Spark
By Jason Kumpf, Strategy Advisor · November 12, 2026 · 5 min read
Y Combinator's newest Request for Startups points founders toward AI that expands expertise, specialized models and the physical buildout behind AI. Here is what is on the list and why it matters.
Y Combinator publishes a short list of ideas it wants founders to pursue, and the latest edition, its Request for Startups for the Winter 2027 batch, reads like a map of where the next decade of company building is headed. AI that expands expertise, specialized models, robots and the physical buildout behind AI all feature prominently.
Read alongside YC's other 2026 moves, the list offers founders a clear signal. The most exciting early-stage opportunities now combine software intelligence with real-world impact.
YC now runs four batches a year. Its 2026 Demo Days fell on March 24 for Winter, June 16 for Spring and September 10 for Summer, with the Fall 2026 batch presenting on December 2. That cadence means a founder with a strong idea rarely waits more than a few months for the next opportunity to apply.
The partnership has grown alongside the program. In June 2026, YC President and CEO Garry Tan named Diana Hu a Managing Partner. Hu, a former YC founder whose AR company Escher Reality was acquired by Niantic, has worked with nearly 230 companies across 18 batches and logged more than 2,100 office hours. YC says those companies are worth a combined $7 billion, with breakouts including Avoca, Reducto, David AI, Salient, Stepful and HappyRobot. Her focus areas are AI, robotics and hard tech.
Her background mirrors the kind of founder YC now seeks. Hu grew up in Chile, studied computer vision and machine learning at Carnegie Mellon and later led the AR Platform team at Niantic, which brought augmented reality to the more than 100 million players of Pokémon GO. Deep technical skill paired with product instinct is a recurring theme across the newest requests.
Several requests share a hopeful theme: AI as a ladder for human talent. Partner Brad Flora's "The $15-an-Hour Expert" asks for AI that lets workers without formal credentials perform work that today requires specialists. The opportunity is enormous, because skilled expertise is scarce in almost every industry.
Partner Jared Friedman adds "LLMs That Write Well," a call for a model optimized for natural, pleasant writing. It is a reminder that quality and craft remain wide-open frontiers, even as general models grow more capable.
Both ideas start from the same insight. The most valuable AI products will help people do work they could not do before, at a level of quality they are proud of. Founders who have watched talented colleagues held back by a missing credential or a missing tool are well placed to build here.
A second cluster centers on models built for one domain and owned by the companies that use them. YC's latest list includes:
Together these point to a durable opportunity. As AI becomes core to how every company operates, many will want models tuned to their data and running where they choose.
The third cluster is about atoms as much as bits. Diana Hu's "Robots Made in America" invites founders to build domestic robot hardware and supply chains. Jared Friedman's "Data Center Supply Chain" looks for companies that expand the supply of transformers, switchgear, cooling equipment and electrical steel.
Brad Flora's "Wall Street for the Physical Buildout" pairs naturally with both. It calls for lenders, leasing companies and other financial products designed for robots, GPU clusters and data centers. Each of these ideas reflects a simple truth: the AI era needs a vast amount of physical infrastructure, and that creates room for new companies at every layer.
Partner Ankit Gupta looks even further out. "Startups in a World of Abundant Space Access" asks founders to imagine business models that become possible once reusable rockets make access to orbit dramatically more abundant. It is an invitation to think a decade ahead and be ready when the window opens.
Taken together, the requests share a few clear signals about where YC sees the strongest opportunities:
YC is also widening the on-ramp for the next generation. In January 2026 it announced the YC AI Stack, a starter pack for students who attend YC university events. It includes more than $20,000 in cloud credits on Azure and AWS, more than $5,000 in credits for models including Claude and GPT and tools from 14 YC companies covering voice, browser automation, web search and more.
YC highlighted tools such as Firecrawl, Vapi, Reducto, Exa, Browser Use and Gumloop as examples of what students can build with. The idea is straightforward. Give talented students the same tools that funded startups use, and more of them will build something worth funding.
YC's long-term results show what patient company building can produce. In January 2026, YC congratulated EquipmentShare, a Winter 2015 company, on going public. Founded to help contractors share construction equipment, it has grown into a nationwide platform with 373 locations that helps contractors manage fleets, track machines and run their operations.
If you are a founder, read the list as a set of open invitations rather than a checklist. The common thread is meaningful work: expanding expertise, building specialized intelligence and supplying the physical world that AI depends on. Pick the idea where your own experience gives you an unfair head start.
If you are an operator or investor, use these themes to frame your next conversations. Ask the founders you meet which of these shifts they see firsthand. The best early-stage opportunities usually appear where deep industry knowledge meets a newly possible technology, and those insights tend to surface when the right people compare notes.
