# Representing Public AI at Global Tech Summit 2025 Innovation Dialogue India
###### *This article was written with AI assistance and reviewed and verified by the author.*
--
On December 10th and 11th of 2025, I had the pleasure to attend and represent Public at the Global Tech Summit 2025 Innovation Dialogue, Closed doors event.
Here's what I gathered;
### Public AI: Inference Utlity
### What is Public AI and why Public AI?
Like Electricity, AI is aimed to become a utlity in the coming years as AI adoption increases world wide.

Public AI, supports Open Source AI community with AI Infrastructure as Public goods.

The Approach Pubic AI takes is of "Airbus for AI" and currently they run inference for Soverign AI initiatives by Nations

Examples of public AI include sovereign models like Apertus and SEA-LION, which are built by public institutions such as universities and research centers. These models are released under open licenses, allowing them to be freely used, modified, and shared.

Public AI aligns with the principles of openness, collaboration, and the common good. It encourages the development of AI that is fair, transparent, and beneficial to all, rather than a select few.

### What are Inference requirements for India's AI mission
The inference requirements for India's AI mission emphasize achieving widespread **Public Access** to computational resources through a **local, grounds-up public infrastructure** model focused on affordability and decentralization.

Key requirements and strategies for AI inference within the mission include:

### 1. Affordability and Scale
The foundational requirement is extreme affordability to enable population-scale adoption. India's objective is to make AI work by sustaining a cost of approximately **one rupee per inference** to reach a billion people. This goal is crucial for expanding AI usage in diverse sectors like education, healthcare, and agriculture.

### 2. Decentralized Public Infrastructure
India is strategically pursuing a public infrastructure model for compute provision that is distinctly different from centralized facilities, largely pivoting toward a flexible, market-focused approach that subsidizes purchases rather than relying on direct government provision.

* **Open Cloud Compute (OCC):** This model is classified as a decentralized provision approach, focusing on creating **networks of smaller facilities distributed across regions**.
* **Micro and Edge Data Centers:** The model strongly supports the development of **micro and edge data centers** as a highly sustainable approach suitable for India's diverse conditions and population density.
* **Open Networks:** The OCC model aims to facilitate a shift from monolithic computing systems to **open-network systems**. This approach eliminates restrictions on data portability, which is considered essential for promoting innovation among startups and MSMEs (Micro, Small, and Medium Enterprises).

### 3. Compute Resource Democratization
The IndiaAI mission is designed to democratize access to the "ingredients" required for AI development, which includes inference capability.

* **GPU Capacity:** The mission ensures the availability of necessary computational power by coordinating the procurement of GPU capacity in the country. This effort includes arrangements to source **over 18,000 GPUs** from bidding companies and providing adequate **funding support for the compute** needed for model training.
* **Ecosystem Accessibility:** Beyond hardware, the mission aims to make **compute, datasets, algorithms, and models accessible** to foster active stakeholder participation in developing AI-based solutions, rather than limiting the public to being mere consumers.

By integrating these requirements with the existing digital infrastructure principles (DPI), India seeks to position itself as the **global capital of AI use cases**.

## India's AI Mission and Its Four Broad Objectives

#### **Objective 1 — Democratization of AI**

Making AI accessible globally, especially to countries and communities currently left out.
This is about **bridging the AI divide**.

#### **Objective 2 — Indigenous & Local AI Development**

Growing recognition that:

* Countries need **their own local AI ecosystems**
* Even within countries, solutions must reflect **local needs, data, culture, languages, constraints**

Localized AI = sovereignty + relevance + resilience.

#### **Objective 3 — AI for Good (Real-World Citizen Impact)**

Ensuring AI benefits every citizen, across high-impact sectors:

* Healthcare
* Agriculture
* Governance
* Education

This aligns with the “impact” theme the chairs mentioned earlier.

#### **Objective 4 — Local Standards, Global Governance Models**

By democratizing AI and building local solutions, we must also:

* Bring local standards to the table
* Shape global governance
* Ensure innovation remains open, responsible, and globally inclusive
* Empower *all voices* in the global AI discourse — not just major powers or big tech

This is a **global justice and representation** component.

### India vs Singapore vs Swiss

The IndiaAI Mission is a government-led initiative aimed at scaling AI in India through a series of investments and public-private partnerships across the AI technology stack. The mission's overarching strategic vision emphasizes democratic, inclusive, human-centric, and low-cost AI, aiming to maximize impact and productivity for India and the Global South.

## India's Sovereign AI Strategy: Pillars and Approach

India's approach, often articulated as building **"Digital Public Intelligence,"** is grounded in its decade-long success with Digital Public Infrastructure (DPI) like Aadhaar and UPI, which accelerates AI adoption and minimizes the gap between global AI developments and India's progress.

Key elements of India's AI mission and sovereign strategy include:

1. **AI for Impact and Inclusivity:** The strategy prioritizes the deployment of AI to solve **citizen-scale problems** and achieve social impact, particularly in sectors like **healthcare, education, and agriculture**.
* This includes ensuring AI systems are **bias-free** and that citizens are active **stakeholders and partners** in developing solutions, rather than just consumers.
* The forthcoming AI Summit, the **AI Impact Summit 2026**, focuses on demonstrating tangible results and showing AI's potential to improve the productivity of India's vast workforce.

2. **Sovereign Infrastructure and Compute:** The mission recognizes compute access as a primary requirement for AI development.
* The IndiaAI initiative focuses on expanding access to over **18,000 Graphics Processing Units (GPUs)** sourced from ten bidding companies, which retain ownership but provide discounted rates to IndiaAI affiliates.
* The vision is to establish sovereign infrastructure capabilities, recognizing the importance of **Indian IT, regulations, and privacy frameworks** within the AI ecosystem.
* It also includes supporting the development of **indigenous AI models** trained on indigenous languages that are often neglected by larger private actors.

3. **Open Source and Accessibility:** India advocates for making **compute, datasets, algorithms, and models accessible**, viewing open source as the potential way forward to broaden AI diffusion.
* The strategy supports projects like **Bhashini**, a public digital infrastructure API platform for Indian languages, which enables accessibility through translation and voice commands, aiming to achieve population-scale adoption by moving away from English/Hindi-only interactions.
* The target operating cost for widespread AI adoption in India is extremely low: about **one rupee per inference**.

4. **Policy and Governance:** India promotes **innovation over regulation** (avoiding measures that throttle opportunity) while simultaneously ensuring systems are built safely, securely, and responsibly. The government seeks to act as an enabler and orchestrator, supporting the private sector's drive for innovation.

## Differentiation from Swiss and Singaporean Approaches

While all three nations are actively building "public AI" or "sovereign AI" resources, their strategies differ significantly in scale, focus, financing, and technological reliance:

| Feature | India's Approach (IndiaAI Mission) | Swiss Approach (Swiss AI Initiative/Apertus) | Singapore's Approach (AI Singapore/SEA-LION) |
| :--- | :--- | :--- | :--- |
| **Model** | Outsourced Provision / Networked Collaboration / Innovation Competition | Networked Collaboration / Full-Stack Public AI (Highest Level of Publicness) | Networked Collaboration / Outsourced Provision / Public Options |
| **Primary Goal** | Impact, Mass Adoption, Economic Prosperity, Bridging Global South Divide, Low-Cost AI | **Sovereignty, Transparency, Diversity** (Multilingualism), Public Good, Building AI Expertise | Establishing a Global Leader Role, **Domain/Language Specificity**, Improving Public Services |
| **Compute Strategy** | Government subsidizes the **use** of commercially owned GPUs (18,000+ units) at aggressively low rates. Aims for low cost ($\sim$1 rupee/inference). | Leverages the world’s most AI-capable supercomputer, committing **over 10 million GPU hours** for foundation models for science, health, and education. | Relies on the subsidized National Supercomputing Center (NSCC); partners with Nvidia to expand capabilities; 80% allocated to publicly funded research. |
| **Model Development** | Supports development of indigenous LLMs (via a competitive bidding process) and public dataset platforms. | Released **Apertus**, a fully open-source, transparent, multilingual LLM. Trained on **15 trillion tokens** across over 1,000 languages, including underrepresented ones like Swiss German. | Built the **SEA-LION** family of open-source LLMs specifically for Southeast Asian languages, trained on local linguistic data. |
| **Transparency & Openness** | Advocates for **open source** as the path to democratization. | Sets the highest bar for **full transparency**, making architecture, weights, training data, and methods fully documented and accessible. | Models are open-source and focus on local business/language needs. |
| **Global Positioning** | Leader of the **Global South** (AI for All/Impact). | Focuses on shared scientific projects and leveraging **Swiss values** (transparency/multilinguality) for global governance. | Focuses on **regional leadership** (Southeast Asia) and global partnerships, acting as a neutral hub. |

The Swiss approach (Apertus) represents the **highest degree of "publicness"** on the gradient of public AI, as it is built with **public data, public models, and public computing infrastructure** (MareNostrum 5 supercomputer or LUMI/EuroHPC). This model aims to be free from commercial dependencies and prioritizes transparency and democratic control.

India’s strategy, by contrast, relies on a massive **scale-through-market** mechanism, subsidizing private computing services to ensure low-cost access, reflecting a "preference for subsidizing purchases rather than direct provision". This model emphasizes the rapid application of AI for societal transformation, building on existing digital foundations.