Y Combinator Sees Hard‑Tech Startup Surge Fueled by AI

Y Combinator (YC) has observed significant shifts in the startup landscape over the past 12-18 months, particularly in the types of companies being funded and their growth trajectories. These changes are largely driven by advancements in AI and evolving macroeconomic trends.

Resurgence of Hard Tech

One of the most striking trends is the dramatic increase in "hard tech" companies within YC batches. The percentage of hard tech companies has jumped from 8% to 20%. Hard tech refers to companies that deal with physical atoms rather than just bits, encompassing areas like robotics, industrial manufacturing, defense, semiconductor stacks, photonics, and power infrastructure.

Key areas within hard tech seeing substantial growth include:

• Robotics: Increased from 1% to 6-7% of the batch.
• Industrial Manufacturing: Grew from 4% to 10%, driven by a trend of bringing manufacturing back to the US.
• Defense: Rose from 1.5% to 5%, with a new generation of founders building solutions for modern warfare.
• Compute Infrastructure: Companies building semiconductor stacks and photonics increased from 1% to nearly 4%, addressing the massive compute needs of AI.
• Power Infrastructure: Grew from 1% to almost 3%, supporting the energy demands of new data centers.

This shift is partly attributed to:

• More Technical Founders: The current summer batch shows that one in six founders holds a PhD, a significant increase from historical numbers. This expertise is crucial for complex hard tech fields like silicon photonics.
• AI Acceleration: AI, particularly code generation (codegen), is making hard tech development faster and more accessible. Previously, top-tier software engineers were a limiting factor for full-stack hardware development. Now, AI can automate much of the software component, reducing the need for massive engineering teams and changing the economics of hard tech.
• Macro Trends:
  • Space Exploration: The success of companies like SpaceX has inspired a new generation of founders to build in the space industry, covering various aspects from satellite internet (e.g., Exosat) to power solutions for satellites (e.g., Beyond Reach Labs).
  • Defense Innovation: A new generation of founders, influenced by global conflicts, is building advanced defense technologies. Examples include Icarus, developing solar-powered U2 spy planes for surveillance and communications, and Nine Mothers, creating anti-drone defense systems. These startups are leveraging AI and new building methods to create solutions that traditional defense contractors often cannot, and are finding success with the Department of Defense.
  • Dual-Use Technologies: Many defense-related startups are "dual-use," selling to both government and private sectors. This includes companies like Knox Metals, which is rebuilding America's metal manufacturing supply chain in places like Detroit, addressing a critical need for materials in defense tech.
  • Compute Demand: The skyrocketing demand for AI compute has made data centers a heavy physical atoms process. The cost of GPUs, like the Nvidia A100, is appreciating due to high demand and limited supply. This has led to a surge in startups building data centers, from construction and planning software to power solutions and alternative silicon. Companies like Lamb Labs are developing new processors, and Bot is building custom hardware architectures using ternary representation for models, recognizing that LLMs don't always need full floating-point precision. Dipole Labs is innovating with fully optical switches for data centers to overcome bottlenecks in electronic switches.
  • Robotics Revolution: There's a strong belief that a "ChatGPT moment" for robotics is imminent. Many companies are building the stack around robotics, from vertical-specific applications to infrastructure and data collection for new robotics labs. This includes companies like Boost Robotic, which builds robots for data centers, and others focused on collecting egocentric data for training robotic foundation models.

Accelerated Growth and New Software Paradigms

YC companies are experiencing unprecedented growth rates. The median YC company, which typically starts at zero revenue, now reaches $20,000 in monthly recurring revenue (MRR) by the end of the batch, up from $8,000 previously. This acceleration is largely due to:

• Agentic Coding: The rise of agentic coding, particularly since Opus 4.5, allows founders to build more mature products faster. Founders can leverage AI to run numerous coding agent sessions, significantly reducing development time.
• Full-Stack, End-to-End Solutions: The percentage of YC companies offering full-stack, end-to-end solutions or automating entire tasks has grown from 10% to over 25%. These solutions, where an AI agent performs the entire job rather than just providing a point solution, are driving significant revenue growth. Examples include AI-powered insurance brokers or clinical intake systems.
• Increased Product Value: Products that automate entire jobs are inherently more valuable than those that merely track or assist with tasks. Companies are willing to pay more for solutions that deliver complete automation. For instance, Juicebox, an AI recruiting tool, started as an LLM-powered people search but has evolved to an agent product that not only searches but also contacts and schedules interviews, significantly increasing its per-account revenue. Recruiters are embracing these agents as they free them to focus on more skilled and interesting aspects of their job, like culture fit.
• Rapid Revenue Milestones: Some companies are now achieving seven-figure revenues within the three-month batch period, a milestone that previously took 18 months or more.

Data and RL Environments for AI Labs

A stealthy but rapidly growing category of companies is selling data and reinforcement learning (RL) environments to AI labs.

• Significant Market: This market, which was niche a few years ago, has exploded. YC has funded over a dozen companies in the last two years that are each making over $10 million annually, and in many cases, hundreds of millions, by selling data or RL environments to labs. These companies are often only a couple of years old.
• Examples: Notable companies in this space include Afterthought and DataCurve.
• Driving Factor: Data is a critical component of the AI scaling law. Without high-quality data, models cannot improve. RL environments, often customized for specific use cases like finance, are also crucial for training large models.
• Robotics Data: Labs are also heavily investing in data for robotics, particularly egocentric data and tele-optic tasks, to help AI interact with the physical world. Companies like Practis Robotics and Deep Reach are collecting data from industrial operations globally.

The Rise of the Experienced and Solo Founder

YC has observed a significant shift towards more experienced and solo founders.

• Experienced Founders: Many successful founders are now in their late 30s, 40s, or even 50s. These experienced founders, who have "been around the block" and possess strong opinions and taste, are uniquely positioned to identify and build what people want. Their experience in managing teams, for example, translates well to managing coding agents.
• Solo Founders: The percentage of solo founders accepted into YC has jumped from 5% to 18-19%, the highest spike ever. This is attributed to AI making it easier for individuals to build and launch products without needing a co-founder for every skill set. While co-founders still increase the chances of success, the barrier to entry for solo founders has significantly lowered. Many successful solo founders, like those behind Instacart and Coinbase, eventually bring on co-founders as their companies gain traction.

Advice for Aspiring Founders

The current era, marked by rapid AI advancements, presents an exciting opportunity for builders. The advice for aspiring founders is to "just start prompting." With tools like GPT-6, AI agents are becoming increasingly capable, able to resolve bugs and tackle complex tasks that were challenging just a month ago. This period of rapid AI progress is expected to continue for the foreseeable future, making it an opportune time to build.