Analysis

A Walk Home in Cambridge

On March 11, 2005, Paul Graham and Jessica Livingston were walking home from dinner in Harvard Square when they decided to start a new kind of seed firm. Graham had just sold his startup Viaweb to Yahoo in 1998, and he and Livingston wanted to learn how to be angel investors. They recruited Robert Morris and Trevor Blackwell, Graham's old collaborators from the Viaweb era and from MIT, and pooled $200,000 of their own money—$100,000 from Graham, $50,000 each from Morris and Blackwell. The firm was briefly called Cambridge Seed before being renamed Y Combinator to shed any regional connotation.

The four were an unusual investment committee. Morris was a computer scientist famous for writing the 1988 Morris worm and later a tenured MIT professor; Blackwell was a roboticist and engineer; Livingston had come out of investment banking and PR; Graham was a Lisp hacker turned essayist. None of them were career venture capitalists, and that was the point: they intended to fund companies the way they wished someone had funded them.

Funding Like a Batch Process

Instead of meeting founders one at a time, they ran the Summer Founders Program: a single open call, a cohort funded simultaneously, and a fixed three-month sprint in Cambridge. The first batch in the summer of 2005 had just eight startups, each given roughly $6,000 per founder—pocket money even then, enough only to cover ramen and rent while founders built. The amounts were deliberately tiny because the scarce resource was not capital but attention and conviction.

The batch structure was originally a shortcut—Graham later wrote that a summer program for undergraduates 'seemed the fastest way' to learn angel investing—but it turned out to be the product itself. Funding many startups at once spread risk across a portfolio, created a peer group of founders who pushed each other, and let the partners give intense, hands-on help in compressed bursts ending in a Demo Day. That first cohort included Reddit, Loopt (co-founded by a 20-year-old Sam Altman), Infogami, and Kiko, seeding outcomes that would later validate the whole approach.

Why Hackers, Not Suits

The founding thesis was that the best judges of technical founders are technical founders. Graham argued that investors should make more, smaller bets, on younger people, and that the gatekeepers of finance systematically underrated programmers who could build but couldn't yet pitch. Because three of the four partners were hackers who had shipped real products, they could evaluate a team on whether it could actually build something people want—a phrase that became YC's motto—rather than on polish or pedigree.

This was less a financial innovation than a cultural one. By having founders run the fund, YC inverted the usual power relationship: the people writing the checks had themselves sat in the founder's chair, eaten the same ramen, and shipped the same buggy first versions. That credibility is what let YC ask for only a small slice of equity, demand intensity, and still attract the ambitious—an arrangement traditional VCs, optimized for capital deployment rather than empathy with builders, were poorly positioned to copy.

From $20k for 6% to a Moving Target

When Paul Graham and his co-founders ran the first batch in the summer of 2005, the offer was deliberately small: roughly $20,000 in exchange for about 6% of each company. The sum was barely enough to live on—the point was the program, the network, and the demo day, not the cash. As YC's reputation compounded, the number climbed in fits and starts. In 2011 Yuri Milner and SV Angel began layering an additional uncapped note on top of YC's own check; by 2014 the headline figure was $120,000 for 7%; in 2018 it rose to $150,000 for 7%; and in June 2020, citing the pandemic and a desire to fund thousands more companies, YC trimmed it back to $125,000 for 7%.

Two things mattered more than any single dollar figure. First, the equity stake settled at 7% and stayed there—a round, memorable number that became the anchor for every conversation about YC dilution. Second, in late 2013 YC introduced the SAFE (Simple Agreement for Future Equity), and in 2018 the post-money SAFE, replacing convertible notes with a short, standardised instrument. The SAFE collapsed a fundraise down to one or two negotiable variables, and because YC published it as an open template, it spread far beyond YC itself to become the default early-stage instrument across Silicon Valley.

The $500k Split: Certainty Plus Optionality

On January 11, 2022, YC announced the deal that defines the program today: a total of $500,000, split across two separate SAFEs. The first is the familiar one—$125,000 for 7% on a post-money SAFE, a fixed and known price. The second is the new piece—$375,000 on an uncapped SAFE carrying a Most Favored Nation (MFN) provision. 'Uncapped' means no valuation is attached up front; the MFN clause means that this SAFE automatically adopts the terms of the lowest-cap (or otherwise most favorable) SAFE the company issues between the start of the batch and its next priced equity round.

The design splits the investment into two psychologically distinct halves. The $125k half is certainty: a founder knows exactly what they gave up. The $375k half is deferred and founder-friendly by construction—YC explicitly takes whatever terms the founder later negotiates with other investors, rather than dictating a price during the batch when the company is at its most fragile and information-poor. In effect, YC hands over real operating capital while postponing the valuation question to the moment the founder has the most leverage to answer it well.

Why One Deal for Everyone

A single, public, identical deal does quiet but heavy work. It removes negotiation entirely: a 19-year-old first-time founder and a serial entrepreneur with two exits sign the same paper, so neither time nor leverage is spent haggling, and no one wonders afterward whether they were taken advantage of. It also makes YC's economics legible at scale—when you fund hundreds of companies per batch, you cannot bespoke-negotiate each one, so standardisation is what allows the model to grow toward thousands of startups a year without proportionally more deal-making.

The trade-offs are real and worth naming. A flat 7% is comparatively expensive for an already-traction-rich startup that could have raised on better terms elsewhere, and the uncapped MFN SAFE, while founder-friendly in tone, still adds $375k of future dilution whose size depends on a round the founder hasn't run yet. Critics note that 'standard' also means 'non-negotiable,' and that uncapped MFN paper can complicate cap tables and downstream rounds. But for the median founder—uncertain, under-leveraged, and prone to underpricing themselves—a known, uniform deal is precisely the kind of guardrail that converts a high-stakes negotiation into a signature.

Three Months, One Rhythm

A YC batch runs roughly three months—an intensive cycle that culminates in Demo Day. The mechanics are deliberately simple and repeatable: each batch is split into small sections of about six to ten companies, each shepherded by YC partners through group office hours and one-on-one sessions, so founders get an intimate setting inside a much larger cohort. The connective tissue is the weekly dinner, where founders gather to hear talks from successful founders and investors. There is little classroom teaching; the program is built around fast iteration, talking to users, and the pressure of a fixed deadline.

Above the schedule sits the four-word doctrine 'make something people want.' Co-founder Paul Graham traces the phrase to a hard-earned lesson from his own pre-YC startup, Viaweb: he spent months building software—an early product to put art galleries online—that nobody actually wanted. The batch is engineered to surface that mistake early. Office hours relentlessly push founders toward users and growth, and the whole three months are arranged so that a team can fail, pivot, and re-aim before the deadline rather than after raising money.

Demo Day as the Forcing Function

Everything converges on Demo Day, when founders pitch to a large audience of investors at the end of the program. The format dates to the very first batch in the summer of 2005—eight companies, including Reddit, the precursor to Twitch from Justin Kan and Emmett Shear, and Sam Altman's Loopt—who after three months presented to a room of investors. The deadline is the real product: a fixed, public date that compresses a year of normal startup procrastination into a quarter and turns a roomful of competing VCs into leverage for the founder.

Over time YC standardized the surrounding deal so the machine could run at scale. The terms became a published formula rather than a negotiation—today $500,000 in total, structured as $125,000 on a post-money SAFE for roughly 7% equity plus an additional $375,000 on an uncapped SAFE. Identical paperwork for every company means a partner can advise eighty teams without re-cutting eighty bespoke contracts, which is precisely what let the batch grow.

From Eight to Hundreds—and the Costs of Scale

The batch grew relentlessly. Eight companies in 2005 became dozens, then hundreds; the Summer 2021 cohort presented 377 companies on Demo Day, and batches hovered near 250 to 300 for years after. Scale brought obvious gains—more shots on goal, a denser alumni network, a self-reinforcing brand that pulls in the next round of applicants—but also strain. A 'class' that knew each other by name became a crowd; Demo Day stretched, then fragmented and partly moved online; and critics argued YC was spraying capital across too many similar companies, diluting the signal a YC stamp once carried.

The most telling move was the correction. In 2024, under president Garry Tan, YC shifted from two batches a year to four—adding spring and fall to the long-standing winter and summer cohorts. The yearly intake stayed around the same, roughly 500 companies, but each individual batch shrank back toward 100-some startups, near where it had been a decade earlier. In effect YC decided that the batch had an optimal size, and that the way to fund more founders was to run the machine more often rather than make each run bigger.

Bookface: the network made software

Bookface is YC's private internal platform — part directory, part forum, part deal desk — and it is the closest thing the firm has to a real product. Much of it was built by Garry Tan during his time as a YC partner from 2011 to 2015, before he returned in 2022 as president and CEO. The mechanics are deliberately mundane: a searchable directory of essentially every person who has ever gone through YC, filterable by batch, city, university, prior employer, and industry; a forum where founders ask and answer questions with an unusually high signal-to-noise ratio because every participant is a peer who has lived the same problem; and a distribution layer for partner-written playbooks and negotiated software discounts. Observers describe it as a blend of LinkedIn, Quora, and Facebook restricted to one community.

What makes Bookface load-bearing rather than a perk is the trust norm wrapped around it. YC asks founders not to repeat what they read there, and the candor depends on that wall: people post real fundraising numbers, blunt ratings of specific VCs, and admissions of near-death moments they would never publish. A 2021 TechCrunch piece probed exactly this tension — whether 'what happens at YC stays at YC' — because the value of the forum is inseparable from its privacy. The honest reading is that the secrecy is both the moat and the risk: it concentrates extraordinary information inside the network while making that same network harder to audit from outside.

Founder-to-founder help at scale

The directory's leverage comes from sheer accumulated mass. YC has funded more than 5,000 companies whose combined valuation exceeds $800 billion, including over 100 unicorns; by YC's own count the community was approaching 10,000 founders in early 2023, with broader tallies citing 9,000 to 11,000-plus people across 4,000-plus companies. A first-time founder logging in can, within minutes, find an alum who has navigated a SOC 2 audit, a crypto-friendly accountant, or a warm path to a specific partner at a specific fund. Topic-specific channels and WhatsApp groups segment this further — hardware, biotech, fintech, international, and groups for women, Black, and Hispanic and Latino founders — so help arrives from people who share the same niche, not just the same logo.

It is worth being precise about what this is and is not. The forum does not manufacture good companies; selection does most of that work, and survivorship makes the network look more magical than the median experience. What the network reliably compresses is search time. A question that might take a solo founder weeks of cold outreach — who is a fair lead for a $2M seed, which vendor will not lock you in, how aggressive is this term sheet — collapses to a single post answered by someone with direct experience. The edge is not secret knowledge so much as the removal of friction and the credible signal that the person answering has skin in the same game.

Work at a Startup and the compounding loop

The network also reaches outward through Work at a Startup, YC's job platform, where one application can be sent to roles across more than 1,000 active YC companies — from week-old batch startups offering meaningful equity to late-stage firms with thousands of employees. This turns hiring, normally a founder's most painful bottleneck, into a shared utility: talent that joins one YC company is already inside the orbit, and engineers who leave often start companies that apply back to YC. Founder verification, surfaced through YC's 'Verify Founders' tooling, lets outside parties confirm someone really went through the program, extending the network's trust signal beyond Bookface's walls.

The Aggregate Ledger

Since Paul Graham, Jessica Livingston, Robert Morris and Trevor Blackwell launched the first batch in 2005, YC has funded more than 5,000 companies, and the headline figures are genuinely large: depending on the source and methodology, the combined valuation of the portfolio is cited somewhere in the range of roughly $600 billion to $800 billion, with on the order of 80 to 100 companies having crossed the $1 billion unicorn threshold and a handful reaching decacorn scale. These numbers should be read with care. They mix public market caps that move daily with private valuations frozen at the last funding round, and they are dominated by a small number of outliers rather than spread evenly across the cohort.

That concentration is the real story of the ledger. A few public names, led by Airbnb, DoorDash, Coinbase and Instacart, account for the large majority of the realized public-market value, while the most valuable private company in the family, payments firm Stripe, was marked at $91.5 billion in a February 2025 employee tender and higher in subsequent secondary sales. The thousands of other YC startups, most of which never reach a billion dollars, are the base of a power-law distribution in which the top of the curve carries almost all the weight.

The IPO Class of 2020-2024

The portfolio's public face was forged in a compressed window. On the same day, December 9, 2020, both Airbnb and DoorDash debuted, each closing far above its offer price in the pandemic-era boom. Coinbase followed with a direct listing on Nasdaq in April 2021, foregoing a traditional raise. After a long IPO drought, Instacart went public in September 2023 near a $10 billion valuation, well below its 2021 private peak, and Reddit priced at $34 a share in March 2024 for roughly a $6.4 billion valuation, in the first major social-media listing since Pinterest in 2019. Earlier, Dropbox had been one of 2018's largest tech IPOs.

What this run demonstrates is less a single triumph than a cycle. The 2020-2021 cohort priced into euphoria; the 2023-2024 cohort priced into a colder, higher-rate market that demanded a path to profit, and several of these stocks traded well below their listing-day highs in the years after. Taken together they show that YC's model can carry companies all the way to the public markets across very different macro regimes, while also exposing how much of the celebrated valuation is mark-to-market and therefore reversible.

Three Verticals, Reshaped

Across consumer, fintech and B2B software, YC alumni did not merely participate in markets; in places they reset the defaults. In consumer, Airbnb normalized staying in a stranger's home and DoorDash helped make on-demand delivery an everyday utility, while Reddit and Instacart anchored online community and grocery logistics respectively. In fintech, Coinbase became the most visible regulated on-ramp to crypto, and Stripe turned accepting a payment into a few lines of code, lowering the barrier for a generation of internet businesses to charge money at all.

The B2B layer is less famous but arguably stickier, with companies such as Gusto in payroll, Brex and Ramp in corporate spend, and a long tail of developer and infrastructure tools embedding themselves into how other startups are built. This is the recursive effect that gives the chapter its title: YC-funded infrastructure increasingly powers the next YC batch, so the accelerator's footprint compounds not only through valuation but through the tooling and distribution that later founders inherit by default. The honest caveat is that influence is not the same as dominance; in every one of these categories YC companies compete with, and are often outsized by, incumbents and non-YC challengers.

The PG / Jessica Era (2005–2013): A Workshop, Not a Firm

Y Combinator was founded on 11 March 2005 by Paul Graham, Jessica Livingston, Robert Morris, and Trevor Blackwell, with roughly $200,000 of the partners' own money. The model itself was the innovation: instead of writing large cheques to a few companies, YC put small amounts into many, gathered them into batches, and ran a fixed three-month program ending in Demo Day. The first cohort, Summer 2005, took just 8 startups out of about 225 applications and included Reddit. Graham supplied the public voice through his essays and the technical credibility; Livingston, often underweighted in the telling, built the culture, ran operations, and authored the founder-empathy that became YC's signature.

This era's defining quality was intimacy and taste. Selection ran on the partners' personal judgment, batches were small enough to fit around a dinner table, and the program optimized for a single output — make something people want, then talk to users. It was a workshop that happened to deploy capital, not a fund that happened to mentor. That hand-built quality was also its ceiling: the format scaled with the partners' attention, which is exactly the constraint the next era was built to break.

The Altman Scaling Era (2014–2019): From Batch to Platform

Sam Altman, who had joined as a part-time partner in 2011, became president in February 2014 when Graham stepped back. His ambition was explicitly industrial: he spoke of funding up to 1,000 companies a year and pushed YC toward 'hard tech' — biotech, hardware, energy — alongside software. Batch sizes climbed steadily into the hundreds, and the program grew an institutional scaffolding around the core: Startup School, the Series A program to help alumni raise their next round, the YC Growth program, Work at a Startup, and a short-lived YC China.

The keystone move was YC Continuity, a roughly $700 million growth-stage fund launched in October 2015 to make pro-rata follow-on investments in successful alumni. This converted YC from a pure entry-point accelerator into a multi-stage capital platform that could ride its winners up the curve. The trade-off was real: the personal, taste-driven workshop became a larger, more process-driven machine, and the gravitational pull of late-stage capital began to compete with the early-stage mission. Altman departed in March 2019 to focus on OpenAI, where he had become CEO — a foreshadowing of which technology would dominate YC's next chapter.

The Interim and the Tan Reset (2019–): Refocusing on Day One

The handoff was a two-step. Geoff Ralston, a long-time partner, served as president from 2019 to 2022, steering YC through the pandemic — which forced batches fully online and, paradoxically, swelled them to record sizes above 400 companies. Michael Seibel ran the accelerator as its CEO until 2020 and then became managing director of early stage; the partnership flattened away from having multiple CEOs. It was a stewardship period more than a reinvention: the platform Altman built kept running, larger and more remote, but without a sharp new thesis.

Garry Tan — himself a YC alum (Posterous) and founding partner of Initialized — was announced in August 2022 and took over as president and CEO in January 2023 with an explicit mandate to refocus. Within months he wound down the Continuity Fund and cut roughly 17 late-stage roles, reversing the scaling era's signature expansion to put resources back into the core batch. The new thesis was the early stage sharpened by AI: batches shifted to smaller, in-person cohorts in San Francisco's Dogpatch, founders skewed younger, and by the Winter 2024 batch about half the companies were building with AI. YC, in effect, chose to be a workshop again — only this time at the center of the AI boom.

Knocking on the world's door

Y Combinator began as a deeply local institution, centred on Mountain View dinners and a network that assumed founders would simply move to the Bay Area. That assumption started loosening in 2016, when partners toured roughly eleven countries — including India, Nigeria, Argentina, Chile, Mexico, Germany and Russia — to court founders abroad. The outreach intensified region by region: more active European support arrived around 2018, a sharper push into Latin America and South Asia followed from 2019. By 2019 international companies made up close to 40% of a typical batch, a figure unthinkable a decade earlier.

The bet was partly philosophical and partly empirical. YC's own data on the 2015–2018 cohorts showed international startups landing on its top-companies list at a comparable or slightly higher rate than US and Canadian ones, undercutting the idea that talent was concentrated in one zip code. Standouts like Brazil's NuBank, Argentina-rooted fintechs, India's Razorpay and Africa's Paystack and Flutterwave gave the thesis concrete proof points and made YC a genuinely global brand rather than a Valley club.

The remote interlude

COVID-19 forced the issue overnight. The Winter 2020 Demo Day was pushed forward and moved online into a stripped-down, one-slide-per-startup format; the Summer 2020 batch then became YC's first fully remote cohort, with its 31st Demo Day held live over Zoom and reportedly generating over 28,000 founder–investor introductions. Without the gravitational pull of physical relocation, the funnel widened dramatically. Summer 2021 swelled to 377 startups from 47 countries — the largest batch to date and the most geographically diverse, with India alone fielding 33 companies, the UK 18, Mexico 17 and Singapore 12.

Remote operation did more than scale headcount; it redistributed access. Roughly half of the S21 cohort was based outside the US, African participation hit a single-batch record, and founders in Lagos, Bangalore, Mexico City or Jakarta could join without uprooting their lives or burning runway on Bay Area rent. For a brief window, geography stopped being a tax on ambition, and YC looked less like a place and more like a protocol any founder anywhere could plug into.

The pull back to San Francisco

The opening did not last. In 2022 YC told its Summer batch to show up in person, kicking off with a Sonoma retreat and weekly in-person meetups, and simultaneously cut cohort size by about 40% — from around 414 to roughly 250 companies — citing the funding downturn. Under Garry Tan, who became president in early 2023, the in-person doctrine hardened: YC relocated its headquarters from Mountain View to San Francisco in spring 2023, and Tan argued bluntly that founders "have to be in San Francisco," close to the frontier of AI, urging them toward neighbourhoods like Dogpatch and Potrero Hill.

The reversal is visible in the numbers. India's YC count, which peaked at 74 startups across 2021's two batches, fell to roughly 40 in 2022 and 18 in 2023, with YC itself acknowledging the in-person return as a factor. San Francisco's share of companies, around 21% during the remote peak, climbed back toward dominance as batches shrank and US focus reasserted itself. The remote era proved YC could be global; the post-2022 era revealed how much it still believed proximity — to capital, to peers, and to the AI frontier — was the real product.

From a trickle to a flood

AI was always present at YC, but for most of the 2010s it was a minority sport. The inflection came with the W21 batch onward. Winter 2021 was YC's first fully remote, supersized cohort, with roughly 350 companies, and only a modest slice of them were genuinely AI-first. Two years later the picture had transformed. In June 2023 Bloomberg reported that 35% of YC's latest batch was made up of AI startups, framing it as a record. By Winter 2024, Crunchbase noted that about half the batch was built with AI, and by Summer 2024 multiple outlets put the AI share near three-quarters of all funded companies.

The trigger was external as much as internal. OpenAI's release of ChatGPT in November 2022, on the heels of GPT-3 and a wave of open foundation models, collapsed the cost and skill required to build a useful AI product. Suddenly two-person teams could wrap a frontier model in a vertical workflow and show real usage within weeks. That maps cleanly onto the batches: the AI share climbs steeply across S22, W23 and S23, exactly the cohorts applying after the generative-AI moment became unavoidable.

Who got funded

The AI batches are not monolithic. The early surge was dominated by copilots and assistants layered onto existing jobs. Vapi (W21) built voice AI for developers; Julius (S22) marketed itself as an AI data scientist; Juicebox (S22) used AI for talent sourcing. By W24, voice and contact-center plays like Retell AI showed how quickly a single capability, in this case natural-sounding phone agents, could anchor a company. The common thread is narrowness: pick one painful task in one industry, and let the model do the heavy lifting.

What is notable is how little these companies resemble the deep-tech AI labs of the prior decade. Almost none train their own frontier models. They are application-layer businesses betting that distribution, workflow integration and proprietary data, not model weights, are where durable value sits. That is a deliberate bet, and a fragile one: it leaves many of them exposed to the same foundation-model providers they depend on, a tension YC partners openly acknowledge.

How RFS steered the fleet

YC did not merely ride the wave; it shaped it through its Requests for Startups (RFS), a long-running tradition of publishing the ideas partners most want to see built. After Garry Tan became president in January 2023, the RFS leaned hard into AI. Successive lists asked for AI copilots, then AI agents, and eventually full AI-native companies, with Tan publicly urging founders to treat AI agents 'not as features but as the core operating system of brand-new companies and industries.' Specific verticals were named repeatedly: insurance brokerage, accounting, tax, compliance and healthcare administration.

RFS is non-binding, framed as extra validation rather than a requirement, but its gravitational pull on a self-selecting applicant pool is real. When the institution that funds you publishes a list dominated by AI, and revenue at AI startups in the batch reportedly grows 10-20% weekly versus the historical 2-4%, the incentive is unambiguous. The recent batches read less like an open call and more like a curated thesis on where software is heading.

Scale vs. Dilution: Has the Batch Gotten Too Big?

The most persistent fair criticism is arithmetic. YC's batches grew from 46 startups in Winter 2013 to more than 400 in Winter 2022 — its largest cohort ever — before the firm cut Summer 2022 by roughly 40% (from 414 to about 250 companies), citing the funding downturn. In 2024 it restructured again, moving from two batches a year to four and halving each cohort to around 125. Critics, and even YC's own limited partners, ask the obvious question: at that volume, is the brand still a mark of selection, or just a mark of attendance? The complaint that bigness 'waters down' the signal — making it harder for any single founder to stand out at Demo Day in front of a fatigued investor audience — is a recurring, almost cliché, conversation in Silicon Valley.

The defense is that the funnel, not the batch, defines selectivity. YC still rejects the overwhelming majority of applicants (an acceptance rate often cited near 1%), and partners like Garry Tan argue the quality of selected companies has held or improved even as raw numbers rose — pointing to a unicorn rate (roughly 4.5–5.5% of funded companies) well above the industry norm. Both things can be true: YC can admit far more startups than any rival and still maintain a higher hit rate than the market, because it is filtering from a vastly larger and self-selected applicant pool. The honest tension is that scale optimizes for YC's portfolio returns — where a handful of outliers pay for everything — more cleanly than it optimizes for the median founder's experience inside an ever-larger room.

The Standard Deal and the Power Imbalance Debate

In January 2022 YC overhauled its 'standard deal': $125,000 for 7% on a post-money SAFE, plus an additional $375,000 on an uncapped SAFE carrying a Most Favored Nation (MFN) clause — meaning that second tranche automatically converts at the most favorable terms of any SAFE the company issues before its next priced round. Critics, especially angels and smaller seed funds, argued this was YC leveraging its dominance: the MFN provision lets YC capture the best terms anyone else negotiates without doing the negotiating, which some framed as effectively crowding value-added early investors out of the cap table. Founders, the argument goes, may also be disincentivized from accepting small but useful angel checks, and risk over-dilution when a priced lead's ownership target stacks on top of YC's position.

The counterargument is that a non-negotiable standard deal is a feature, not a bug — it spares first-time founders from negotiating against a far more sophisticated counterparty, and the extra $375,000 is real, founder-friendly runway that reduces the pressure to raise on bad terms right after Demo Day. The MFN, in YC's framing, simply declines to be a sucker: it ensures YC isn't paying a worse price than the angels who follow it in. Whether that is fair leverage or unfair leverage depends largely on where you sit. To a founder it can read as a clean, take-it-or-leave-it term sheet from a brand that adds genuine value; to a competing seed investor it can read as a gatekeeper using its position to write its own terms into rounds it didn't lead.

Hype, Herding, and the Series A 'Guarantee' That Never Was

Two related critiques target YC's culture rather than its cap table. The first is herding: when YC publishes 'Requests for Startups' and tilts hard toward a theme — AI now dominates, with the majority of recent cohorts being AI-focused — batches can fill with derivative bets chasing the same clogged markets, and outlets like Fast Company have framed this as an 'identity crisis' or 'spray and pray' at scale. Garry Tan has explicitly rejected the spray-and-pray label, but the structural reality is that funding a thousand companies a year and concentrating them on a hot theme will, by construction, produce visible clusters of near-identical startups and the occasional public copycat dispute between two companies in the same batch.

The second is the myth of the guaranteed Series A. YC's strongest signal — that its imprimatur makes the next round easier — is real but conditional: roughly 45% of YC companies reach a Series A versus a market baseline closer to a third, which is a meaningful edge and not a promise. In tougher markets the gap narrows in absolute terms; by early 2025 investors were noting a 'vibe shift,' with YC founders raising smaller rounds and the brand alone no longer clearing the bar. The accurate read is that YC compresses timelines, lifts valuations, and opens doors — but it cannot manufacture product-market fit, and a founder who treats the badge as the finish line rather than the starting gun typically leaves with little more than the check.

Inventing the Batch

When Paul Graham, Jessica Livingston, Robert Morris and Trevor Blackwell founded Y Combinator on 11 March 2005, the prevailing venture model favoured large cheques written slowly to a few well-connected teams. YC inverted it: that summer it funded eight startups at roughly $6,000 per founder, ran weekly dinners out of Graham and Livingston's house, and ended with a Demo Day pitch to investors. The radical move was the cohort itself, treating startups as a batch to be processed twice a year on a fixed three-month clock rather than as one-off deals.

The standardisation was the genius. By offering every company identical, founder-friendly terms, YC removed negotiation as a bottleneck and made early-stage funding programmable at scale. The deal has since grown to a $500,000 standard offer set in 2022, $125,000 for 7% on a post-money SAFE plus $375,000 on an uncapped Most Favored Nation SAFE, and the institution has now funded more than 5,000 companies including Airbnb, Stripe, Coinbase and Dropbox.

A Template the World Copied

Within a year the format had a rival and a name. Techstars, founded in 2006 in Boulder by David Cohen, Brad Feld, David Brown and Jared Polis, took YC's batch-plus-Demo-Day structure and grafted on a deep mentor network and a city-by-city franchise logic. 500 Startups, launched in 2010 by Dave McClure and Christine Tsai, pushed the model toward high-volume global cohorts; Antler, founded in 2017 in Singapore by Magnus Grimeland and Fridtjof Berge, stretched it earlier still, assembling founders before they even had teams or ideas.

Together these established the accelerator as a recognised asset class, spawning hundreds of regional and vertical imitators from Y Combinator's blueprint. The diffusion was uneven: many imitators replicated the three-month cadence and the Demo Day theatre without the dense alumni network and brand gravity that make YC's signal valuable, and academic studies of accelerator outcomes have been decidedly mixed. The format proved easy to copy; the network effects underneath it did not.

Democratising the Curriculum

The model's second export was knowledge, not capital. In April 2017 YC launched Startup School as a free online course, its first cohort drawing roughly 3,000 companies from 141 countries; it built on the popularity of Sam Altman's earlier Stanford lectures and bundled essays, video sessions with founders like Dustin Moskovitz and Stewart Butterfield, and group accountability into an open programme. Alongside it, YC's library of essays, the Startup Library and tools like the standardised SAFE made the tacit playbook of Silicon Valley legible to anyone with an internet connection.

This was both generous and strategic. Free curricula widened the global funnel of founders who already think in YC's vocabulary, reinforcing the accelerator's gravity even for those it never funds. The deeper effect is cultural: by codifying startup wisdom into repeatable lessons, YC helped shift the act of company-building from an apprenticeship reserved for the well-connected toward something closer to a learnable discipline, for better and, when the advice ossifies into orthodoxy, sometimes for worse.

The Aggregate at a Glance

This dataset records 5,988 companies launched across 50 batches between Summer 2005 and the most recent 2026 cohorts. Of those, 4,129 are still tracked as active, 794 have been acquired, 1,042 have shut down, and 23 are public companies on the dataset, names ranging from Airbnb, Coinbase, DoorDash and Dropbox to GitLab, Instacart, Ginkgo Bioworks and Rigetti. A further 91 are flagged as top companies and 237 sit in the curated top tier. Those counts are the skeleton everything else hangs on, and they are unusually clean for a venture portfolio because YC defines membership precisely: you were either in a batch or you were not.

The valuation numbers are where precision gives way to estimate. Y Combinator's own marketing has at times claimed a combined alumni valuation 'over $1 trillion,' while independent trackers and press accounts more often cite a range from roughly $600 billion to $800 billion. Unicorn counts vary just as widely: public lists track on the order of 80 active billion-dollar companies, while YC and some analysts cite 250 or more unicorns historically, a cumulative figure that counts every company that ever crossed the line, including those since acquired, taken private, or marked back down. None of these is wrong; they answer different questions, and the gap between them is itself the first lesson of the chapter.

From Eight to Four Hundred

The most legible trend in the data is batch size. The founding Summer 2005 cohort had just 9 companies; Winter 2006 had 7. For years the batches stayed small, an intimate dinner-and-office-hours format that scaled into the dozens. Then the curve bent sharply upward: by the early 2020s the largest single batches reached 398 (Winter 2022) and 391 (Summer 2021), with Winter 2021, Winter 2023 and Winter 2024 all clearing 250. Since 2024 YC has run four batches a year rather than two, so the annual intake now dwarfs anything from the first decade.

That growth changes what the accelerator structurally is. A 9-company batch is a workshop; a 400-company batch is closer to a seed-stage index, writing a standardized check into a very wide slice of the year's startups and accepting that most individual bets will not pay off. The bet is not on any one company but on the portfolio's right tail being fat enough that one Stripe, Airbnb or Coinbase per era covers the rest many times over. Scaling the funnel raises the absolute number of shots on goal, but it does nothing to guarantee the hit rate, and it dilutes the hands-on attention that the small early batches could lavish on each team.

What the Hit-Rate Reveals

Read as a survival table, the portfolio is a textbook power law. Of the 5,988 companies, 23 reaching the public markets is about 0.4 percent; if 80-odd are current unicorns, that is on the order of 1 to 1.5 percent at any one time, and even the broader historical unicorn tally lands near 4 to 6 percent. Against that, 1,042 recorded shutdowns and a long base of never-scaling startups make up the bulk of the cohort. The arithmetic is deliberately lopsided: the model assumes most companies return little or nothing and is engineered so that the rare outlier pays for the whole book.

This is the 'index fund of startups' logic stated plainly, and it cuts both ways. On the upside, a roughly 4 percent unicorn rate is several times the seed-stage baseline often cited near 2.5 percent, which is a genuine edge from selection, brand and network effects. On the downside, an index only works if you can hold the whole index, and YC's economics depend on the handful of names at the top, the same concentration that makes the headline valuation impressive also makes it fragile, since a repricing of two or three giants moves the aggregate more than the launch of an entire new batch. The numbers do not show a machine that reliably manufactures winners; they show a wide net, cast deliberately wide, whose value is decided by how heavy a few fish in the tail turn out to be.

One bay, gravitationally

Whatever the passport on the application, the centre of mass has barely moved. Roughly 85–90% of all YC companies are headquartered in the United States, and within that share the San Francisco Bay Area is dominant rather than merely large: external trackers put San Francisco's slice of recent batches above half of all companies — around 51% in 2024 and 53% in 2025. Add New York, Seattle, Austin and the rest of the US, and the international cohort is left contesting a minority of seats. YC is, in the most literal sense, a San Francisco institution that recruits everywhere.

This concentration is not an accident of where applications come from; it is a property of where companies end up. The application pool is genuinely planetary — founders apply from dozens of countries each cycle — but the program's gravity bends trajectories toward a single metro. The relevant geography of YC is therefore two maps laid over each other: a wide, dispersed map of origin, and a narrow, dense map of destination. Most of the analysis that calls YC 'global' is reading the first map; most of the value YC sells is priced off the second.

The growing edges of the map

Around that core, the rest of the world has gone from anecdote to standing fixture. India is now the single largest non-US source by cumulative count, with well over a hundred and fifty YC-funded companies; Latin America anchors a fintech-heavy contingent that runs from Brazil's Nubank to a steady stream of cross-border payments and banking startups; Europe contributes a deep developer-tools and B2B SaaS bench led by the UK, France and Germany; Africa is carried by the Nigerian fintech lineage of Paystack and Flutterwave; and Southeast Asia, with Singapore and Indonesia, rounds out a genuinely five-continent applicant base. Each region brings a recognisable specialisation rather than a generic copy of the Valley playbook.

But 'growing share' has to be read against a moving baseline. India and Southeast Asia together fielded scores of companies a year at the pandemic peak — on the order of 96 in 2021 — before that pipeline thinned sharply: by 2023 only a handful of YC's roughly 349 startups came from South and Southeast Asia combined, and 2024 batches counted Indian participation in the low single digits. The emerging-market presence is real and structurally larger than a decade ago, yet it is also volatile, sensitive to batch size, to YC's AI pivot after ChatGPT, and to whether a given cycle is run with relocation in mind. The edges are growing, but they expand and contract with the centre's policy, not independently of it.

Relocation as the product

The cleanest way to understand YC's geography is to stop treating relocation as friction and start treating it as the offer. Under Garry Tan, who became CEO in 2023, the program moved its headquarters from Mountain View to San Francisco's Dogpatch, declared the batch fully in-person, and stated plainly that founders 'have to be in San Francisco' — encouraged to live in Dogpatch or Potrero Hill, with the 2022 jump to a $500,000 standard deal partly framed as runway for Bay Area rent. The argument is proximity: to AI labs like OpenAI and Anthropic, to capital, and to a dense peer group of other founders. From this view, asking a Lagos or Bangalore founder to move is not a barrier to inclusion but the mechanism of it.

That framing is coherent, and it is also a filter with sharp edges. Relocation-as-feature works cleanly for the founder who is young, unencumbered and chasing the frontier; it works less cleanly for one solving a payments problem specific to Brazil, a logistics problem specific to Indonesia, or a market that simply cannot be operated from 8,000 miles away. So YC's geography settles into a stable equilibrium: it scouts the entire world for talent, accepts a meaningfully global slate, and then routes the resulting energy through one zip code — exporting its brand outward while importing its founders inward. The map of where YC looks keeps widening; the map of where YC lives stays almost exactly where it began.

Reading the Funnel

Of the roughly 5,988 companies in this dataset, about 69% are still tagged Active, around 13% have been Acquired, and roughly 17% are marked Inactive or shut down. Only 23 companies, about four-tenths of one percent, have reached the public markets. Laid out that way, the YC outcome funnel looks less like a launchpad to Wall Street and more like a wide field in which most startups are simply still standing, a smaller slice has been absorbed by a larger company, a comparable slice has quietly closed, and a vanishingly thin sliver has rung the opening bell. The dominant verb in the portfolio is not 'IPO'd' or 'died' but 'continues'.

That 69% Active figure deserves a skeptical eye, because 'Active' is the most ambiguous label in the set. It bundles together genuine breakout companies, steady profitable businesses, and so-called zombies that are technically alive but have stopped growing and may never raise again. Because most of these companies are young, the funnel is also incomplete: a startup founded in a recent batch has had little time to be acquired, to fail, or to list, so the Active bucket is partly a measure of age rather than of health. The honest reading is that the dataset captures a snapshot of an unfinished process, not a final scoreboard.

What the Tails Look Like

The 23 public companies are an unusually varied list. The headline names everyone cites are the consumer and fintech giants, Airbnb (W09), DoorDash (S13), Coinbase (S12) and Instacart (S12), joined by infrastructure plays like Dropbox (S07) and GitLab (W15). But the public column also holds companies in biotech, nuclear energy and quantum computing, such as Ginkgo Bioworks, Oklo and Rigetti, several of which listed via SPAC during the 2020 to 2021 boom and have since traded far below their debut prices. Going public, this list quietly reminds us, is the start of a new and often harsher chapter rather than a guaranteed happy ending. The acquired tail is broader and arguably more representative, running from Twitch, sold to Amazon for about $970 million in 2014, through Reddit, Heroku, Segment, Cruise, Optimizely and Twitch's many less famous peers.

Acquisition is where the realistic version of 'success' usually lives at YC. A clean sale to a Stripe, a Salesforce or an Amazon can return capital to investors, reward founders and staff, and close the story on a positive note, even when the price never approaches unicorn territory. The shutdown column, meanwhile, is the largest of the three completed outcomes at roughly 17%, and it is the one the ecosystem talks about least. Most of those closures are not dramatic blowups but ordinary fade-outs: a product that never found a market, a founding team that ran out of runway or patience, a company that was acqui-hired for talent and then wound down. Plenty of widely admired startups sit in that column, which is precisely why it should not be read as a roll call of failure so much as the base rate of early-stage risk.

The Power Law and the Mirror

The deeper truth the funnel hides is that these outcomes are not weighted equally in dollars. Venture returns follow a power law: a handful of extreme winners generate the overwhelming majority of the value, while everything else, including most acquisitions and every shutdown, contributes comparatively little. By widely cited estimates, just a few of YC's billion-dollar companies account for the large majority of its entire portfolio value, and across the family a small set led by Airbnb, Stripe, Coinbase, DoorDash and Instacart carries returns that the other thousands of companies, in aggregate, do not. This is not a flaw in the model; it is the model. YC writes many small checks precisely because no one can pick the few that will matter in advance, so the strategy is to fund broadly and let the distribution do the sorting.

This is also where survivorship bias quietly distorts the story. The companies that get studied, quoted and turned into playbooks are the survivors and the breakouts, while the shut-down 17% and the stagnating share of the Active 69% rarely get written up at all. Reading the funnel honestly means holding two facts at once: that the typical YC outcome is modest or unresolved rather than triumphant, and that the rare triumphs are large enough to make the average look far better than the median. The dataset is most useful not as a promise of where any single company will end up, but as a mirror showing the real shape of early-stage risk, in which being still in the game is the common case, a clean acquisition is a good case, quietly closing is an ordinary case, and the public-market giant is the spectacular exception that the whole system is built to occasionally produce.