Y Combinator and the 90-Day Bet That Rewired How Startups Get Built

Two people were walking home from dinner in Harvard Square in 2005 when they decided to fund startups the way nobody else would - in batches, in cash, on a fixed schedule. Paul Graham and Jessica Livingston, joined by Robert Morris and Trevor Blackwell, pooled $200,000 and offered eight teams a summer's worth of runway: about $6,000 per founder, a number pegged to an MIT summer research stipend. One of those eight companies was Reddit. The experiment was called Y Combinator, after a function in computer science that runs other functions. It was a programmer's inside joke that turned into the most copied idea in the startup world.

Strip away the mythology and Y Combinator does something fairly simple to describe. It gives a young company money and three months of concentrated attention, then puts it on a stage in front of investors. The money today is $500,000. The three months happen in San Francisco. The stage is called Demo Day. What is hard to describe is why that combination has produced Airbnb, Stripe, Coinbase, DoorDash, Dropbox, Instacart, Twitch and thousands of others - a portfolio whose combined value now clears $600 billion.

01 / WHAT IT ACTUALLY DOESAn investor first, a school second

It is tempting to call YC a school. It publishes essays, runs a free online curriculum called Startup School, and hands out advice with the confidence of a coach. But YC does not sell education. It buys equity. Every company that gets in takes the same deal, and YC makes its money when a handful of those companies eventually go public or get acquired. The teaching is real, but it sits on top of a venture fund, not a tuition model.

The deal itself is deliberately boring, which is the point. YC invests $125,000 for 7% of the company on a post-money SAFE, then adds another $375,000 on an uncapped note designed to give YC the same terms as whoever leads the next round. There is no milestone to hit and no negotiation. The offer is identical whether you are two dropouts or a team of ex-Google engineers.

The YC Standard Deal - $500,000

$125K / 7%

$375K  -  uncapped SAFE (MFN)

$125,000Fixed 7% equity, post-money SAFE

$375,000Uncapped, converts on your next round

The invention YC is quietly proudest of is not a company at all. It is a document. The SAFE - Simple Agreement for Future Equity - was created at YC to replace the tangle of convertible notes that used to slow down early rounds. YC gave it away for free, and it became the default way pre-seed startups raise money around the world. Standardizing a form nobody enjoyed filling out turned out to be a form of leverage.

02 / WHO IT SERVESThe founder, and then all the other founders

YC's customer is the early-stage founder - often technical, often young, sometimes with nothing more than a prototype and a story about who it is for. Four times a year now (YC added a spring session in 2025), a new batch arrives. Partners run weekly group office hours, push teams to launch before they feel ready, and repeat the same few pieces of advice until they stick: talk to your users, do things that don't scale, measure whether anyone actually wants the thing.

But the money and the mentorship are not the part alumni talk about years later. They talk about the network. YC runs a private social platform called Bookface where thousands of founders answer each other's questions, make introductions, and become one another's first customers. A batchmate you can text at 2am is a real asset, and it is the one competitors find hardest to copy.

03 / THE PROBLEM IT SOLVESStarting is the hard part

Most startups do not die because the idea was wrong. They die because the founders ran out of time, money, or nerve before they found out. YC attacks all three at once. The $500,000 buys runway. The batch structure imposes a deadline - Demo Day is a fixed date, and nothing focuses a team like a countdown. And the peer group provides the nerve, the social proof that quitting a stable job to build something risky is a reasonable thing to do.

There is also a signaling problem YC solves almost by accident. Getting into YC is a credential. It tells the next round of investors that someone with a good track record already looked at this team and wrote a check. For a founder with no name and no network, that stamp can be worth more than the cash.

04 / HOW IT'S DIFFERENTVolume, plus help attached

There are plenty of accelerators - Techstars, 500 Global, Antler, Entrepreneur First, and a long tail of university programs. YC did not win by picking better than everyone else on average. It won by making more bets, more often, with more support attached to each one, and by building a brand so strong that the best founders apply there first. Deal flow feeds reputation, reputation improves deal flow, and the loop tightens.

The constraint on batch size, YC now says, is not money and not applications. It is how many founders its roughly fifteen investing partners can actually help. YC accepts under 1% of applicants - a lower rate than most Ivy League universities - yet the application takes about ten minutes and asks unglamorous questions about what you are building and who uses it. The bar is not polish. It is whether you are actually making something.

A rough map of YC's biggest outcomes (illustrative)

~$95B

Stripe

~$75B

Airbnb

~$50B

DoorDash

~$40B

Coinbase

$10B+

Reddit

05 / PRODUCTS & SERVICESMore than a three-month program

The accelerator is the core, but YC runs a small ecosystem around it. Hacker News, launched in 2007, became one of the most influential technology forums online - technically a side project of a venture firm. Startup School offers the curriculum for free to founders anywhere. Co-Founder Matching helps solo builders find a partner. Work at a Startup connects engineers with YC companies that are hiring. And its Requests for Startups - published lists of problems YC wants people to tackle - now span AI, hardware, defense, agriculture, and space.

06 / THE BUSINESS MODELA few winners pay for everything

Venture returns follow a power law, and YC lives at its extreme. The top handful of companies - Airbnb, Stripe, Coinbase, DoorDash - are estimated to account for the majority of all returns. YC can afford to be wrong about most of the batch precisely because it needs to be very right a few times per decade. That is why the free content, the forums, and the advice all make sense as a business: they widen the top of the funnel so that the rare outlier is more likely to apply to YC than to anyone else.

07 / EXPERTISE & LEADERSHIPThe Garry Tan era, built on AI

Paul Graham handed the presidency to Sam Altman in 2014; Geoff Ralston led it next; and in 2023 Garry Tan - himself a YC alumnus and an early Coinbase investor - took over as President and CEO. Tan trimmed the late-stage activities that had grown under previous leadership and pointed YC back at the earliest stage. He also leaned hard into artificial intelligence.

The results are striking. YC has said that in a recent batch, roughly a quarter of companies had codebases that were about 95% AI-generated. The Spring 2025 cohort was reported as around 90% AI companies and the fastest-growing, most profitable in fund history, at about 12% weekly revenue growth. Applications from founders aged 18 to 22 jumped 110% year on year. The accelerator that funded the last software era is now running a live experiment on the next one.

08 / WHERE IT FITSThe default first step

For a generation of founders, applying to YC is simply what you do when you decide to start a company - the same way an aspiring academic applies to grad school. That position, more than any single check, is what YC built. It sits at the entry point of the startup pipeline, and everything downstream - seed funds, later investors, acquirers - has learned to read the YC stamp as a filter. Whether AI reshuffles that order is the open question, and for now YC is betting it can lead the reshuffle rather than be flattened by it.

Two decades on, the pitch has barely changed from that walk in Harvard Square. Give founders a little money, a hard deadline, blunt advice, and a room full of peers - then get out of the way. The checks got bigger and the code writes itself now, but the index card still reads the same: make something people want.