The Global State of Play
There is little doubt that the world is at the cusp of a new revolution in information technology with the advent of Generative Artificial Intelligence (GAI). This is symbolized by the launch of the US-based Nvidia’s ChatGPT in November 2022.
As Indian Prime Minister Modi appropriately stated in his inaugural remarks at the recently convened February 2025 Artificial Intelligence Action Summit (AIAS) in Paris which was attended by over 90 countries and top global tech company leaders, among others, which he co-chaired with President Macron of France, AI “is already shaping our polity, our economy, our security and even our society. AI is writing the code for humanity in this century….it is very different from other technological milestones in human history.”
GAI should be differentiated from Traditional Artificial Intelligence (TAI) which is around 10-15 years old. The two can be differentiated thus: TAI can and is being used for analytics, fintech, credit scores and the like i.e. it is being used to analyse data and make predictions. GAI, on the other hand, is used to create new content. This is a fundamental difference between the two.
The conventional wisdom thus far has been that the US is the undisputed global leader in AI. This was recently reinforced by the Biden Administration’s January 2025 “Framework for Artificial Intelligence Diffusion,” (FAID) effective 13 January 2025. This was issued as an “interim final rule” with comments on revisions and additions to be received by the US Commerce Department’s Bureau of Industry and Security (BIS) no later than May 15, 2025, after which full compliance is expected. FAID was promulgated literally in President Biden’s last week in office.
Its key premise is that US national security requires that it maintain technological leadership in the global AI industry. The Framework restricts the export of AI chips and Graphic Processing Units (GPUs) from the US, categorizing countries in three-tiers with specific restrictions for each tier. While China, together with North Korea and Iran has been placed in tier-3 explicitly reserved for US adversaries, with the most restrictive controls on the dissemination of advanced AI technologies, India has been placed in the intermediary tier-2, also subject to heavy export restrictions on the number of GPUs and advanced AI chips (roughly 50,000 through 2027) and therefore computing power, it can import from US based companies.
President Trump, on his second day in office during Trump 2.0, clearly indicated that he plans to expand the global dominance of US companies in AI.
Till very recently, US companies such as OpenAI, Microsoft and Google had been setting the GAI narrative because they had access to the most cutting-edge hardware made by another US company, Nvidia. They also had access to the best AI talent pool in the world. Moreover, OpenAI, Softbank, Oracle, Microsoft and Nvidia are coming together over the next four years to build AI infrastructure for OpenAI through a new project and company, Stargate, with an anticipated investment of USD 500 billion.
However, in January 2025, an earthquake reverberated through the global AI industry which has the potential to change all of this, although it should be cautioned that it is still premature to arrive at a definitive conclusion on this. It was triggered by a Chinese AI laboratory when it released DeepSeek, a low-cost, OpenSource foundational model, costing only around USD 6 million. This appears to demonstrate that AI foundational model development may not be as expensive as previously thought, notwithstanding the credible belief of some experts that the financial resources required for DeepSeek may have been significantly higher.
Nevertheless, the development of DeepSeek has threatened to upend the economics of GAI development since, even if the USD 6 million figure is significantly understated, its total cost will represent only a very small fraction of the USD 6.6 billion that was spent to train OpenAI models.
DeepSeek’s R1 ‘reasoning’ model is now being proclaimed as almost on par with OpenAI’s new o1 ‘reasoning model’. China’s e-commerce giant, Alibaba, also released a new AI model, QwQ in November 2024. This is said to rival OpenAIs GPT-o1 series models in reasoning capacity, making it yet another credible Chinese competitor to the US’ OpenAI. DeepSeek is also open source, unlike OpenAI’s models. This allows developers to easily build on top of its models.
Not only is DeepSeek competing with OpenAIs latest models on several parameters, but it is noteworthy that while it still relied on Nvidia GPUs, it was based on an older technological version because of China’s limited access to cutting-edge Nvidia AI hardware resulting from US sanctions.
The Chinese DeepSeek shock will, no doubt, be further reinforced by credible recent research findings which suggest that China is ahead of every other country in the world in almost every strategic and critical technology area. TK Arun, referencing the Australian Strategic Policy Institute (ASPI) which has both identified 64 critical technologies and tracked the progress of different nations in generating research in areas of strategic capacity belonging to several broad clusters which directly impact the global race for future power wrote that, according to ASPIs end-August 2004 report, China has been leading in 57 out of 64 strategic technologies since 2019. The US only leads in seven.
In the case of China, this was an exponential increase from it being the global leader in only three such critical, strategic technologies between 2003-07 (https://the federal.com/category/opinion/india-ai-mission-deepseek-169536, February 1, 2025).
Can India Catch Up with China?
Policy work on AI in India began only in 2018-2019. China has had a head start over India in AI for well over a decade. Moreover, AI projects funded for strategic purposes such as military and law enforcement, smart automation and intelligence have been part of China’s technology plans for more than three decades with AI being included in the country’s five-year plan back in 2006. China has also used deep tech for military, law enforcement, image recognition and facial recognition purposes for some time now and AI automation for strategic autonomy since 2008/09.
China’s total research output is also several multiples of India’s and becomes more substantial when measured by the quality of publications against their quantity. While ASPI judges the quality of Chinese papers to be high, it assesses the quality of Indian papers to be generally low. Fintech is the one AI area of India’s competitive advantage vis-à-vis China, but this is a TAI application, not GAI development.
Moreover, unlike China, India does not have good examples of government led IT services or infrastructure. Its main role has been to keep out of the way of the private sector in this area while providing them with some incentives (e.g. DPI, UPI, Open Source).
India and AI
Despite the objective facts presented above, there is a belief by some in India that it is in the forefront of AI development and application. This may have been reinforced by Prime Minister Modi when, in his opening remarks at the recent Paris AI Action Summit (AIAS), he disingenuously stated that India was the world’s leader in AI adoption without clarifying that he was referring to TAI “consumption,” (e.g. using Google Search) not AI development.
In this context, it is also important to note that while private sector companies such as Microsoft and Intel have, indeed, used AI in India for data analytics, fintech, credit scores, human resources development and even some R&D especially in Global Capability Centres (GCCs), these are examples of TAI not GAI.
While India ranks in the global top five on Stanford University’s AI Vibrancy Index which ranks countries on metrics such as patents, funding, and policy and research, experts all agree that India is many years behind the two global leaders, China and the USA. Most notably, India lacks its own hardware and compute infrastructure.
The Government recently announced an ambitious outlay for the India AIMission of Rs 10,371 crores (a little over USD 1.1 billion). While this was clearly a declaration of intent focused on AI research, developing computing power and industry collaboration, it is only a few years old, gaining momentum with funding proposals only two years ago in 2023. As per the Ministry of Finance’s expenditure budget documents, the AIMission was sanctioned the relatively modest amount of Rs 551.75 crores (USD 74.5 million) during the Union Budget 2024-25. Even this relatively modest amount was significantly underutilized, leading to a budget revision to a much-reduced figure of Rs 173 crores (USD 23 million).
Rs 2000 crore (USD 270 million), nearly 25% of the entire India AIMission budget has been approved as part of the most recent Union Budget 2025-26 announced on February 1. 2025. This comes on the heels of the Indian government announcement that it was planning to fund the development of one or more foundational models just days after the release of China’s DeepSeek.
The government has also empanelled 10 companies in India from whom it will now procure and supply 18,693 Graphic Processing Units (GPUs) for usage by end-users, institutional and individual. This is much more ambitious than the original aim of the AIMission under which the government was looking to procure only 10,000 GPUs. The government has also announced that it will establish a new AI centre of excellence for education with an outlay of Rs 500 crores (USD 65 million), in addition to the three centres of excellence in AI for agriculture, health and sustainable cities the Union Finance Minister had announced in 2023.
While the intent is admirable and clearly in the right direction, it will clearly not be enough to narrow the lead that China and the US have on India. Moreover, the promise of developing India’s own foundational model, including major Indian languages in just 10 months, a timeline which appears to have been pressured by China’s recent DeepSeek announcement, appears to be a dream too far. This is because Bengaluru’s IT companies which should have been at the forefront in this area have never really shifted their focus from cheap service-based IT work to developing foundational GAI technologies.
Lack of high-quality India specific data sets required for building training AI models in national or regional Indian languages such as Hindi, Marathi or Tamil is another constraint. Start-ups, expected to fill this gap, cannot do so. The best India may be able to do in the short or medium-term is to build on top of OpenSource platforms like DeepSeek.
There are also numerous other challenges which GAI presents a country such as India with its 1.4 billion population which it will also need to urgently and simultaneously address. For example, it will need to immediately prepare and resource appropriate high value-added skill replacement programs to address the likely huge domestic job disruptions, in the hundreds of thousands, if not millions of people. This is in addition to addressing the country’s already dire unemployment, underemployment, unemployability, ill-suited employment and informal unpaid work crises.
GAI is also likely to rapidly lead to tech service disruptions. “Knowledge arbitrage” industries like IT services, law and medicine are among the most vulnerable to GAI. These jobs are expected to be among the first to be displaced by GAI. India is also in danger of being displaced in its global IT services leadership role because this is still largely at the low-end, or at best middle data entry level of the supply chain, with significant negative consequences for India’s USD 300 billion tech services exports.
Global Regulation and India’s Emerging Regulatory Framework
The global regulatory policy response so far has been different in different jurisdictions with the EU having taken a tougher stance by proposing a regulatory framework through its EU AI Act. It has also established several AI Safety Institutes. The UK is seen to be at the other end of the spectrum with a laissez-faire or ‘light-touch’ approach which seeks to foster, not stifle, innovation in this field, while the US so far has been in-between but is now likely to become much more deregulated under Trump 2.0. China too has released its own set of measures to regulate AI.
In India, regulation has been mixed. The government says that it supports a balanced AI regulation framework aimed at fostering technological innovation, safety and trust with a focus on regulating social media, addressing risks like deepfakes and misinformation for which it is developing guidelines. However, there are, yet no separate AI laws or regulations. India is in the process of developing a Digital Infrastructure (DI) Act for this purpose which is expected later this year.
In the meanwhile, the 2000 IT Act has had very heavy-handed regulation which led to self-censorship by Netflix, Amazon Prime and others because of the risk of being sued and even subject to police action. Of concern is that the new Indian Penal Code is now being applied to AI.
India’s Potential Niche in the Global South
India cannot hope to be a potential bulwark to China since it is way ahead in both AIs industrial and military uses. China’s leadership aspiration is global in these and other areas, not merely leadership of the Global South.
India advocated ‘AI for All’ during its 2023 G20 Presidency, emphasizing inclusive and equitable access. It proposed an international AI governance framework like the global Climate accords. It also encouraged collaborative AI research with global partners, promoting ethical AI use in developing countries. These are not issues that China has publicly or globally championed.
India was asked to Co-Chair the recent Paris AIAS by President Macron largely because of perceived shared values, despite China being at the technological forefront of AI. This was a welcome development which India needs to build on quickly from the perspective of and for the benefit of the Global South.
At the global level, India and France should build on the 2015 initiated joint India-French International Solar Alliance to promote sustainable AI for the global public good which should not only mean using clean energy but the promotion of India GAI foundational models which are sustainable in size, data needs, resource requirements and efficiency.
India has been keen to globally leverage its Digital Public Infrastructure (DPI) which it has used to bridge the digital divide in some IT areas through its path-breaking Unified Payments Interface (UPI). It should seek to extend DPI by building GAI computing infrastructure in a transparent, inclusive, responsible and equitable manner for the benefit of the Global South. It should also seek to bridge the AI divide to benefit the Global South.
India should also prioritize AI safety and ethics in the Global South. The Government’s Ministry of Electronics and Information Technology (MEIT) recently announced the establishment of an AI Safety Institute (AISI) to set standards, frameworks and guidelines for AI development. It will soon join the global network of AISIs currently being established which India should aspire to lead for the Global South, setting an example for the whole world on the ethical and safe use of GAI.
India will, however, only have the credibility and the substantive and technical heft to play these roles for the Global South in the future AI tech driven world if it urgently and successfully develops its own AI hardware capabilities, a journey on which it is yet to effectively embark.
ABOUT THE AUTHOR
Kamal Malhotra is Non-Resident Senior Fellow at the Boston University Global Development Policy Center and has recently also Guest Lectured at the NALSAR University of Law, Hyderabad and the School of Interwoven Arts and Sciences (SIAS), Krea University, India. Prior to his retirement from the United Nations in September 2021, Mr. Malhotra had a rich career of over four decades as a management consultant, in senior positions in international NGOs, as co-founder of a think-tank, FOCUS on the Global South, and in the United Nations (UN) including as its Head in Malaysia, Turkiye and Vietnam (2008-21). He was UNDPs Senior Adviser on Inclusive Globalization, based in New York, USA, for most of the prior decade. Mr. Malhotra is widely published.
