NSF X-Labs Brings In Ginkgo and Foundations: A $5M Cloud Lab for Winners

Imagine this: you have assembled a strong team and won a grant, but half the money goes to buying equipment, renting space and negotiating access to data. This is exactly the problem the US National Science Foundation (NSF) is trying to solve with its X-Labs program. In October, the foundation announced that partners are joining the program to cover part of the infrastructure costs.

What X-Labs Is and What the Partners Added

The program started in May. The idea is to give large, high-risk grants not to traditional university labs, but to independent organizations created for a specific goal and a limited time. These structures are called FROs (focused research organizations).

In October, NSF named philanthropic foundations, companies and investment funds as "consortium partners." They will provide winners with lab space, software, shared equipment, data access and expert consulting. The partners receive no equity in the projects.

The scheme has two stages. The first winners will be announced in November or December. Each will receive $1.5 million for nine months of planning. Those who show ambitious results will be able to compete for multi-year grants of $50 million.

There are five themes: instruments for sensing and imaging, photonic chips for quantum systems, AI for physical devices, decoding biological sequences, and the physical limits of computation.

What Is Actually Offered and Why It Matters to AI Teams

The most concrete offer comes from Ginkgo Bioworks of Boston: up to $5 million in access to a "cloud lab." Experiments there are run remotely, much like renting cloud servers instead of owning a data center. The company will also provide lab space and internal datasets that can be used to train AI models. For teams building AI on biological data, this means a tangible saving of time and money.

The Sloan Foundation has already supported several FROs and is now interested in how the environment in which researchers work is structured. Investor Matt Chimaglia, associated with partners Occam Foundry and the Florida Quantum Center, explains the point this way: there is no need to reinvent the wheel and spend half of a grant on basic infrastructure.

Ginkgo CEO Jason Kelly goes further. In his words, X-Labs could "pull platforms into existence," as the Human Genome Project once did. And since modern companies are far larger than they were in the days of Vannevar Bush in the 1940s, leaving them out of government-funded science is, in his view, foolish.

An Example: Self-Learning AI on Mining Machinery

Natalie Risso, an engineer at the University of Arizona, works on automating mining equipment. She is applying under the "AI for physical systems" theme and wants to test self-learning control of heavy machinery.

Her problem will be familiar to many who have worked with AI in the real world. A university grant is not enough for such experiments. An industrial partner is unwilling to risk expensive equipment: according to Risso, mines won't let her use their processing plants because they are too expensive. X-Labs is meant to close precisely this gap between the lab and real production.

Limitations and Open Questions

The program demands autonomy. The project leader must work on it full-time and make hiring and spending decisions independently of the university or parent company. Technology policy historian Arthur Demmrich calls this approach "gig work for science": it will suit some people and some tasks, but not others.

There is also the question of public benefit. Former NASA manager Zachary Pirtle asks whether such teams need economists and sociologists. There is no answer yet.

What the material leaves out matters too. The winners have not yet been announced. Partner support is not the same as money, and the terms of access to resources have not been disclosed separately. For an ordinary developer of applied software, the program changes little directly.

What to Do About It

If you are a deep tech startup founder or lead a research team in AI for physical systems, biological data or computing hardware, it is worth studying the program's themes and autonomy requirements closely. The main shift here is organizational: NSF is ready to fund a new independent organization rather than an individual piece of scientific work. As Joshua Greenberg of the Sloan Foundation puts it, this is more like founding a company than submitting a grant application.

For everyone else, X-Labs is a signal of how the US government wants to work with private companies in science. Expect the list of winners in the coming weeks: it will show which AI directions NSF considers worthy of such bets.