Y Combinator has posted its Winter 2027 Requests for Startups, a list of 11 ideas its partners want founders to work on. The page opens: “AI’s biggest bottlenecks are moving beyond the models.”
The page carries no date. In captures by the Wayback Machine, the Winter 2027 list first appears on 27 October 2026; a capture from 25 October still shows the Fall 2026 list.
Short explainer · YC Requests for Startups
A vertical news Short narrated from the published text of this piece and related FourWeekMBA pieces.
An evergreen Business Engineer explainer on the difference between an AI model and the harness around it. It explains the general idea and does not discuss YC.
The key insight: As we read it, YC’s Winter 2027 list moves most of its requests away from the frontier models themselves: seven of the 11 are about hardware, financing, government and work, and the four about models ask for models a company can own or specialise.
YC frames the list this way: “We’re excited about a new wave of startups building the hardware, infrastructure, financial systems, and specialized software needed to bring intelligence into the real economy, from factories and data centers to government and space.”
The page also says the ideas “represent just a fraction of what we fund” and that founders do not need to work on them to apply. Each request is signed by a YC partner.
Y Combinator’s 11 Winter 2027 Requests for Startups, each signed by a YC partner, grouped by FourWeekMBA into three themes. The grouping is ours; titles and names are YC’s. Source: ycombinator.com/rfs, read 28 Oct 2026.
The Physical Buildout and Its Money
Four requests deal with hardware and its financing. In “Wall Street for the Physical Buildout,” Brad Flora writes: “Software is cheap to start. Robots, GPU clusters, and data centers aren’t.” He says YC wants founders “building lenders, leasing companies, securitization, and hedges for the physical AI economy.”
In “Data Center Supply Chain,” Jared Friedman lists bottlenecks: “Large power transformers have a three-year lead time. Switchgear, circuit breakers, and liquid cooling equipment are all sold out.” He adds that the grain-oriented electrical steel used in transformer cores “is made in only a single plant in the US.”
In “Robots Made in America,” Diana Hu writes that “the bottleneck is moving to the hardware.” She cites 38,400 industrial robots installed in the US in 2025 and says China made 94 percent of the world’s sintered rare-earth magnets in 2024. Ankit Gupta’s request, “Startups in a World of Abundant Space Access,” asks what business models could work “when launch costs fall 100x or more.”
Models a Company Can Own
Four requests deal with the models themselves. In “Own Your Intelligence,” Jon Xu writes: “Today, nearly every company building with AI rents its intelligence.” He says “Open-weight models are now competitive with the top frontier models on meaningful benchmarks.”
In “Vertical Foundation Models,” Diana Hu writes: “I think Jev hit a nerve because it gives developers a typed decision and a probability instead of generated text.” She asks for specialised models for manufacturing, energy and biology that match general models “at orders of magnitude less inference compute.”
In “AI for the Air-Gapped Economy,” Brad Flora writes that defense contractors, banks, hospitals and nuclear plants “can’t send it to an AI API.” In “LLMs That Write Well,” Jared Friedman writes that if forced to choose, the major labs “will always choose to focus on coding and agentic workflows over style.”
Government and Work
Three requests deal with government and jobs. Gustaf Alströmer’s “AI for City Hall” says that if building permits were approved in 10% of the current time, “US cities would make $16 billion a year in additional tax revenue,” a figure YC describes as its own math. His “Plaid for Government Benefits” cites a GAO estimate that fraud costs the federal government “up to half a trillion dollars a year.”
Brad Flora’s “The $15-an-Hour Expert” says: “Most people talk about AI replacing workers. We’re excited about the opposite: AI that turns an ordinary worker into an expert.”
What’s new: YC’s partners have each signed a request, under YC’s own line that “AI’s biggest bottlenecks are moving beyond the models.”
The Structural Read
As we read it, money is a theme of its own: Brad Flora writes that the financing for robots, GPU clusters and data centers mostly does not exist yet.
As we read it, the model requests point away from renting: Jon Xu writes that nearly every company building with AI rents its intelligence, and that many will also own specialized models.
As we read it, government is treated as a market: Gustaf Alströmer writes that the conventional wisdom that you can’t sell software to government used to be true and isn’t anymore.
Y Combinator, Requests for Startups, Winter 2027
“AI’s biggest bottlenecks are moving beyond the models.”
Three Implications
FOR FOUNDERS YC says the requests represent just a fraction of what it funds and that founders don’t need to work on these ideas to apply.
FOR INVESTORS IN THE BUILDOUT Brad Flora writes that every great buildout, from railroads to electricity to telecom, created giant financial companies alongside it.
FOR MODEL PROVIDERS Jon Xu writes that companies will keep renting frontier intelligence for their hardest problems, while many will also own specialized models.
The framework says: “It is a layered industrial stack — and at any given moment, the binding constraint sits at a different layer than it did six months ago.”
As we read it, YC’s list is a bet on where that constraint sits now, in its own words beyond the models. The framework was written before this list and does not discuss YC.
What Is Not Established
YC does not say how much it will invest against these ideas or how many companies it funded from earlier lists. The figures in each request are the partners’ own, and we did not check their underlying sources. A post on X on 27 October listed different themes, such as “Company Brain” and “Software for Agents,” that do not appear on the current page.
This Week In The AI Supercycle
Execution Fault Lines, Concentrated Demand, and Geopolitical Firewalls
This week’s issue: the AI build is not short of demand for its parts. It is short of cheap money, and its end revenue rests on a narrow base of large, often venture-funded buyers.
Y Combinator’s Winter 2027 list names 11 startup ideas, from data center supply chains and financing for robots and GPU clusters to company-owned models and AI for city permits. YC says the ideas are a fraction of what it funds.
A note on sourcing. Request titles, partner names, figures and quotations are from Y Combinator’s Requests for Startups page, read 28 October 2026; the timing comes from Wayback Machine captures of that page. The three-way grouping and the links to our coverage are ours. This is not investment advice.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
