Rethinking AI for Science Funding

As science stands at an inflection point, how do we organize capital to shape the future of discovery?

Photo by Google DeepMind on Unsplash

Renaissance Philanthropy and Google DeepMind are partners in the $30M Google.org Impact Challenge: AI for Science, a supercharged initiative at the intersection of AI and scientific discovery. In this article, the Google DeepMind team dives into how we can organize funding for AI for science to unlock progress rather than limit it.

We’re at a critical moment. Venture capital is straining to deliver returns at the pace it once promised. Government R&D budgets, especially for science, continue to erode. Philanthropy is struggling to strike a balance between filling the urgent gaps of the present and pursuing the moonshots needed to power the future.

Yet science — and particularly science accelerated by AI — has never had more potential. The most transformative science-based organizations emerging today aren’t “software eats the world” apps. They’re building new materials, mapping tumor biology, or synthesizing global datasets to accelerate cures.

That said, these catalytic scientific initiatives are not without risk, and they are not cheap. These companies often need $5 million -$10 million just to get to a seed-stage proof of concept. And right now, there’s no single capital source set up to support them.

So how about we stop trying to fund 21st-century science with 20th-century structures?

Science Needs Blended Finance

Venture capital can’t go it alone. At the earliest stages, science is too risky, too slow, and too expensive for traditional VC to bear by itself — especially when timelines stretch, capital intensity rises, and manufacturing challenges demand significant upfront investment. At the same time, government funding often stops where commercialization begins, leaving a gap at the very point where broad deployment might require the government’s reach and public accountability. That’s a missed opportunity.

We need a layered approach that creates new markets and adds value beyond just monetary returns. An approach that sees:

• Philanthropy and government de-risk early exploration and infrastructure, absorbing initial high risks and enabling foundational research (especially tackling fundamental “root node” problems with broad applicability). Often through grants and public investments in critical scientific infrastructure.

• Venture capital amplifies and scales validated breakthroughs, leveraging the de-risked environment to achieve necessary multiples and sustainable business growth.

• Crowdfunding or High Net Worth (HNW) capital fills gaps and aligns incentives, particularly when public interest and market upside intersect. This provides flexible capital that complements traditional sources and engages a broader community.

This isn’t a theory. It’s already happening. The blended finance model is already fueling significant breakthroughs across the scientific landscape. Examples include:

• Regenerative medicine — Longeveron. The company raised approximately $26 million from private investors while also securing around $20 million in NIH grants during its first five years. Co-founder Dr. Joshua Hare described this non-dilutive grant funding as life-saving in the early stages, perfectly illustrating how government capital can de-risk a venture and pave the way for commercial investment. Longeveron has since advanced its cell therapy through clinical trials and even reached public markets.

• Clean energy – ARPA-E: The U.S. Department of Energy’s ARPA-E program functions similarly, providing grants that de-risk novel energy technologies to make them attractive for the private sector. Since its founding in 2009, ARPA-E has invested roughly $2.9 billion in over 475 high-risk projects, which have gone on to attract more than $7.6 billion in private-sector follow-on funding. This model — government taking the early risk so industry can follow — has seeded new energy companies and entire cleantech markets.

• Life sciences – Venture philanthropy: Organizations like the Alzheimer’s Drug Discovery Foundation (ADDF) have institutionalized this model. The ADDF alone has funded over 50 neuroscience companies in the past decade, both by awarding early grants and by participating in later-stage venture rounds. “As a venture philanthropy, the ADDF seeks out and invests in the most promising, high-risk/high-reward science ... that would go underfunded without us,” explains ADDF’s Chief Philanthropy Officer. This kind of philanthropic capital not only validates nascent science but also attracts follow-on investment from VCs and pharma, creating a powerful flywheel: public and philanthropic funds test and validate high-risk, high-impact science, making it possible for venture capital to then write the larger checks needed to scale and commercialize the resulting breakthroughs.

Why Al Changes the Game

AI is making this kind of capital model more urgent and more feasible. It dramatically lowers cost and time, shortening development cycles through simulation and synthesis, which cuts down traditional long lead times. It reveals hidden patterns in scientific data, accelerating the pace of discovery. And most importantly, it’s already delivering tangible breakthroughs.

The Nobel-winning work behind AlphaFold is a landmark example. Its success was made possible by leveraging the Protein Data Bank (PDB), a global archive of 3D protein structures built over decades as a public good. DeepMind’s privately funded AI, fueled by this immense publicly funded dataset, created a tool exponentially more powerful than its inputs — a true symbiotic breakthrough. This synthesis of foundational public data and private AI innovation demonstrates a powerful new paradigm for science. What government and philanthropy provided (a huge, de-risked data repository and early research), AI amplified into a world-changing result.

The question we must ask now is: How do we intentionally replicate the AlphaFold model at scale? The clear lesson is the immense value of creating large-scale, high-quality, open-access datasets as public goods. Imagine a “PDB for genomics,” a “PDB for materials science,” or a “PDB for climate data”—foundational repositories intentionally built with public and philanthropic capital. These platforms would serve as launchpads for a new generation of AI-driven companies, creating an ecosystem where public and private innovation can be unleashed on publicly de-risked data to solve our biggest challenges.

Now is the time to operationalize that success at scale.

To make this work, we need to reduce the silos between capital types and consciously design financial architecture that optimizes for cures, not just unicorns.

Some immediate steps:

• Create shared visibility: Build a common platform showing which projects have been de-risked by philanthropy and are ready for follow-on investment by VCs and other funders. This creates a high-quality, vetted pipeline and deal flow so that promising projects don’t languish unseen after early grant successes.

• Incentivize co-funding: Structure philanthropic contributions to enable matched VC or government investment, and encourage philanthropies to co-invest on shared initiatives rather than each pursuing isolated projects. This approach maximizes leverage and impact by pooling resources (the absence of coordination today often leads every foundation to chase its own shiny pilot program instead of building common infrastructure). Government grants and public subsidies can be strategically deployed to complement private capital, creating a more robust, coordinated funding ecosystem.

• License permissively: Make open-source datasets and tools easy for startups to build on, commercially. We should ensure that publicly funded innovations (databases, research tools, algorithms) come with permissive licenses that allow entrepreneurs to adopt and scale them. Lowering the friction here means a faster transfer of knowledge from lab to market.

• Rethink outcomes: Explore financing tied to real-world milestones and performance incentives, not just valuations, as recommended by Renaissance Philanthropy’s playbooks on market-shaping and incentive prizes. Think Advanced Market Commitments (AMCs) or outcomes-based tranches, or dynamic public-private contracts that unlock capital based on tangible progress (e.g., “make the playoffs, get a new tranche of funding”). This moves beyond the “grand slams only” mentality of traditional venture and rewards hitting critical singles and doubles that save lives or the planet.

  • Examples of AMCs in Action: These have proven effective in aligning financial incentives with public-good outcomes:

    • Pneumococcal Vaccine AMC: Launched in 2009 by GAVI, the World Bank, WHO, UNICEF, five national governments, and the Gates Foundation, this AMC committed $1.5 billion to guarantee a market for pneumococcal vaccines for low-income countries. The result? Pharma companies like GSK and Pfizer stepped up to develop and scale production, ultimately accelerating vaccine distribution by five years and saving an estimated 700,000 lives by 2020. It also drove down long-term prices for everyone by ensuring volume.

    • COVID-19 Vaccines (Operation Warp Speed & COVAX AMC): During the pandemic, governments (like the US with Operation Warp Speed) and global initiatives (like COVAX) made substantial pre-purchase commitments for vaccines. These guarantees provided crucial financial certainty to pharmaceutical companies, enabling them to invest heavily in parallel R&D efforts and rapidly scale manufacturing capacity. The unprecedented result was multiple effective vaccines delivered in under a year – a timeline previously unimaginable – with billions of doses distributed worldwide in the following year.

    • Stripe’s Frontier Climate: This ongoing AMC, launched in 2022 by a consortium led by Stripe, commits nearly $1 billion through 2030 to purchase permanent carbon removal. By guaranteeing future demand for nascent carbon-capture technologies, Frontier aims to stimulate innovation and scale in an industry that barely exists today. It sends a clear signal to researchers and startups that if they build viable carbon removal solutions, a market will be there.

    • California’s Low-Cost Insulin AMC: California recently entered a 10-year, $50 million contract with nonprofit manufacturer Civica Rx to produce affordable insulin. By acting as a committed buyer of low-cost insulin, the state provides market certainty, allowing Civica Rx to invest in production infrastructure and achieve economies of scale, with the goal of reducing insulin unit costs for patients.

  • Diversify deal flow: Finally, investors should consciously diversify what “success” looks like in their portfolio. This means stratifying risk — having some bets that are moonshots, balanced by others that steadily advance core scientific infrastructure — and doubling down on a few anchor investments that serve both science and society. For example, a fund might accept a lower financial return on a platform that enables dozens of other breakthroughs (like a new genomic dataset or a battery materials testbed), because that platform multiplies downstream innovation (and yes, eventual profits). It’s a different way of thinking about ROI: return on innovation as well as investment.

This Isn’t Charity. It’s Infrastructure.

The truth is, if we want to build science companies that solve problems for people we love, we need to redesign the financial architecture around them. The current system optimizes for unicorns, not cures.

Science creates public value; often its most important outcomes (longer lifespans, a cleaner planet, a more resilient society) can’t be fully measured in dollars. Profit is not the primary goal to optimize for; it is the side effect of a successfully commercialized scientific breakthrough. When we solve a real-world problem, create a life-saving therapy, or invent a world-changing material, the market will eventually reward that value. The problem with our current financial architecture is that it often attempts to optimize for profit first, forcing companies to chase short-term financial metrics instead of focusing on the hard, long-term work of scientific discovery.

This is precisely why we have launched our second AI for Science Fund and are actively working with a coalition of forward-thinking funders to build this flywheel. We’re not arguing for less investment discipline; we’re arguing for smarter risk-sharing, longer time horizons, and clearer metrics of real-world progress that ultimately lead to both public good and sustainable financial return.

In a world where AI is increasing our capacity to ask and answer the hardest questions, our capital systems should keep up. It’s time to fund science like the future depends on it — because it does.

---

Dorothy Chou is Director of Public Engagement at Google DeepMind.

Zoë Brammer is Strategic Foresight Manager at Google DeepMind.

---

SCI PHI explores how visionary philanthropy enables the breakthrough science that will define the 21st century. In addition to publishing essays, reflections, and interviews from the team at Renaissance Philanthropy, SCI PHI also features perspectives and field notes from experts and philanthropists working to advance ambitious science.

This is the first in a series of SCI PHI articles looking at AI for Science. In late 2025, Renaissance Philanthropy worked closely with the UK government's Department for Science, Innovation & Technology on their AI for Science datasets RFP - keep an eye out for an upcoming SCI PHI piece diving into what we’ve learned through that work.

You can find more from our visionaries here. To get in touch with the team at Renaissance Philanthropy, contact us info@renphil.org.