New Nonprofit Launches with $500M to Modernize Science for AI Era

Science is at an inflection point. Artificial intelligence tools like AlphaFold have demonstrated what becomes possible when rigorous data infrastructure meets machine learning — a single system predicting the structures of over 200 million proteins, earning its creators the 2024 Nobel Prize in Chemistry [1]. But behind that success lies an inconvenient truth: most of the scientific ecosystem lacks the digital plumbing to replicate it. Now, a wave of philanthropic money — led by a new $500 million nonprofit called Radial — is flooding into the unsexy but essential work of modernizing how science generates, shares, and builds upon its data.

The Launch of Radial

On March 11, 2026, scientist and philanthropist Seemay Chou announced the creation of Radial, a new division within her AI-focused Astera Institute, with at least $500 million in committed funding over the next decade [2]. The initiative is led by CEO Becky Pferdehirt, a former investing partner at Andreessen Horowitz's Bio + Health fund who holds a PhD in Molecular and Cellular Biology from UC Berkeley and previously worked as an R&D scientist at Amgen and completed postdoctoral studies at Genentech [3].

Radial's mission is deliberately unsexy. Pferdehirt described the organization's focus as building the "unglamorous, unsexy infrastructure and tools" needed to change how scientific data are generated, shared, and built upon [2]. In practice, this means funding projects that traditional grantmakers routinely overlook — database architectures, data formatting standards, open-source analytical pipelines, and the connective tissue that allows disparate research communities to build on each other's work.

"We'll never see the value of AI fully" until these foundational systems are updated, Chou told STAT News [2]. The argument is straightforward: AI models are only as good as the data they train on, and most scientific data today exists in fragmented, poorly formatted, or inaccessible silos.

The Protein Data Bank Lesson

Radial's thesis is grounded in one of AI's most celebrated scientific achievements. Google DeepMind's AlphaFold system was trained on over 170,000 protein structures deposited in the Protein Data Bank (PDB), a publicly accessible repository built and maintained over decades through painstaking, largely unglamorous curation work [4]. Without that database, AlphaFold could not have existed.

"Public data were essential to the development of AlphaFold," DeepMind has acknowledged. "The careful curation of such large data resources, representing the collective output of an entire subfield of biology, is exactly what enables our machine learning models to generalise well" [4].

One of Radial's flagship initiatives, the DiffUSE project, aims to create the next-generation structural biology database — one focused not on static protein structures, as the PDB captured, but on protein dynamics [5]. The three-year, $5 million collaborative effort involving Vanderbilt University, Cornell University, and the University of California, San Francisco, seeks to make diffuse X-ray scattering a routine tool for understanding protein motion, which is fundamental to drug discovery and basic biology [5]. The project operates as a distributed team experimenting with open publication practices that bypass traditional journals, sharing methods and findings as early as possible through logbooks and direct scientific dispatches [6].

The Crypto-Science Power Couple Behind It All

Radial's funding flows from a sprawling philanthropic network built by Chou and her partner, Jed McCaleb, the blockchain pioneer who co-founded Ripple, Stellar, and the early cryptocurrency exchange Mt. Gox [7]. Forbes estimated McCaleb's net worth at approximately $2.9 billion as of April 2025 [7]. Together, the couple has committed to giving away $3.5 billion through the Giving Pledge [3].

Their philanthropic apparatus is extensive. The Navigation Fund, legally known as the Navigation Charitable Fund, holds a $1.3 billion endowment and focuses on scientific research, farm animal welfare, criminal justice, climate change, and digital sentience [7]. It distributed $36.8 million in grants in 2024, with expectations to potentially double that in 2025 [7]. In a notable move, the Navigation Fund in late 2023 used roughly $500 million to purchase NVIDIA GPUs and establish Voltage Park, a for-profit AI cloud computing company whose profits flow back to the fund [7].

Chou also co-founded and serves as CEO of Arcadia Science, a research organization that has deliberately abandoned journal-based publishing in favor of a fully scientist-led open publishing system [8]. At Arcadia, researchers stopped designing experiments for journals and began sharing all results — including failed attempts and abandoned inquiries — through an open platform called the Icebox [8]. Chou has written that traditional publishing "is failing scientists and the public alike, costing the global community $10–25 billion per year in subscription and processing fees" [8].

A Broader Reorganization at Astera

Radial's launch comes amid a broader strategic reorganization at the Astera Institute, which McCaleb founded in August 2020 as a 501(c)(3) nonprofit dedicated to developing high-leverage technologies [9]. In February 2026, Astera published a strategy document titled "Sharpening Astera's Focus," narrowing its work to two pillars: intelligence (both biological and artificial, including neuro and AGI research) and AI-enabled life sciences [9].

The restructuring separates Astera's foundation from its technical divisions. Each division now operates semi-independently, "like start-ups," with leaders who hold "CEO-like authority" [9]. The foundation handles shared operational infrastructure while technical teams pursue their missions with significant autonomy.

Astera's 2026 Residency Program reflects this ambition. Applications opened in early 2026 for 12-to-18-month funded engagements offering salaries between $125,000 and $250,000, with project budgets of up to $1.5 million per resident, focused on neuro, AI, and life sciences [9].

Part of a Billion-Dollar Philanthropic Wave

Radial does not operate in isolation. It emerges within a remarkable surge of philanthropic investment aimed at retooling science for the AI era.

Source: STAT News / HHMI / FFST / MacArthur Foundation

Data as of Mar 12, 2026CSV

The Howard Hughes Medical Institute announced AI@HHMI in August 2024, committing $500 million over 10 years to embed AI throughout the scientific process across its network of over 300 affiliated laboratories [10]. Centered at the Janelia Research Campus in Ashburn, Virginia, the initiative pioneered an "AI-in-the-Loop" model that places artificial intelligence at the heart of experiment design, data collection, and model building. HHMI President Erin O'Shea described the goal as "bringing human curiosity and artificial intelligence closer together at every phase of experimentation and data collection" [10].

The Fund for Science and Technology, created from the estate of Microsoft co-founder Paul G. Allen, announced plans to deploy at least $500 million over four years across bioscience, environment, and artificial intelligence [11]. Its inaugural grants in 2025 included $15 million to Seattle-area scientific centers of excellence, with investments in cellular immunotherapy trials at Seattle Children's Hospital and immune system research at Fred Hutchinson Cancer Center [11].

Meanwhile, the Humanity AI initiative brought together 10 major foundations — including the MacArthur Foundation and the Mellon Foundation — to commit $500 million over five years to ensure people have a stake in AI's future [12]. And across the for-profit sector, AI captured close to 50 percent of all global venture funding in 2025 — $202.3 billion, a more than 75 percent increase year-over-year [13].

The Federal Funding Crisis

This philanthropic surge arrives against a backdrop of severe federal funding cuts to science. The White House's FY2026 budget proposal called for slashing NSF funding by approximately 57 percent — from $9.06 billion to $3.9 billion — and cutting NIH funding by 41.9 percent, from an estimated $47.3 billion to $27.5 billion [14]. In 2025, the administration froze or ended about 5,300 NIH and NSF research grants totaling over $5 billion in unspent funds [14]. More than 25,000 people departed science agencies, many at early career stages [14].

Source: FRED / Bureau of Economic Analysis (Series Y694RC1Q027SBEA)

Data as of Mar 12, 2026CSV

Congress has pushed back. The Senate Appropriations Committee's proposed FY2026 bills reduce NSF spending by only 0.7 percent and would give NIH a $400 million raise, in direct defiance of the White House's proposed cuts [14]. But even if Congress preserves existing federal funding levels, the gap between what public science receives and what it needs to compete in the AI era continues to widen — creating a vacuum that philanthropies like Radial are stepping in to fill.

The Infrastructure Gap

The core problem Radial seeks to address is structural. Modern AI systems require vast quantities of high-quality, well-formatted, accessible data. Yet scientific research has historically rewarded novelty and publication over data curation and infrastructure maintenance. Faculty members who invest time in making software robust and data properly described often receive no credit toward tenure or promotion [15]. Data sharing mandates exist but are hampered by poor formatting, proprietary software restrictions, financial paywalls, and a cultural reluctance to release data that researchers fear losing control over [15].

This creates what researchers call the "infrastructure debt" of science — decades of accumulated data locked in formats that AI cannot easily process, scattered across institutional silos, and maintained (if at all) by underfunded teams with no long-term support.

Radial's bet is that solving these infrastructure problems will yield outsized returns, much as the Protein Data Bank's decades of unglamorous curation work made AlphaFold possible. The question is whether a private nonprofit, however well-funded, can accomplish what the scientific community's own institutions have struggled to do for decades.

Skepticism and Open Questions

Not everyone in the scientific community views this surge of private science philanthropy uncritically. A 2023 investigation by the Associated Press and The Philanthropy found that "quick grants from tech billionaires" aiming to speed up science research drew pushback from scientists who questioned whether philanthropists' priorities aligned with the most pressing scientific needs [16]. Critics worry that billionaire-driven science risks tilting research agendas toward the interests of funders rather than the broader public good.

There are also governance questions. The Navigation Fund's creation of Voltage Park — using nonprofit assets to purchase $500 million in NVIDIA GPUs for a for-profit AI compute business — illustrates how these structures can blur the line between philanthropy and commercial enterprise [7]. While Voltage Park's profits flow back to the fund, the arrangement raises questions about whether charitable resources are being deployed for their stated purposes.

For Radial specifically, the challenge will be demonstrating that infrastructure investment can produce the kind of measurable outcomes that justify half a billion dollars. Scientific databases and data standards are inherently collective goods — their value is diffuse and long-term, accruing across entire fields rather than to any single laboratory or organization.

What Comes Next

The launch of Radial represents a concrete experiment in a thesis that has been gaining momentum across the scientific establishment: that the era of AI demands not just new algorithms, but new institutional plumbing. Whether the experiment succeeds may depend less on the money involved — $500 million is substantial but modest against the scale of global research — and more on whether Radial can catalyze changes in culture and incentive structures that have resisted reform for decades.

Chou and Pferdehirt are betting that by building the databases, tools, and data standards that AI needs, they can prove that the most transformative investments in science are often the least glamorous ones. If the Protein Data Bank and AlphaFold are any guide, they may be right.