengard has generated $10 million in revenue and attracted high valuations due to its cost-effective manufacturing approach.
Lamb Labs is working on custom inference chips that hardcode AI model weights directly into silicon. This approach aims to reduce the energy consumption and memory bottlenecks typical in traditional AI hardware, potentially improving efficiency for AI inference tasks.
Another startup collects real-world video data of humans performing work tasks to create training datasets for robotics applications. This data-driven approach supports companies exploring which tasks are best suited for automation versus human labor.
Nori Robotics has introduced an affordable humanoid robot priced around $1,600, designed to assist with household chores like cleaning and folding clothes. Despite being newly launched, it has already achieved nearly half a million dollars in sales, presenting a more accessible option compared to higher-priced humanoid robots.
Ramp Robotics develops autonomous machines capable of heavy lifting, with current applications including solar panel installation across the U.S. The company envisions its technology as foundational for future Mars colonization efforts, targeting exploratory missions by 2028.
Parasma is exploring the use of human brain cells to power computing, aiming to create more energy-efficient alternatives to existing AI hardware.
Waddle Labs offers an API that enables natural language commands to generate robot control code autonomously. By integrating large language models, the platform allows developers to program robots quickly without extensive manual coding, positioning itself as a potential breakthrough in robotics programming.
This cohort’s focus on cutting-edge technologies reflects a shift in startup innovation toward solving complex, real-world problems with advanced science and engineering. The attention from investors underscores the growing interest in deep tech solutions that could shape future industries.