Can the EU and India build AI together?

EU–India cooperation on AI is taking the form of diplomatic engagement, joint research projects, and company-level collaboration. The potential of this cooperation lies mostly in applied systems, such as the development of multilingual tools, industrial AI, and AI safety.

Key takeaways:

In June 2026, more than 100 European and Indian technology companies gathered in New Delhi for the first EU–India Tech Business Forum. The meeting covered artificial intelligence, semiconductors, cybersecurity, data governance, and digital public infrastructure. This forum signals growing interest in cooperation on emerging technologies on both sides, with artificial intelligence (AI) as one of the most notable areas.

Yet the emerging EU-India partnership on AI should not be read as a belated effort to compete in the US-China frontier-model race. Rather, both parties are focused on developing trusted, affordable, multilingual, and sector-specific AI systems for practical deployment. This approach is relevant to most countries, as only a few will develop frontier models, while most need to implement AI across public services, healthcare, industry, energy, agriculture, and disaster management.

During the second EU-India Trade and Technology Council meeting in New Delhi in February 2025, both parties agreed to enhance cooperation between the European AI Office and the IndiaAI Mission across large language models, AI for human development, responsible AI, semiconductors, and high-performance computing. The conclusion of the EU-India free trade agreement negotiations in January 2026 adds an economic framework to facilitate deeper market access, investment, and business collaboration. Additionally, in January 2024, the EU and India signed a semiconductor Memorandum of Understanding to reinforce supply chains, foster innovation, and enhance resilience.

This EU-level framework is further supported by bilateral initiatives with individual member states. France and India have signed an AI declaration encompassing industrial partnerships, research on large language models and the social consequences of AI, and AI norms and standards. Germany and India have established an AI pact focused on collaboration in industrial AI across sectors, the development of responsible AI, and the exchange of experience on the EU AI Act, as well as research and talent exchanges, AI for governance efficiency, and public services. Italy and India’s 2025–2029 strategic action plan emphasizes expanding collaboration in artificial intelligence. Sweden and India have launched the Sweden-India Technology and Artificial Intelligence corridor to connect government, industry, startups, and academia, especially in AI deployment and applications. Finland and India have identified AI, high-performance and quantum computing as priority areas for cooperation in their joint statement. Similarly, the joint statement for the Czechia-India Strategic Partnership on Innovation also lists AI as one of the key research priorities. Collectively, these efforts show a broader European interest in collaboration with India rather than a single Brussels–New Delhi channel.

Employing leading AI, while developing domestic capabilities

India’s AI strategy follows a dual-track approach. On one side, Indian firms and public institutions adopt leading global AI models, including American and open-source Chinese systems, wherever appropriate. Simultaneously, New Delhi is developing domestic capabilities through the IndiaAI Mission, which integrates public-private computing resources, develops indigenous large, multimodal, and domain-specific models, establishes a dataset platform for start-ups and researchers, encourages AI applications in state institutions, increases AI courses and labs at universities, provides startup financing, and supports the development of safe AI tools.

Computing infrastructure constitutes a critical component of India’s AI strategy. The IndiaAI initiative was initially structured around a public AI compute infrastructure comprising 10,000 or more GPUs. Subsequent official updates indicate that India has now onboarded over 38,000 graphics processing units (GPUs), most likely a combination of various Nvidia, AMD, Intel, AWS, and Google chips. Although this does not position India as a frontier-compute superpower, given that a single leading US private cluster may utilize tens or hundreds of thousands of advanced chips, it does establish a domestic experimentation environment where Indian companies can train, fine-tune, and deploy systems tailored to local requirements. For comparison, OpenAI’s single Stargate facility in Abilene, Texas, is planned to contain 450,000 NVIDIA B200 chips.

There are multiple companies already providing locally developed large language models (LLMs), such as Sarvam AI, BharatGen and Krutrim, all marketed as supporting all 22 official languages of India. Under the IndiaAI Mission, Sarvam AI released 30B and 105B open-source models, trained entirely in India using IndiaAI compute. The model family also includes tools for speech, translation, and document understanding. BharatGen, a government-supported initiative led by an academic consortium, has introduced a multilingual 17B Param2 model, intended for public-sector applications. Krutrim has developed a multilingual model and cloud platform for Indian developers and businesses, including Krutrim-2, a 12-billion-parameter model supporting English and Indian languages.

While these models are smaller than globally dominant LLMs, whose parameter counts are usually undisclosed but are estimated in the hundreds of billions or more, the Indian models are more locally specialized, with an emphasis on Indian languages, domestic use cases, and deployment within India’s AI infrastructure.

However, India’s AI capabilities continue to face significant gaps, underscoring the importance of international collaboration to achieve its AI goals. India ranks third in the world for AI research output, but only eighth for patents and fifteenth for citation impact. Only 16% of Indian AI papers have a foreign co-author, the lowest share among the top ten research nations. The country spends only about 0.6% of GDP on research and development, compared with 3-4% in leading innovation economies, and lacks private investment in AI. Top-tier researchers remain scarce, while India’s GitHub developer base is on track to become the world’s largest by 2028. Annotated Indian-language data is thin. Consequently, more research cooperation, data, and advanced compute are precisely where collaboration with European countries can contribute.

What Europe brings

Europe’s value to India is often seen as mainly regulatory, but that view is too limited. Whereas the EU’s AI Act provides Europe with a strong legal framework for AI rights, safety, and compliance, it also offers industrial environments for AI deployment, research networks, standards expertise, public funding, and plans for advanced computing infrastructure.

The EU’s AI Continent Plan prioritizes the development of computing infrastructure, enhancement of data and skills, regulatory simplification, and the promotion of AI adoption in key sectors. Despite these initiatives, Europe holds a relatively small share of global AI computing power. According to Epoch AI, as of May 2025, the EU accounted for approximately 4.8% of global AI cluster capacity, compared with 74.5% in the United States and 14.1% in China.

Besides compute plans, Europe has a growing AI company landscape, though it is smaller than that of the United States or China. France’s Mistral AI is the leading European contender in the frontier model domain, offering both open and commercial models, enterprise agents, private deployments, and custom model development. Germany’s Aleph Alpha has advanced toward sovereign enterprise AI through PhariaAI, prioritizing deployment, government and enterprise applications compliant with the EU’s General Data Protection Regulation (GDPR), and control over data and infrastructure. DeepL, headquartered in Germany, specializes in AI for translation, writing, voice, and enterprise workflows in over 100 languages. Helsing, based in Germany, represents Europe’s emerging AI defense sector, developing AI-enabled autonomous systems, battlefield software, and precision platforms for governments and militaries.

For European companies, cooperation with India can mean access to a large deployment environment, demand for applied AI in public services, and a rapidly expanding base of developers and researchers. For European policymakers, India is also an important partner in shaping AI governance beyond the transatlantic space, especially in areas such as safety evaluations, data governance, and responsible deployment.

Emerging patterns of EU-India cooperation

The GANANA project serves as a notable example of already happening research cooperation in AI. Led by KTH Royal Institute of Technology in Sweden and funded through Europe’s high-performance computing program, the project links European supercomputing centers with Indian research institutions, including India’s national advanced computing center, the India Meteorological Department, and the AIRAWAT AI supercomputing platform. The collaboration focuses on applications in biomedicine, natural hazards, weather, and climate.

Industrial cooperation is also becoming more concrete. HCLTech, one of India’s major technology services companies, has launched a Business AI lab in Munich with SAP, the German enterprise software group, to co-create business applications. The same companies have expanded into physical AI, including warehouse operations, fleet management, and other industrial processes. Infosys, another Indian technology services company, is providing its generative AI platform to Germany’s Handelsblatt Media Group to support the group’s global economics and financial analysis, to Denmark’s STARK Group for its AI digital transformation, and to SAP through an enterprise innovation lab in Düsseldorf.

Multilingual AI should also become a flagship area for collaboration. For example, the India-France AI Declaration explicitly references language diversity and LLMs. Europe faces its own linguistic diversity and challenges with low-resource languages, while Indian companies have experience developing multilingual LLMs. A joint EU-India program on multilingual AI benchmarks, public-sector translations, voice interfaces, and document analysis would be both technically valuable and politically salient.

Regulatory differences

The EU and India do not regulate AI in the same way. Whereas Europe has chosen a horizontal, risk-based legal framework, India does not currently have a standalone AI law. Its 2025 AI Governance Guidelines reflect what the government calls a balanced and pragmatic techno-legal approach, relying on existing laws such as the Information Technology Act and the Digital Personal Data Protection Act, stating that a new horizontal AI law is not required at this stage. Alongside these guidelines, India supports eight strategic AI safety projects focused on tools such as bias mitigation, privacy-enhancing technologies, explainability, algorithm auditing and deepfake detection.

Published in November 2025, the guidelines are built around seven guiding “sutras,” including trust, a human-centric approach and oversight, responsible innovation, fairness, accountability, understandability and safety of AI systems. They favor voluntary commitments backed by techno-legal tools, propose a graded liability system and a national database of AI incidents, and set up new bodies to develop and implement AI governance frameworks, including an AI Governance Group, a Technology and Policy Expert Committee, and an AI Safety Institute responsible for technical expertise.

This regulatory divergence need not be an obstacle to mutual cooperation. The EU is not required to impose its AI Act on India, and India should prioritize governance and safety from the outset. The optimal path forward involves collaboration on shared testing methodologies, multilingual benchmarks, model evaluations, AI safety assessments, and risk management in specific sectors. This is particularly important, as significant risks in applied AI often emerge during deployment, including biased outcomes, unreliable medical or administrative recommendations, deepfakes, cybersecurity vulnerabilities, and failures in less commonly used languages.

Regulatory sandboxes, supervised environments where companies and public agencies can test AI prior to full-scale deployment, are particularly valuable. These environments enable innovators to identify risks at an early stage and provide regulators with empirical evidence regarding the technology’s functionality. The India-EU joint initiative on explainable and robust AI exemplifies this approach by emphasizing the need to ensure that systems are safe, reliable, and comprehensible in practical contexts.

From diplomatic statements to actual cooperation

The EU and India are unlikely to lead the global AI race by outspending the United States or China on cutting-edge models. Instead, the most effective strategy is to collaborate on applied AI by combining Europe’s computing power, regulatory expertise, industrial demand, and research networks with India’s expanding talent pool, experience with multilingual LLMs, and cost-effective engineering. The partnership already benefits from diplomatic support, but it needs to move beyond joint statements to collaboratively develop models, share infrastructure, conduct research projects, and invest across borders.

Several initiatives are already in place. Sweden and India have initiated a technology and AI corridor to connect public agencies, companies, startups, and universities. The Berlin-based FIWARE Foundation and the German Indian Innovation Corridor are developing open-source digital infrastructure, data spaces, digital twins, and sovereign AI. These examples illustrate the types of institutional channels that support joint AI projects, as well as the willingness of both sides to pursue bilateral partnerships.

Ultimately, the extent to which the EU-India AI partnership will foster shared capabilities remains to be determined. If both parties combine computing resources, data, research capabilities, industrial demand, and developer talent, EU-India cooperation has the potential to influence the global deployment of AI beyond the US-China frontier.