Google.org Impact Challenge: AI for Science — $500,000 to $3 Million Grants

Historical reference: this application round is closed. Google.org’s Impact Challenge: AI for Science accepted applications until 1 May 2026. The official page remains useful for understanding the fund’s scope and selection criteria, but it does not publish a later deadline or an open application window. Do not treat this page as an invitation to apply now.

This was a specific themed round, not an undated promise of an annual grant. Google.org described it as a $30 million global open call for scientific projects that use artificial intelligence to accelerate discovery. The programme targeted work at the intersection of AI and science, with particular attention to health and life sciences, climate resilience, and environmental science. The published page also says that selected organisations could opt into a Google.org Accelerator with six months of pro bono technical support from Google experts and access to Google Cloud credits.

The broader Google.org Impact Challenge brand is still visible on Google.org’s programme index, alongside the separate AI for Government Innovation challenge. That does not amount to an announcement of another AI for Science round. A future call could change its subject areas, award range, eligibility rules, or submission dates. Keep the closed date in the front matter so this page works as an archive entry rather than appearing to advertise a missed deadline.

AI for Science at a glance

The amount is important when assessing fit. This was not a small pilot award: the published range started at half a million dollars and reached $3 million. An applicant needed a project, team, budget, and operating structure capable of using a substantial grant responsibly. The fund size also meant that a strong application needed to explain what the money would change in scientific practice, not only describe an interesting model.

What Google.org was looking for

The official criteria set four tests. They are the most reliable guide for anyone preparing for a later, similarly structured call.

1. Scientific ambition and impact

The project had to pursue high-impact research and explain why the work could matter beyond one organisation. Google.org asked applicants to be evidence-based and to define clear, quantifiable success metrics. A general statement such as “apply AI to improve research” would not explain the scientific result. A stronger proposal would identify the question, the present limitation, the proposed method, and the measurable result that would show progress.

For a health or life-sciences project, that metric might relate to the quality of a biological prediction, the speed of a diagnostic process, or the usefulness of a model for understanding disease mechanisms. For climate or environmental work, it might concern the accuracy of a forecast, the quality of a map, the discovery of a biological relationship, or the performance of an intervention. The exact measure would depend on the science; the requirement was that the applicant name it and show how it would be assessed.

2. Innovative and responsible use of AI

AI needed to be central to the project, not a decorative feature added to a conventional research plan. The official page gives two routes to fit: the proposed solution could use AI as a core component and be shared through open-source licensing for public benefit, or the work could enable later AI use through an asset such as a foundational open dataset.

That open approach has practical consequences. Teams should decide in advance which code, model, documentation, and data products they can release. They should also identify restrictions created by privacy, consent, biosafety, intellectual property, or third-party data licences. A proposal that promises open outputs without checking whether the underlying data can legally be shared is not ready. Responsible AI alignment should be reflected in testing, documentation, access controls, and a plan for handling errors or harmful use.

3. Feasibility

The application needed a realistic execution plan, timeline, and budget. The team also had to possess the technical and domain expertise needed to do the work. For interdisciplinary research, that normally means showing who understands the scientific question, who can build and evaluate the AI system, and who can manage the data and operational work between those disciplines.

The budget should connect directly to the work plan. Explain the personnel, compute, data collection, laboratory or field work, evaluation, documentation, and maintenance costs. If the project depends on a partner, name the partner’s role and confirm what access or contribution is actually available. A large award ceiling is not a reason to request the maximum amount; the request should follow from a defensible plan.

4. Scalability and sustainability

Google.org asked applicants to show potential for impact beyond the immediate project. That could mean a method that other research groups can reuse, an open dataset that supports work in several fields, a model that can be adapted in different geographies, or a discovery that changes how organisations approach a major scientific problem.

The published criterion also asks how outputs will be discovered, adopted, and maintained across scientific domains and geographies. That makes stewardship part of the proposal. State who will maintain repositories, correct documentation, support users, version datasets, and measure adoption after the funded work ends. “Open” is not the same as “sustainable” unless another researcher can find, understand, run, and build on the result.

Focus areas

The first focus area was AI for Health & Life Sciences. Google.org described projects that decode fundamental mechanisms of life and produce foundational models, agents, open datasets, or predictive understanding of biology. The goal was scientific progress with a plausible connection to better human health, rather than a generic healthcare software product.

The official page gives three previously funded examples. Spore.Bio is building a foundational microbiological emulator that combines biophotonics with deep learning to detect antimicrobial resistance more quickly. The Technical University of Munich is integrating multi-scale biological data with an LLM interface to give physicians a spatially grounded view of cellular processes. The University of Washington is using Fiber-seq and machine learning to create high-resolution maps of the human genome and improve prediction of how genetic variation affects health.

The second focus area was AI for Climate Resilience & Environmental Science. This covered projects addressing unresolved questions about the planet’s living systems or developing new ways to preserve those systems. The examples on the official page include the Innovative Genomics Institute, which is developing an AI foundation model trained on cultivated rumen microbiomes to predict collective bacterial behaviour and identify genetic interventions related to enteric methane emissions. The Sainsbury Laboratory is using AI-guided pipelines and AlphaFold to predict disease-resistance genes from plant and pathogen genomes. The University of Liverpool is combining autonomous robotics, AI agents, and human expertise to discover porous materials for atmospheric carbon capture.

Google.org said it would consider exceptional proposals in other scientific fields when they showed strong alignment with the criteria. That is an invitation to make a well-supported case, not a reason to ignore the named priorities. A proposal outside the two focus areas would need unusually clear scientific importance, a convincing AI contribution, and a strong public-benefit argument.

Eligibility and fit

The official applicant categories were nonprofits, social enterprises, and academic institutions. The call was global. An individual researcher would therefore need to apply through an eligible institution or organisation rather than as a private individual. The same distinction matters for a commercial company: a purely commercial venture was not one of the applicant types named on the challenge page.

Good fit required more than an eligible legal structure. The project had to connect AI to a serious scientific question, have access to the necessary expertise and data, and explain a credible public benefit. A team should be cautious if its AI component only automates an administrative task, if its research question is not measurable, if it cannot share the proposed outputs, or if it has no plan for validation by domain experts.

Application steps for a future round

The 2026 window is closed, so there is no current submission to complete through this page. If Google.org announces a new round with comparable rules, use the official challenge page as the controlling source and work through this sequence:

1. Confirm the new call. Check the named theme, deadline, eligible organisation types, award range, and any new instructions. Do not carry the 2026 date or criteria into a later round without checking them.
2. Define the scientific question. State the unresolved problem, the evidence that it matters, and the result that would count as progress. Tie the question to health, life sciences, climate resilience, environmental science, or the new round’s published focus.
3. Show why AI is necessary. Describe the data, model, agent, foundation system, or computational workflow and explain what it makes possible that existing methods do not. Include an evaluation design, not just a technology description.
4. Prepare the responsible-release plan. Identify the open-source code, model, dataset, or other public asset. Resolve privacy, consent, safety, licensing, and governance issues before submission.
5. Build the work plan and budget. Assign responsibility to people with both scientific and technical expertise. Include milestones, compute and data costs, partner commitments, risk controls, and a maintenance plan.
6. Explain adoption. Name the communities that could use the result, how they will discover it, and what support or documentation will make reuse possible across institutions and countries.
7. Submit only while the new window is open. Use the application route linked from Google.org’s current official page. Save a copy of the submitted materials and confirm that the organisation’s legal and financial information matches the eligible applicant entity.

Frequently asked questions

Can I apply to the round listed here? No. The AI for Science application window closed on 1 May 2026. This page is a historical reference until Google.org publishes a new call.

Was this an open call or an invitation-only award? The official page describes it as a global open call. It names nonprofits, social enterprises, and academic institutions as applicants and describes a review process involving Google.org, subject-matter experts, and external specialists.

How much could a selected organisation receive? The published award range was $500,000 to $3 million USD, from a $30 million global initiative.

Was Accelerator participation mandatory? The official page says selected organisations had the option to participate in the Google.org Accelerator. It describes six months of dedicated pro bono support from Google experts and access to Google Cloud credits.

Could a project outside health, life sciences, climate resilience, or environmental science apply? Google.org said it remained open to exceptional proposals in other fields when they aligned strongly with all of the selection criteria. The named focus areas were still the clearest indication of priority.

What should applicants watch for next? Watch the Google.org Impact Challenge index and the official AI for Science page, but wait for a new announcement before using a new deadline, amount, or eligibility rule. Until then, the facts above describe the closed 2026 round only.

Official source

The verified programme page is Google.org Impact Challenge: AI for Science. It records the 1 May 2026 close date, $30 million fund, $500,000–$3 million award range, eligible applicant categories, focus areas, review partners, selection criteria, and Accelerator support. The Google.org Impact Challenge index confirms that AI for Science is a named challenge within the broader programme, but it does not announce a replacement application deadline.