Current AI's $400M Bid to Build a Public Web for AI
Nonprofit Current AI wants a free, public alternative to Big Tech's AI models, and it has $400 million and a chatbot to show for it so far.
6 min read
Every frontier AI model in wide use today is owned by a handful of companies in the United States and China. Current AI, a nonprofit born out of the 2025 Paris AI Action Summit, wants to change that math with public money instead of venture capital, and its pitch is blunt: build AI the way the internet's early architects built the web, open and free, before anyone gets to lock it down.
TechCrunch reported on the group's progress this week, and the numbers are the story. Current AI has $400 million committed, a working open-source chatbot built with ten partner organizations, and a hardware prototype already running in the field in India. It also has a $2.5 billion five-year target it's nowhere close to hitting.
TL;DR
- Current AI is a nonprofit founded in February 2025 to build open, public AI infrastructure as an alternative to commercial models
- CEO Ayah Bdeir, formerly of Mozilla and littleBits, says "there has to be a public alternative" to AI owned by a handful of companies
- The group has $400 million committed against a $2.5 billion five-year target, from the French government, Ford Foundation, MacArthur Foundation, Google DeepMind, and Salesforce
- Its flagship product, Alpha Chat, is an open-source chatbot assembled in seven weeks by ten partner organizations including Hugging Face and MIT Media Lab
A Different Kind of AI Company
Current AI doesn't train frontier models from scratch. It coordinates money, code, and compute that other people already have, then routes the result toward languages and communities that commercial labs tend to skip. That distinction matters more than it sounds, because it's the entire basis for comparing this project against the companies it's positioning itself against.
| Current AI | Typical frontier AI lab | |
|---|---|---|
| Structure | Independent nonprofit | For-profit or capped-profit corporation |
| Funding so far | $400 million (of $2.5B five-year target) | Tens of billions per funding round |
| Flagship product | Alpha Chat, assembled from partner contributions | Proprietary frontier model trained in-house |
| Compute disclosed | None published | Dedicated data centers, often hundreds of thousands of GPUs |
| Stated goal | Public infrastructure, no profit motive | Commercial products, investor returns |
Who Actually Funds It
The French government seeded Current AI with $100 million at launch. Ford Foundation, MacArthur Foundation, Google DeepMind, and Salesforce added to that pool, bringing the committed total to $400 million. CEO Ayah Bdeir, who joined in January 2026 after leading Mozilla's AI strategy and founding littleBits (acquired by Sphero in 2019), is careful about how she describes those backers.
Ayah Bdeir, now CEO of Current AI, previously led Mozilla's AI strategy and founded hardware startup littleBits.
Source: wikimedia.org
"There are different models and proposals for who owns data in various communities, but one thing is sure: it shouldn't be a company in Silicon Valley trying to make a select few thousand people wealthier."
Funders are "not investors; they're funders," she told TechCrunch, a distinction that matters less if two of the five named backers, DeepMind and Salesforce, are themselves commercial AI companies.
What Current AI Has Actually Shipped
Announcements from a well-funded nonprofit are easy to write. Shipped products are harder, and Current AI has two worth naming.
Alpha Chat
Alpha Chat is an open-source chatbot that ten organizations, including Hugging Face, Mozilla, and MIT Media Lab, put together in seven weeks and launched at the AI for Good Global Summit in Geneva earlier this month. Each partner contributed a piece: a language model here, safety tooling there, compute somewhere else. It's less a new model than proof that models already tracked on our open-source LLM leaderboard can be assembled into a working product without a nonprofit owning any of the underlying weights.
Suno Sutra
The more concrete project so far is Suno Sutra, a handheld, offline device built with Bhashini, the Indian government's AI language division, and unveiled at the IndiaAI Impact Summit in February 2026. It runs three models locally, a vision model, a speech-to-text and translation model, and a text-to-speech model, across all 22 official languages of India, with no internet connection required. It shares no relation to the AI music company Suno beyond the name; "Sutra" translates roughly to "chronicles" in Hindi.
Bdeir frames the language gap as the core problem Big Tech has no incentive to fix. "In India, there are hundreds of different languages and dialects, and right now AI is not representing them," she said. "Big tech builds multilingual models to expand their market, regardless of consent or context."
Current AI has also struck a partnership with Tokyo-based Sakana AI, the startup behind evolutionary model-merging techniques, to build shared open-source infrastructure aimed at Japanese-language and Global South communities that mainstream model releases tend to treat as an afterthought. Separately, a $3.2 million grant round funded four organizations working on AI and cultural preservation in Kenya, Lebanon, and Brazil.
Current AI's homepage frames its work as a "global movement" for AI built in service of people, not profit.
Source: currentai.org
What It Does Not Tell You
The TechCrunch piece and Current AI's own materials are light on the parts that would let an outsider judge whether this scales. Three gaps stand out.
The funding gap is enormous. $400 million is 16% of the $2.5 billion five-year target, and the organization has not said what happens if the rest doesn't appear. Governments and foundations can walk away from multi-year pledges more easily than a venture-backed startup can walk away from a term sheet.
Compute is the part nobody mentions. Alpha Chat exists because partners donated pieces of an already-built stack, not because Current AI operates its own training infrastructure. Our open-source vs proprietary comparison shows the quality gap between open and closed models has narrowed fast, but the compute and money gap hasn't. Current AI is coordinating a resource it doesn't own and can't expand on its own schedule.
The funders include the companies it's positioned against. Google DeepMind and Salesforce sit on the list of backers for an organization whose CEO argues AI shouldn't enrich "a select few thousand people" in Silicon Valley. Current AI hasn't addressed that tension directly in public remarks, and neither TechCrunch nor the group's own site raises it.
None of that makes the mission dishonest. It makes it unproven, which is a different thing and one Current AI itself would probably concede: the $2.5 billion target implies the organization knows $400 million buys coordination, not competition.
Public-interest infrastructure built this way has one real advantage over anything a lab like OpenAI or Anthropic can offer: nobody has to trust a shareholder-driven roadmap to use it. Whether that advantage survives contact with a funding shortfall, or with partners whose day jobs are building the very systems Current AI says need a public alternative, is the question worth watching over the next year, not the $400 million headline number.
Sources:
