Y Combinator
A case study in Startup Education that Built Silicon Valley
Original Business Case Weekly articleLately I was scrolling through the Spring 2026 batch acceptances and noticed that nearly every company getting in was building agentic AI. My first reaction: every accelerator in 2026 is accepting AI-anything startups — and YC had finally become indistinguishable from others.
A Fall 2024 company called Starcloud — which is building data centers in orbit, raised a Series A at $1.1B about a year and a half after Demo Day. That is the fastest unicorn in YC’s history [1]. And across the firm’s twenty-one years, cumulative capital deployed sits around $1.5 billion. Combined portfolio NAV (Portfolio Net Asset Value is the total value of all assets held in an investment portfolio, minus any liabilities) sits around $600 billion.
That is a 400× blended multiple on 5,668 companies [1][2], which is a number that should not exist in financial markets and yet somehow does. My second thought: no — these guys still have a track record, including in recent batches, of picking and scaling winners that nobody else matches.
So how a $200,000 educational expense that four people ran out of a Cambridge living room — they literally called it “a combination of an educational expense and a charitable donation” [3] — became the institution that underwrites a meaningful slice of the public technology sector?
A Dinner in Harvard Square, 1998–2005
On the evening of March 11, 2005, Paul Graham and Jessica Livingston walked home from dinner in Harvard Square toward their apartment in Cambridge, Massachusetts.
Your call
You are Graham in March 2005, sitting on $7M from the Viaweb exit. You can write four well-chosen $50,000 angel checks a year for the next decade, or run an unproven group program that requires you and your spouse to host eight teams in your house every week for a summer. What information would you need to choose between them?