Diana says hard tech rose from 8% to 20% of companies accepted over the last 18 months. Robotics, manufacturing, defense, semiconductors, photonics, and power all grew, with several categories reaching roughly three to five times their earlier share. Jared adds that one in six founders in the Summer 2026 batch has a PhD, while AI coding tools are reducing some software engineering bottlenecks for hardware startups.
The State of Startups in 2026
Speakers
Y Combinator’s The Lightcone hosts and guests Diana, Jared, and Gary discuss startup trends and advice for builders with other YC speakers.
The speakers point to space, defense, and AI compute as forces behind the rise in companies building physical products and infrastructure. Examples include satellite power and connectivity, defense startups such as Icarus and Nine Mothers, and Nox Metals, which is rebuilding metal manufacturing in Detroit. They also describe startups building data centers, chips, and optical networking equipment as demand for compute outpaces supply.
The speakers argue that SaaS remains valuable when it keeps its data advantage and becomes a place where people or agents complete work. YC’s share of accepted companies building end-to-end products rose from 10% to more than 25%, and median monthly revenue at the end of a batch rose from about $8,000 to $20,000. Juicebox is an example: its recruiting agent searches for candidates, contacts them, and schedules interviews, with the company expecting revenue per account to double or triple.
YC has funded more than a dozen companies in the last two years that each make over $10 million annually selling data or reinforcement learning environments to AI labs; some reportedly make hundreds of millions. The speakers describe demand for specialized data, including real-world robotics tasks, and suggest that companies with proprietary data may fine-tune open models for specific uses. They note that robotics models often need customization for a particular environment and must respond in real time.
YC’s solo-founder share rose from about 5% to roughly 18% or 19%, which the speakers link to AI tools making it easier for one person to build and sell a product. They also see experienced founders doing well because they bring product judgment and experience managing teams, skills that can transfer to directing coding agents. Their advice to aspiring founders is to start building and revisit problems that improved models may now be able to solve.
Takeaways
- Look for hard tech opportunities where AI reduces engineering bottlenecks and customers need new physical infrastructure.
- Build software that completes a customer’s workflow, rather than only tracking it.
- If your product has valuable proprietary data, explore whether it can improve a model for a specific use case.
- A solo founder can get started with AI tools and add co-founders as the company grows.
- Experienced builders can apply product judgment and team management skills to directing coding agents.
Mentioned in this talk
Search inside this talk
What was claimed 20
- metric
Y Combinator: Median monthly revenue by the end of the batch rose from about $8K historically to $20K.“now companies are getting to $20,000 in monthly revenue, as opposed to $8K.”
YouTube Unverified1 source· Talk
- metricY Combinator: Hard tech companies increased from 8% to 20% of accepted companies over the last 18 months.
“The number of hard tech companies in the batch... has gone from 8% to 20%.”
YouTube Unverified1 source· Talk
- metricY Combinator: Semiconductor and photonics startups increased from about 1% to close to 4% of the batch.
“from about 1% of the batch a year ago to close to 4%.”
YouTube Unverified1 source· Talk
- metricY Combinator: Defense increased from about 1.5% to about 5% of the batch.
“It has gone from about 1.5% to about 5% of the batch.”
YouTube Unverified1 source· Talk
- metricY Combinator: Industrial manufacturing increased from about 4% to 10% of the batch.
“It has gone from about 4% to 10% of the batch.”
YouTube Unverified1 source· Talk
- metricY Combinator: Robotics increased from 1% to about 6–7% of the batch.
“It has gone from 1% of the batch to about 6–7%.”
YouTube Unverified1 source· Talk
- featureExosat: Building a sovereign Starlink solution.
“Exosat that's trying to build a sovereign Starlink solution.”
YouTube Unverified1 source· Talk
- featureNine Mothers: Anti-drone defense system described as a shotgun turret with computer vision.
“Nine Mothers' anti-drone defense. It's basically a shotgun turret with computer vision.”
YouTube Unverified1 source· Talk
- metricIcarus: Secured seven-figure contracts with the Department of War.
“They've been able to get seven-figure contracts with the new Department of War.”
YouTube Unverified1 source· Talk
- featureIcarus: Building a solar-powered U-2 spy plane for overwatch and communications.
“Icarus is doing a solar-powered U-2 spy plane that provides overwatch and can also do communications.”
YouTube Unverified1 source· Talk
- featureNox Metals: Rebuilding metal manufacturing and the metal supply chain in America, based in Detroit.
“Nox Metals is rebuilding America's metal supply chain.”
YouTube Unverified1 source· Talk
- featureBot: Building custom compute hardware using ternary representation for models.
“Bot that's trying to build a new custom hardware architecture using ternary representation for models.”
YouTube Unverified1 source· Talk
- featureDipole Labs: Building a fully optical switch for data centers, routing between GPUs with photons.
“building the first fully optical switch, with photons all the way from GPU A to GPU B.”
YouTube Unverified1 source· Talk
- metricY Combinator: Full-stack, end-to-end companies increased from 10% to over 25% of accepted companies.
“has gone from just 10% to over 25% of the batch.”
YouTube Unverified1 source· Talk
- launchedJuicebox: An agent product that contacts candidates and schedules interviews.
“Recently, they've launched an agent product that's really taking off.”
YouTube Unverified1 source· Talk
- metricJuicebox: Expects agent product to double or triple revenue per account.
“this will double or triple the revenue they make from a single customer.”
YouTube Unverified1 source· Talk
- metricY Combinator: Some companies went from zero to seven figures in revenue during a three-month batch.
“companies going from zero to seven figures in revenue during the batch, a span of three months.”
YouTube Unverified1 source· Talk
- metricY Combinator: Funded more than a dozen companies in the last two years that each make over $10 million a year selling data or RL environments to labs.
“YC has funded more than a dozen companies each making more than $10 million a year selling data or RL environments to the labs.”
YouTube Unverified1 source· Talk
- featurePi Labs: Robotics models are a starting point that YC companies fine-tune for specific use cases.
“all the YC companies using Physical Intelligence models to deploy robotics are fine-tuning the PI models.”
YouTube Unverified1 source· Talk
- metricY Combinator: Solo founders increased from about 5% to over 18–19% of accepted companies.
“we used to accept only about 5% solo founders, and now we're over 18 or 19%.”
YouTube Unverified1 source· Talk
