Summer 2005, Cambridge, Massachusetts: Eight Companies

Paul Graham and Jessica Livingston launched Y Combinator in the summer of 2005 with a structure that had no direct precedent. The vehicle was organised as a limited liability company rather than a conventional venture fund; the model was to invest a small, fixed sum in multiple companies simultaneously, run them through a shared curriculum for roughly three months, and end with a public pitch event — Demo Day — at which the companies presented to a room of investors. The founding partners also included Robert Morris and Trevor Blackwell.

The investment terms were deliberately minimal. Each founding team received approximately $6,000 per founder, in exchange for a small equity stake — typically around six percent. The figure was calculated to cover living expenses in Cambridge for a summer, not to capitalise a business in any traditional sense. What the programme sold, beyond the cash, was access: to Graham's network, to a cohort of peers working simultaneously, and to the credibility of selection itself.

The first batch ran in Cambridge, Massachusetts — where Graham was then based — and comprised eight companies. Among the first group was Reddit, which Graham and his co-founders encouraged two applicants, Steve Huffman and Alexis Ohanian, to pivot from their original idea into a social news aggregator while their application was being evaluated. Loopt, a location-sharing application founded by Sam Altman, was also in the inaugural cohort; Altman would later return as Y Combinator's president in 2014 and chief executive in 2019. The other six companies from that first batch left a markedly thinner documentary record, a contrast that became part of the lore of early venture selection.

The programme relocated to Mountain View, California, in 2006 and adopted a twice-yearly cycle — Summer and Winter batches — that it maintained for the following decade. By the second cycle the investment amount had already begun to drift upward as competition among accelerators increased and the cost of a viable prototype changed. Y Combinator published the SAFE note — Simple Agreement for Future Equity — in 2013, replacing the earlier convertible note structure and further standardising terms across the industry.

Batch sizes grew substantially over time. By the early 2010s a typical cohort numbered in the dozens; by the mid-2010s some batches exceeded one hundred companies. The fixed-class structure created network effects that compounded: each cohort added to an alumni base that became a recruiting pool, a customer network, and a source of follow-on capital for later batches. By 2021, according to Y Combinator's own published figures, the combined valuation of its portfolio companies exceeded $400 billion.

The Demo Day format migrated across the industry. Techstars, founded in Boulder, Colorado in 2006, adopted a similar structure. Station F in Paris, which opened in 2017, housed multiple accelerator programmes that used the batch-and-demo architecture. The label "accelerator" — distinguishing the class-based model from the earlier, more informal "incubator" — entered common usage largely as a result of Y Combinator's visibility.

What the 2005 experiment established was less a financial instrument than a calendar logic: the idea that a startup's early months could be structured around a semester, a cohort, and a single high-stakes public presentation. That logic made seed investment legible as a class rather than a series of bespoke bilateral negotiations, and it compressed the timeline between idea and institutional capital in a way that reshaped what a seed round was expected to cost and produce. The eight companies in Cambridge in 2005 were modest in ambition by later standards; the container Graham built around them was not.