Deep Tech Takes Center Stage: Inside the Buzziest Startups of the Latest Y Combinator Demo Day

The latest Y Combinator (YC) Demo Day concluded on Thursday, showcasing a cohort that felt distinct from its predecessors. While every batch introduces a fresh wave of ambitious founders, the startups presenting this week skewed far more toward "deep tech" and hard science than in past years.

As it does every quarter, TechCrunch surveyed early-stage venture capitalists to identify the standout startups of the batch—highlighting both top picks and the deals dominating investor conversations. Across the board, VCs agreed that the technology on display felt "like science fiction." Yet, despite the wild, boundary-pushing ideas, the overarching consensus was that valuations remained refreshingly grounded compared to the hyper-inflated cohorts of recent years.

Main Facts: A Shift Toward Hard Science and Infrastructure

The latest Y Combinator batch is defined by a heavy emphasis on infrastructural bottlenecks, energy scarcity, advanced robotics, and heavy industry. Rather than focusing purely on software-as-a-service (SaaS) or generic consumer applications, these early-stage companies are targeting massive physical and computational constraints.

Key themes dominating the batch include:

• The AI Energy Crisis: Innovations tackling the massive power and cooling limitations of AI data centers.
• Next-Gen Robotics: Companies shifting away from traditional data collection to affordable home hardware, heavy-duty construction automation, and natural language control layers.
• Defense and Deep Science: Disruptive approaches to aerospace, defense manufacturing, and even biological computing.

Chronology: From Concept to Demo Day Spotlight

The journey for these startups culminated on Thursday, but the timeline leading up to their public debut highlights the rapid acceleration of deep tech in the venture ecosystem:

• Months Prior to Demo Day: Founders refined their pitches, built early prototypes—ranging from ocean-faring barge designs to neural cell cultures—and secured early letters of intent or initial customer traction.
• The Lead-Up: Rumors circulated among early-stage VCs regarding the unusually high concentration of hard-tech startups in the upcoming YC batch.
• Demo Day (Thursday): Startups officially presented to a global audience of investors, journalists, and industry leaders.
• Post-Demo Day (Current Phase): VCs scramble to finalize term sheets, while high-performing startups leverage their early traction to command significant market attention.

Supporting Data: The Buzziest Startups of the Batch

The following startups were flagged by at least two early-stage investors as the most compelling companies in the batch, categorized by sector and innovation.

Energy and Data Center Infrastructure

1. Atomarine: Floating Nuclear Data Centers

• The Problem: Power is in increasingly short supply, and local communities frequently oppose the construction of land-based data centers.
• The Solution: Co-founded by an MIT computer science and naval engineer alongside an MIT PhD in nuclear engineering, Atomarine is building nuclear-powered data centers that float at sea. Seawater provides near-free cooling for the intense compute loads.
• Traction & Valuation: The startup plans to launch a gas-powered pilot by 2028 before transitioning to floating nuclear power ships in 2032. Atomarine claims it has already secured over $4 billion in customer interest through letters of intent, making it one of the highest-valued startups in the batch.

2. Dipole Labs: Optical Networking for AI

• The Problem: GPU clusters waste valuable compute time simply waiting for data to move between chips. In traditional networking layers, data constantly converts from light to electricity and back, burning massive amounts of power and generating intense heat.
• The Solution: Dipole Labs has built an optical switch that bypasses conversion entirely, keeping data as light and routing it directly where it needs to go.

3. Lamb Labs: Custom Inference Chips

• The Problem: Traditional AI chips burn excessive energy during inference by constantly fetching model weights from memory.
• The Solution: Co-founded by an Imperial College London AI PhD and an Oxford theoretical physicist, Lamb Labs creates "Model Processing Units" (MPUs) by hardcoding AI model weights directly into silicon, completely eliminating memory-bandwidth bottlenecks.

Defense and Aerospace

4. Isengard Industries: Locally Produced Jet Drones

• The Problem: Traditional defense prime contractors charge exorbitant prices for U.S.-built military hardware, slowing down deployment and scaling.
• The Solution: Isengard aims to mass-produce jet-powered attack and counter-drones directly within allied countries at a fraction of standard costs. 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 itself.

Robotics and Automation

5. Waddle Labs: "Claude Code for Robotics"

• The Problem: The industry is waiting for a "ChatGPT moment" for robotics, but training foundation models on raw video or human teleoperation data remains difficult and fragmented.
• The Solution: Founded by Harvard graduates, Waddle Labs uses a layer of LLM agents to write code and control robots directly. Developers can plug any hardware into Waddle’s API, give natural language instructions, and have AI agents autonomously generate executable control code in about 20 minutes.

6. Nori Robotics: Affordable Home Humanoids

• The Problem: While humanoid robots exist, high price tags—such as Figure’s Neo priced around $20,000—keep them out of everyday consumer reach.
• The Solution: Launched just six weeks prior to Demo Day, Nori has already generated nearly half a million dollars in sales. Its humanoid robot, designed to help clean and fold clothes, is priced at an accessible $1,600 and can be operated via a laptop app.

7. Cosmic Robotics: Heavy-Duty Automation for Earth and Mars

• The Vision: The founders aim to eventually help build a city on Mars. The immediate stepping stone is building rugged, heavy-duty robotic technology for use on Earth.
• Traction: Cosmic Robotics reports that its technology is already installing solar panels across the U.S. and boasts $25 million in contracts through 2027. It hopes to begin an exploratory mission by 2028, racing alongside SpaceX’s timeline.

8. Praxis AI: Real-World Training Data for Robots

• The Problem: Robots require massive amounts of diverse, real-world data to successfully navigate human environments.
• The Solution: Praxis partners with businesses to capture video and data of humans performing daily work, converting it into training material for robotics companies. The startup is already working with publicly traded companies across more than 150 different environments.

Frontier Science

9. Parasma: Biological Computing

• The Problem: Traditional silicon hardware struggles to keep up with the escalating power and energy demands of massive AI models.
• The Solution: Parasma is exploring an unconventional alternative: training human brain cells to one day power compute workloads directly, offering a biological pathway to energy-efficient processing.

Official Responses and Market Perspectives

Venture capitalists observing the event noted a distinct maturation in founder ambitions. While the macro environment over the last few years forced startups to focus on immediate profitability and lean operations, this cohort demonstrates that founders are once again willing to tackle generational, capital-intensive infrastructure challenges.

"The technology feels like science fiction," noted one participating VC, echoing a sentiment shared widely across the investor floor. However, seasoned investors also expressed relief that valuations remained disciplined, avoiding the runaway multiples that characterized previous tech booms.

Implications: What This Means for the Future of Tech

The heavy lean toward deep tech in this YC batch signals a broader realignment within the venture capital ecosystem.

1. Infrastructure as the Ultimate Bottleneck: As generative AI continues to scale, the primary constraints are no longer just algorithmic—they are physical. Solutions addressing power generation (Atomarine), data transmission (Dipole Labs), and chip architecture (Lamb Labs) are poised to capture immense market value.
2. The Democratization of Robotics: With companies like Nori lowering the price barrier for humanoid robots to $1,600 and Waddle Labs simplifying software development via natural language APIs, the timeline for consumer and industrial robotics integration is accelerating rapidly.
3. A Return to Pragmatic Valuations: The discipline observed in YC batch valuations suggests that investors and founders alike have adopted a more sustainable approach to early-stage financing, prioritizing long-term technical defensibility over short-term hype.