Becoming “Global AI garage” or “Digital colony?” India’s AI Policy
Aydin Guven examines India’s AI policy and its ambitious mission to become a “Global AI Garage,” exploring the gap between visionary rhetoric and the country’s practical readiness in AI governance.
India’s bid to lead in artificial intelligence is torn between ambition and dependence. Despite bold visions like becoming a “Global AI Garage,” the country’s reactive policymaking and reliance on foreign hardware risk turning it into a digital colony.
India’s efforts to assert itself in the global AI race are currently defined by uncertainty over intent and capacity. While the government envisions transforming the country into a global “AI Garage” that builds scalable and socially impactful solutions for developing nations, its recent push toward developing a homegrown large language model appears largely reactive—sparked by China’s unexpected success with DeepSeek. Despite increased public investments, new initiatives like AIKosha, and increased GPU acquisition, India still suffers from critical weaknesses: limited access to interoperable data, dependency on foreign hardware, and the absence of a comprehensive regulatory framework for AI.
In this context, whether India becomes a global “AI garage” or remains a dependent “digital colony” will depend on its ability to move beyond reactive policymaking and commit to long-term, autonomous, and sustainable AI capacity-building. India’s current trajectory risks being more symbolic than strategic-comprehensive on matching rivals rather than addressing structural gaps. Despite all efforts, the scale of its financial commitment still lags far behind that of major powers such as the United States and China.
What does India’s National AI Strategy tell us?
If someone looks at the India’s National AI Strategy, they will see that it is closely tied to its broader ambition to of becoming a great power. Viewing AI as a key driver of transformation, the country seeks to establish itself among the world’s leading AI nations through its distinct vision, branded as #AIforAll.
The strategy outlines a vision of transforming the country into an “AI Garage” for the world -serving as a hub for developing scalable AI solutions that can be applied across developing economies, particularly in regions such as Southeast Asia and Africa. As part of its broader push for AI leadership, India was also prioritizing the development of advanced natural language processing (NLP) infrastructure tailored to its diverse linguistic environment.
Furthermore, the strategy considers Large Language Models (LLMs) as public infrastructure, funding a national GPU cloud, open data repositories and grants for “foundational” and “small” models that natively understand all formal Indian languages and many dialects. Projects like Digital India Bhashini and BharatGen/Sarvam-1 help by sharing Indian language data and ready-made AI models for free. The main aim is both social (“AI for All” inclusion via multilingual access) and strategic (tech-sovereign Indic LLMs that cut dependence on foreign platforms).
However, until the announcement of DeepSeek, India had taken few basic steps toward establishing a foundational AI ecosystem, and its progress remained largely limited to policy declarations. That is, while the government has announced that India’s first homegrown LLM will be launched within the next few months, significant uncertainty remains about the underlying drivers of this initiative. It is unclear whether the decision stems from a strategic necessity to build autonomous digital infrastructure, or whether it is largely a reactive move to the success of models like ChatGPT and DeepSeek.
The DeepSeek Effect and India’s AI Awareness
The release of China’s DeepSeek model in 2024 had a transformative impact on India’s evolving AI landscape. Until that point, India’s engagement with AI was largely centered around the application of existing models and tools rather than the development of foundational systems. DeepSeek, however, challenged this passive position by raising AI awareness among Indian policymakers and related sectors, and by demonstrating that developing homegrown models is neither as expensive nor as hopeless as the OpenAI CEO had claimed during his 2023 visit to India.
First, Indian policymakers and sector leaders realized that it is possible to develop such a comprehensive generative AI model with limited financial resources and a relatively small team. The Chinese model was built in less than two years, with just $5.5 million and 200 employees. On the other hand, OpenAI developed ChatGPT with approximately $6.6 billion and 4500 employees. DeepSeek’s success made it evident that cutting-edge AI development was not only reserved for multi-billion-dollar firms. Indian IT Minister Ashwini Vaishnaw publicly praised DeepSeek’s approach, drawing comparisons with India’s own frugal innovation models. His remarks were as a direct challenge to OpenAI CEO Sam Altman’s earlier comments during a 2023 visit to India, where Altman had suggested that foundational AI models were too expensive and resource-intensive for countries like India to develop independently. DeepSeek, in contrast, proved otherwise.
Second, Indian political and economic elites began to publicly frame the global AI race not only as a technological competition but as a strategic imperative tied to national identity and autonomy. Despite having one of the world’s largest AI talent pools and serving as a global hub for IT services, India’s continued reliance on AI models developed by the US and China has sparked concerns that it risks becoming a “digital colony” rather than an autonomous technology leader.
Therefore, Indian policymakers began discussing the importance of achieving homegrown models that could reflect the nation’s unique linguistic, social, and economic characteristics—thereby reclaiming control over digital infrastructure and decision-making.
India’s Growing Investment in AI
The DeepSeek’s effect was not limited to an awareness but also led a sharp increase of public investment in India’s AI sector. Just weeks after DeepSeeks’s release, the Indian Government significantly increased funding for the IndiaAI Mission—raising its annual budget from approximately $66 million to nearly $240 million. While this budget change was seen a sharp increase, India’s overall investment remained modest compared to global leaders. In 2024, after the IT minister praised Deepseek’s success, the Indian government announced a broader $1.25 billion allocation for the IndiaAI Mission aimed at creating compute infrastructure and providing financial support to domestic startups. This was accompanied by an additional $9.1 billion investment under the India Semiconductor Mission, designed to enhance local chip production.
Despite these efforts, the scale of India’s financial commitment lags behind major powers. According to Stanford University's AI Index Report 2025, the US unveiled a $500 billion AI infrastructure plan in partnership with OpenAI, SoftBank, and Oracle, while China launched a $47.5 billion semiconductor fund to further strengthen its position in AI race. In this context, we can say that India has entered the global AI race, but it continues to operate with far fewer resources than its strategic competitors.
Dependency on Foreign Hardware
Hardware infrastructure—particularly access to high-end Graphics Processing Units (GPUs)—has been at the core of large language models (LLMs). As the computational core of machine learning and generative models, GPUs enable the large-scale training and fine-tuning necessary for building competitive AI systems.
No one expected such a comprehensive success with restricted GPU access and limited number of chips. Deepseek was developed with just 2,000 GPUs, and ChatGPT was developed on 25,000 GPUs. However, for countries like India, access to such high-end GPUs remains constrained due to export restrictions.
Being aware of that, since the release of DeepSeek, India has more than doubled its GPU stockpile, rising from roughly 10,000 units to over 18,000, by early 2025. Of these, 15,000 are high-end chips, including 1,480 H200s—placing India’s compute capacity significantly above DeepSeek’s (2,000 GPUs) and approaching that of early versions of ChatGPT (25,000 GPUs).
Also, the government has committed to procuring and distributing advanced GPU models such as the Nvidia A100 and H100 models, which are very difficult for many countries, including China, to access because of U.S. export restrictions. However, collecting external manufactured chips is a short-terms solution. What India truly needs is to increase the incentives for its local semiconductor sector and build deep, sustainable domestic capacity. To solve this problem, Indian government increased the R&D fundings for semiconductor industry. The government allocated a $9.1 billion for “Semicon India” program to boost semiconductor manufacturing. Also, India Semiconductor Mission is reportedly preparing to launch a design-linked incentive (DLI) scheme worth up to $4 billion to promote indigenous semiconductor design and reduce dependence on foreign chipmakers. However, these steps are not enough compared to the scale of investments by the U.S. and China.
Becoming Global AI Garage or a Digital Colony?
As India prepares to release its own large LLM within the next few months, the country faces a series of challenges that could impede the effectiveness of its AI ecosystem. Yet it remains unclear whether this initiative is driven by strategic necessity or simply a reaction to the rapid advances of models like DeepSeek and ChatGPT. Whatever the reason, it comes with a set of challenges.
A primary weakness is about AI infrastructure and data: the lack of data interoperability and public access at scale. Access to diverse and high-quality datasets is a prerequisite for developing competitive AI models. India’s complex linguistic and institutional landscape has historically limited the consolidation of standardized datasets. Although India has already launched the IndiaAI Dataset Platform (AIKosha) in March 2025, to serve as a national repository that collects non-personal, high-quality datasets across sectors, there is still no clear evidence that the platform has yielded meaningful results or significantly improved data accessibility for AI development.
A second concern is about the absence of a clear regulatory framework specifically addressing AI. While the Digital Personal Data Protection Act of 2023 declared as a guidelines for the processing of digital personal data, it does not deal with AI related privacy and ethical concerns. The Act primarily focuses on consent-based data sharing, fiduciary responsibilities, and penalties for data misuse, but it remains weak on broader issues that arise when machine learning models are trained on large, unstructured, and potentially sensitive datasets.
These challenges stand in contrast to the ambition articulated in India’s National AI Strategy. The strategy envisions India becoming a global “AI Garage”—a hub for building scalable, affordable, and socially useful AI solutions for the world. However, until recently, India remained largely a user and adopter of AI models developed by the US and China. To realize its vision of becoming an “AI Garage” for the world, India must move beyond aspirational strategy documents and demonstrate sustained commitment to regulatory clarity, institutional coherence, and computational capacity.
Aydin Guven is a PhD candidate in Political Science at George Mason University’s Schar School of Policy and Government. His research interests include South Asian international relations, Indian strategic autonomy, Indo-Pacific security, Political Economy of AI, and Techno nationalism.
Affiliation: Schar School of Policy and Government, George Mason University
Photo by Pixabay
