The move will become official July 1; arXiv headquarters will remain in Cornell Tech’s Tata Innovation Center. The search is on for an inaugural CEO as well as a board of directors.

“I laud the effort that Cornell has invested in arXiv for the past 25 years, enabling its growth,” said Ginsparg, professor of information science and physics in the Cornell Ann S. Bowers College of Computing and Information Science, and in the College of Arts and Sciences. “I would like to ensure that the memory of that effort remains part of the permanent arXiv history.”

Since it started, arXiv has been a free service for researchers around the world to share their work with the community, and to see what their peers are doing without waiting for potentially lengthy peer-review processes, or paying for academic journal subscriptions.

Among the site’s users are scientists, both in the U.S. and abroad, for whom paywalls can be prohibitive. In a 2022 story marking arXiv’s 2 millionth submission, Tara Holm, professor of mathematics in the College of Arts and Sciences, noted the scale of arXiv’s global influence.

“These 2 million submissions represent 2 million opportunities for humanity to push forward the frontiers of our understanding,” she said. “arXiv has transformed mathematics.”

Making foundational research more accessible

Ramin Zabih, arXiv’s faculty director and professor of computer science at Cornell Tech and Cornell Bowers, said arXiv has had a key role in many disciplines.

A foundational high-energy physics paper, “The Large N Limit of Superconformal Field Theories and Supergravity,” from 1997, was introduced to the world on arXiv by its author, then-Harvard University professor Juan Maldacena, now at the Institute for Advanced Study in Princeton, New Jersey.

In 2002, when Russian mathematician Grigori Perelman solved the Poincaré conjecture, for which he won the $1 million Millennium Prize in 2010, he published the papers on arXiv instead of submitting them for formal peer review. And in computer science, the DeepSeek papers that substantially changed the AI landscape in 2025 were published on arXiv.

Material is not peer-reviewed by arXiv; a submission’s contents are the responsibility of the submitter but are subject to a moderation process that assesses the material’s topicality to the subject area, as well as its scholarly value.

Ginsparg said arXiv’s greatest attribute – aside from the fact that “it doesn’t have advertisements” – is how it facilitates “global low-barrier research, providing access to cutting-edge research, in principle, for anyone with an internet connection, a prime feature it has retained since 1991.”

He recalled that, following the fall of the Berlin Wall and the end of the Soviet Union, when several countries in Eastern Europe were in turmoil, researchers “were already telling me that it’s so wonderful that, with a telephone and an acoustically coupled modem, they can still get access to research.”

In 1991, arXiv was an email-based service housed at Ginsparg’s Los Alamos lab workstation, on a cluttered desk using a NeXTstation (forerunner to Apple’s MacIntosh). It was never meant to be a fixture for science dissemination and discovery.

“It was supposed to be a little one-off, a quick hack, and maybe once a month I’d have to intervene and do something manually,” he said.

It has grown in ways Ginsparg could not have imagined, and now requires a staff of approximately two dozen across the U.S. and around the world to keep it running.

At Cornell, arXiv was hosted by Cornell University Library before being transferred to the Faculty of Computing and Information Science (now Cornell Bowers) in January 2019. When Morrisett, then the dean of Cornell Bowers, moved to Roosevelt Island in the summer of ’19, he took arXiv with him.

Three million submissions and counting

Submissions to arXiv have grown dramatically over the years. As Ginsparg noted in when it hit 2 million submissions: “It took 23 ½ years to get 1 million submissions, seven more years to get to 2 million submissions, and perhaps will be another 4 ½ years to get to 3 million.” It passed that milestone in April, just four years later.

“In multiple fields of science,” Zabih said, “arXiv has become an indispensable resource. The delays associated with traditional scientific publishing, which range from months to years, would severely handicap these fast-moving fields. Many scientific communities benefit enormously from the rapid turnaround arXiv provides.”

The platform has averaged around 27,500 submissions per month since the beginning of 2026. Gifts from the Simons Foundation and Schmidt Sciences, totaling $10 million, are helping arXiv finish migration to cloud infrastructure and modernizing its code. 

But the long-standing financial and logistical realities of maintaining a massive repository of scholarly work in a range of disciplines – including physics; mathematics; computer science; quantitative biology; quantitative finance; statistics; electrical engineering and systems science; and economics – made it clear that Cornell Tech could no longer be arXiv’s sole custodian.

“There was a strong sense that, as a separate organization, we might attract more funding than if it’s tied to a particular university,” Morrisett said.

The transition is happening with major financial backing from the Simons Foundation, a private foundation dedicated to advancing research in mathematics and basic science, and a long-time supporter of arXiv.

“The Simons Foundation and leadership here started thinking about it, and there were a number of things that needed to happen,” Morrisett said. “One was we needed to secure more stable funding. We’ve been running a deficit the last couple years; Cornell Tech was funding that, and we were happy to do so, but couldn’t afford it in the long run.”

Morrisett also said that arXiv needed to be more nimble when it came to hiring – and paying – staff than was possible under the Cornell umbrella. “It’s all about trying to make sure arXiv lasts as long as possible,” he said, “and serves its mission as long as possible.”

There remain significant near-term challenges to arXiv, with the advent of AI and large language models, Ginsparg said.

“Recent developments in AI have great promise but also pose an existential threat to the underlying arXiv methodology, relying as it does on the bonds of human-to-human trust,” he said. “It’s now difficult to prepare for the world three months from now if the median LLM-produced computer science paper is better than that produced by the median grad student.”