AI Discounts and Grants for Nonprofits
A working list of what’s current: vendor discounts, training programs, and grants aimed at nonprofits and mission-driven organizations, who qualifies, and where a discount isn’t the whole answer. Last checked July 2026.
Anthropic, OpenAI, Google, and Microsoft all run nonprofit discount programs now, and several large foundations fund AI capacity work directly. Most of it goes unused, not because it’s hard to find, but because nobody’s laid out who actually qualifies and what a discount doesn’t fix on its own.
This page covers the current vendor discount programs, the grants and fellowships aimed at AI capacity-building, eligibility for each, and where a free or discounted tool is the right move versus where it creates a different problem. Terms change quickly on all of these. Confirm current pricing and eligibility directly with the vendor before applying.
Vendor discounts
Verification for most of these runs through one of two intermediaries, Goodstack or TechSoup. Getting verified with whichever one your primary vendor uses is usually a one-time, few-day process, and it opens the door to other participating vendors on the same platform.
| Program | What you get | Eligibility | Verify through |
|---|---|---|---|
| Claude for Nonprofits (Anthropic) |
Claude Team at about $8/seat/month (2-seat minimum, down from $25), up to 75% off Enterprise, plus a free AI training course and Blackbaud, Candid, and Benevity data connectors. | US 501(c)(3)s and equivalent international charitable registrations. Excludes government agencies, political organizations, and most higher education. | Goodstack Program details |
| Claude for Teachers (Anthropic) |
A free year of premium Claude access for individual teachers, plus a Learning Commons connector and standards-aligned teaching skills. | Verified US K-12 educators. Applications open through June 30, 2027. | Direct application Program details |
| OpenAI for Nonprofits | ChatGPT Business at about $8/user/month on annual billing (roughly 60% off), Enterprise up to 75% off through negotiated sales. | Registered 501(c)(3) public charities only. No academic, religious, or government organizations (a tighter line than the other vendors below). | TechSoup (US) or Goodstack / direct sales (international) Program details |
| Google Workspace for Nonprofits (Gemini + NotebookLM) |
The Gemini app, Deep Research, and NotebookLM free for up to 2,000 users, plus AI built into Gmail, Docs, and Sheets. Advanced Workspace tiers from $3.50/user/month, 70–75% off standard pricing. | US 501(c)(3)s and equivalent. Fiscally sponsored organizations without their own determination letter don’t qualify. Hospitals, schools, and universities are generally excluded, though affiliated charitable foundations sometimes qualify. | Goodstack Program details |
| Microsoft 365 Copilot for Nonprofits | $25.50/user/month on annual billing (15% off), requires a qualifying Microsoft 365 license. Copilot Chat is included free with any M365 subscription. Eligible orgs also get $2,000/year in Azure credits. | Registered nonprofits with a qualifying Microsoft 365 subscription. | TechSoup Program details |
| TechSoup | Not a vendor itself. It’s the verification hub behind 100+ discounted tools, including Canva Pro free for up to 50 seats, Notion AI at $5/user/month, 30% off Grammarly, plus discounts on Otter, Zoom, and Adobe. | Verified 501(c)(3) status or local equivalent. | techsoup.org/ai-tools |
Grants and fellowships
These fund a project or a person, not a software seat. Cycles run in months rather than the same-week turnaround of a vendor discount signup, and most expect a defined scope and a way to show results.
| Program | What it funds | Eligibility | Learn more |
|---|---|---|---|
| Claude Corps (Anthropic, $150M) |
A full-time, AI-literate fellow embedded in a host nonprofit for 12 months: an $85,000 salary plus a $10,000 grant to the organization, employed through CodePath. First cohort of 100 started October 2026; next cohorts January and August 2027. | US nonprofits able to host a full-time fellow. Fellows must be 18+, US work-authorized; no degree required. | details |
| Humanity AI ($500M, 10 foundations) |
Pooled grants across democracy, education, humanities and culture, labor, and security. Ford, MacArthur, Packard, Mozilla, Omidyar, Mellon, Kapor, Doris Duke, Lumina, and Siegel Family Endowment are the founding partners. Over $18M already committed; a $10M open call is launching in summer 2026. | Public-interest and civil-society organizations working in the five focus areas. Open-call criteria to be published. | humanityai.ai |
| Google.org AI grants | Several active programs: the Impact Challenge for AI and government partnerships ($1–3M), AI for Science ($500K–3M), and regional AI-training grants such as a $2M award to Infoxchange for nonprofit training across Asia-Pacific. | Varies by program. Most require a government, research, or intermediary partner rather than a standalone nonprofit applying alone. | google.org |
| Microsoft Elevate for Changemakers | A free AI for Nonprofits credential built with LinkedIn and NetHope, live and on-demand AI skills training, and a fellowship track for nonprofit leaders already running AI projects. | Open to nonprofit staff and leaders; no cost to participate. | details |
| People-First AI Fund (OpenAI, $50M) |
Unrestricted, general-support cash grants, no requirement to build anything AI-related. Wave 1 gave $40.5M to 208 nonprofits, an average award of about $195,000, out of nearly 3,000 applicants. A smaller, board-directed Wave 2 targets organizations already running AI work at scale in health, education, and community resilience. | US 501(c)(3) with an annual operating budget between $500,000 and $10 million. The 2026 open call closed July 15; a new round is expected to follow. | details |
| Patrick J. McGovern Foundation | AI and data-science work for public good across climate, health equity, human rights, and media. Recent grantees include ReFed, Climate Change AI, Black in AI, and the Committee to Protect Journalists. The one program open to direct application is the Data Practice Accelerator, up to $125,000 for nonprofits with complex datasets moving toward AI readiness, on an annual cycle (most recently closed July 1, 2026). | The foundation’s larger strategic grants are invitation-only, based on visibility in the AI-for-good field, not open applications. The Data Practice Accelerator is the exception. | details |
Is it worth your time?
Three of the six above are effectively closed doors for most nonprofits, not because the money isn’t real, but because of who tends to win it. Humanity AI’s first $18M went almost entirely to established policy and research organizations (AI Now Institute, the Center for Democracy and Technology, the Pulitzer Center), not direct-service nonprofits; its $10M open call this summer is the one moment a broader set of organizations could get in. The McGovern Foundation doesn’t take unsolicited proposals for its larger grants at all, and its recipients (ReFed, Climate Change AI, Conservation International) tend to already have visibility in AI-for-good circles before a grant shows up. Google.org’s Impact Challenge requires a government partner as a co-applicant, which rules out most nonprofits working independently.
OpenAI’s People-First AI Fund is the most genuinely open of the group: unrestricted money, no AI expertise required, and a stated goal of reaching community-rooted organizations rather than AI-sector insiders. It’s also the most competitive on raw numbers, about 3,000 applications for 208 grants in its first wave, roughly a 7% hit rate. Claude Corps sits in between: any nonprofit that can host a full-time employee for a year is eligible, but 400 host spots against a much larger pool of interested nonprofits means most applicants won’t get a fellow. Microsoft Elevate is the safest bet on time spent versus odds of return: it’s free, has no real selection process, and the credential and training have value on their own even without a grant attached.
Who actually qualifies
The baseline across nearly every vendor program is the same: a registered 501(c)(3), or the equivalent charitable registration outside the US. Fiscal sponsorship alone often isn’t enough. Google’s program explicitly excludes fiscally sponsored organizations without their own determination letter, and the others tend to follow the same logic even where it isn’t stated as plainly.
From there, the programs diverge. Anthropic and Google both exclude government agencies and political organizations but leave room for healthcare, environmental, and research nonprofits. OpenAI draws the tightest line of the three: registered 501(c)(3) public charities only, with academic, religious, and government organizations excluded outright (categories the other vendors don’t rule out). Higher education is treated inconsistently across the board: mostly excluded, though university-affiliated nonprofits and research centers sometimes qualify depending on how they’re incorporated.
B Corps, social enterprises, and mission-driven for-profits don’t qualify for any of the vendor discount programs above, regardless of how closely their mission aligns. That’s worth stating plainly, since it’s the assumption that trips people up most often.
Who benefits most, and when a different solution is better
A small direct-service organization running on a shared inbox and a spreadsheet gets the clearest win here. Free Gemini and NotebookLM through Google for Nonprofits, or Claude Team at $8 a seat, puts solid capability in front of a five- or ten-person staff for close to nothing. The cost is verification paperwork, not budget, and the risk is low because the day-to-day work (drafting, summarizing, research) usually doesn’t touch sensitive records.
The picture changes for organizations handling case files, safeguarding information, legal or immigration records, or anything where a leak has consequences for the people it describes. The discount is still real, but the harder question (what data is allowed into the tool, what the vendor’s terms actually say about training and retention, who owns that decision) doesn’t go away because the seat is cheap. Treating a signed-up discount as the finish line is how staff end up pasting client records into a consumer chat tool because nobody decided otherwise. An affordable tool makes that decision more urgent, not less.
For B Corps and mission-driven for-profits shut out of the programs above, the honest options are standard commercial pricing, an open-source or self-hosted model where the economics work, or restructuring a specific program under a nonprofit fiscal sponsor. That last one is a path, but it’s a legal and operational one, not a shortcut to attempt informally.
Grants like Humanity AI and Google.org’s Impact Challenges fit a narrower case: a defined project with a timeline and a way to show results, not a way to cover the AI tools budget for the year. They’re worth pursuing when there’s a specific initiative to fund, and the application cycles run in months. Claude Corps is different again: it’s a staffing model, a person embedded for a year, which suits an organization with a capacity gap and the ability to manage a full-time hire, more than one just looking for cheaper software.
Where to start
Getting the discount is the easy part.
Deciding what’s allowed, who owns that decision, and how you’d know if it broke is separate work from signing up for a discount. It’s the piece we help purpose-driven organizations get right.