Y Combinator’s latest Demo Day featured a cohort of startups that leaned significantly toward deep technology, a departure from previous years. Early-stage venture capitalists identified nine startups as the most notable, highlighting innovations that blend advanced science with practical applications.
One standout is Atomarine, which aims to address data center power shortages by deploying nuclear-powered data centers on barges at sea. Utilizing seawater for cooling, the company plans a gas-powered pilot by 2028, with a transition to nuclear-powered vessels by 2032. Atomarine has reportedly secured over $4 billion in customer interest, contributing to its high valuation.
Dipole Labs is developing energy-efficient optical networking hardware designed to improve data transfer speeds within AI data centers. By eliminating the traditional light-to-electricity conversion process, their optical switch reduces power consumption and heat generation, optimizing GPU cluster performance.
Isengard focuses on producing jet-powered strike and counter-drones locally within allied countries, offering a cost-effective alternative to traditional prime contractors. With $10 million in revenue and a strong valuation, the startup benefits from leadership experienced in defense and drone technology.
Lamb Labs is creating custom inference chips that hardcode AI model weights directly into silicon, addressing the high energy consumption typical of traditional AI chips. Their “Model Processing Units” aim to eliminate memory bandwidth bottlenecks, enhancing efficiency.
Another company is collecting extensive real-world video data of humans performing tasks to train robots more effectively. Working with publicly traded companies and capturing data across numerous environments, this startup supports the growing experimentation with robotic automation in various industries.
Nori Robotics recently launched a humanoid robot priced around $1,600, designed to assist with household chores like cleaning and folding clothes. Its affordability and functionality position it as a potential breakthrough in consumer robotics.
Ramp Robotics develops autonomous robots capable of heavy lifting and construction tasks. With $25 million in contracts through 2027, the company envisions its technology supporting future Mars colonization efforts, aiming for exploratory missions by 2028.
Parasma is exploring the use of human brain cells to power computing, seeking a more energy-efficient alternative to current AI hardware. This approach reflects ongoing efforts to reduce the environmental impact of AI model training and inference.
Waddle Labs offers an API that uses large language model agents to write robot control code, enabling developers to command robots in natural language. This approach could accelerate robotics programming by autonomously generating, verifying, and deploying control code within minutes.
This cohort’s focus on deep tech innovations signals a maturing startup landscape where ambitious, science-based solutions are gaining traction with investors. The grounded valuations suggest a cautious optimism about the commercial viability of these advanced technologies.