Why are more founders focusing on hard tech now?

Advances in AI, particularly 'codegen', are accelerating scientific research and making hardware development faster. Additionally, macro trends like successful space companies and increased focus on defense are inspiring founders to tackle physical world problems.

How has AI changed software development for startups?

AI, through tools like 'codegen' and agentic workflows, is making it possible for startups to build more mature products faster. It's shifting the focus from pure coding skills to knowing what to prompt and how to leverage AI to automate entire jobs or tasks.

Are SaaS companies still relevant in the AI era?

Yes, SaaS is transforming rather than dying. Companies that act as 'systems of record' or 'harnesses' for AI agents, and those that automate entire workflows rather than just being point solutions, are seeing significant growth and value.

What is the trend regarding solo founders vs. co-founders?

There's a notable increase in solo founders being accepted into YC, rising from 5% to nearly 19%. This is largely due to AI tools enabling individuals to handle more aspects of building a company alone, though co-founders remain valuable for scaling.

Is it still possible for new founders to succeed in the current startup environment?

Yes, the current era favors founders who know what to build and can leverage AI tools. Experienced founders are seeing a resurgence, and the accessibility of AI development tools means that even solo founders can get traction and build mature products much faster than before.

What kind of companies are experiencing rapid revenue growth?

Companies that automate entire jobs or workflows, especially those leveraging agentic coding and AI, are seeing explosive growth. This includes full-stack solutions that deliver a complete task, moving beyond older point-solution SaaS models.

What role do data and RL environments play in AI development?

Data and Reinforcement Learning (RL) environments are crucial for training advanced AI models. A significant category of stealthy startups is now generating substantial revenue by providing these essential resources to major AI labs.

Key Moments

The State of Startups in 2026

Y Combinator

Science & Technology6 min read37 min video

Sep 17, 2026|2,010 views|201|14

YC Y Combinator

Save to Pod

Want to know something specific about what's covered?

We've already dissected every moment. Ask and we will deliver (with timestamps).

Key Moments

TL;DR

Startups are shifting from software to physical products, with YC seeing a jump in 'hard tech' companies to 20%. This is fueled by AI's ability to accelerate research and development, making complex hardware more accessible to smaller, more experienced teams.

Key Insights

The proportion of YC companies in 'hard tech' (physical products) has grown from 8% to 20% in the last 12 months.

The median monthly revenue for YC companies at the end of a batch has increased from $8K to $20K.

One in six founders in the current YC summer batch has a PhD, indicating a rise in highly technical founders.

AI is reducing the software engineering bottleneck in hardware development, allowing smaller teams to achieve more.

Companies selling data or RL environments to AI labs have become a significant category, with over a dozen YC companies each making more than $10 million annually.

The percentage of YC companies focused on full-stack, end-to-end tasks (where an agent does the job) has increased from 10% to over 25%.

The significant rise of 'hard tech' startups

Y Combinator has observed a striking shift in the startup ecosystem, with the proportion of companies focusing on 'hard tech'—companies that deal with physical atoms rather than just digital bits—growing significantly. This category has expanded from 8% to 20% of accepted companies in the past 18-12 months. This includes a surge in robotics (from 1% to 6-7%), industrial manufacturing (4% to 10%), defense (1.5% to 5%), and the semiconductor/photonics stack (1% to nearly 4%), as well as power infrastructure (1% to nearly 3%). These increases suggest a revitalization of industries that were once considered difficult to fund and bootstrap, moving back towards the core of innovation.

AI accelerates hardware development and lowers barriers

A key driver for this 'atoms' trend is the accelerating power of AI. Advanced AI models, including generative AI for code, are reducing the previously significant software engineering bottleneck in hardware development. Palmer Luckey's insights into how code generation tools, like those at Anduril, can drastically speed up work that previously required large teams of top-tier engineers. This means that startups can achieve complex R&D breakthroughs much earlier and faster. For instance, companies working on silicon photonics or advanced robotics no longer need to hire a thousand engineers to compete with tech giants. The reduced need for massive software teams makes sophisticated hardware development more economically feasible for smaller, more agile startups.

Experienced founders and solo entrepreneurship on the rise

The landscape is also seeing a resurgence of experienced founders, with individuals in their late 30s, 40s, and even 50s starting companies. These founders often bring a wealth of knowledge, understanding market needs and potential pitfalls. Concurrently, there's a notable trend towards solo founders. The percentage of YC companies with solo founders has jumped from around 5% to over 18-19% in recent batches. Historically, building a startup, especially in hard tech, required a diverse skill set best covered by co-founders: a hustler, a salesperson, and a world-class technologist. However, with the advancements in AI agents that can assist with ideation, sales, and even coding, the bar for an individual to get a company off the ground has been significantly lowered. While co-founders remain valuable, the necessity to have them from day one is diminishing.

Accelerated revenue growth for startups

The rate at which YC companies achieve revenue has also accelerated. The median YC company, which used to have around $8,000 in monthly revenue by the end of a batch, now reaches approximately $20,000. A truly shocking statistic is that some companies are now breaking from zero to seven figures in revenue within the 3-month batch period, a feat that previously took 18 months or more. This rapid growth is attributed to founders solving real problems with more mature products, often built with the help of AI agents which can significantly accelerate product development cycles. This accelerated revenue growth validates the idea that end-to-end solutions, which fully automate a job, are more valuable and command higher prices than point solutions.

The booming market for AI data and compute infrastructure

The insatiable demand for AI has created booming sub-sectors. Companies selling data or reinforcement learning (RL) environments to AI labs have become a major category. YC has funded over a dozen such companies in the last two years, each generating over $10 million annually, with some reaching hundreds of millions. Big AI labs are reportedly spending around a billion dollars on these resources. Furthermore, the need for compute power is driving innovation in data centers and semiconductor design. NVIDIA GPUs are appreciating in cost due to high demand, leading to startups focused on building data centers, improving power solutions, and even designing alternative silicon chips. Companies like Dipole Labs are developing optical switches to overcome the bottleneck of electronic switches in data centers, demonstrating how AI's demand is pushing the boundaries of physical infrastructure.

The transformation of software and SaaS

While the trend is towards 'atoms', software and SaaS are not dead; they are transforming. The percentage of YC companies focusing on full-stack, end-to-end tasks—where an AI agent performs the entire job—has risen from 10% to over 25%. This contrasts with older SaaS models that were merely point solutions requiring human operators. Modern software is expected to run autonomously and deliver complete workflows, such as insurance brokerage or medical billing. Companies like Salesforce are seeing renewed growth by positioning themselves as 'harnesses' for AI agents, becoming systems of record where collaboration and AI interactions occur. This shift implies that software valuable to AI agents will be the future, potentially leading to new 'harness wars' among platforms.

The future of robotics and physical intelligence

Robotics is another area poised for significant growth, often described as half AI and half hardware. While a 'ChatGPT moment' for robotics hasn't fully arrived, progress is rapid, with benchmarks showing dramatic improvements. Companies are building the entire stack, from vertical robotics for specific industries to the infrastructure for deploying robots and selling data to robotics labs. A hypothesis is that foundational models for robotics, which model reality in 3D physical space, will require fine-tuning on custom data for specific environments, unlike LLMs which can be more general. This specialized approach is crucial for real-time, responsive robotic actions, as seen with companies fine-tuning models for tasks like data center cabling. The ultimate goal is to enable AI to work effectively in the physical world.

Encouragement for builders in the current era

The current era offers unprecedented opportunities for builders. The acceleration in AI capabilities means that what seemed impossible a month ago can be reality today. The advice for aspiring founders is straightforward: start prompting. The ability to effectively prompt and direct AI agents, combined with domain expertise, is now the high-order bit for building a successful company. This is especially true for experienced individuals who understand market needs and can manage AI agents akin to managing human teams. The message from YC is one of excitement and encouragement: the tools are more powerful than ever, and founders should seize this moment to build.

Mentioned in This Episode

●Software & Apps

●Companies

●Organizations

●Studies Cited

●People Referenced