By: Dhruva Jaishankar and Jeffrey D. Bean
U.S.-India technology cooperation continues to advance despite U.S.-India friction in other areas, sustaining the commitment of the two countries’ leaders in February 2025 to develop a roadmap for accelerating the AI infrastructure build out in India. U.S. technology firms are doubling down on their already sizeable investments in India in artificial intelligence infrastructure. Google has broken ground on a $15 billion gigawatt-scale AI hub in Visakhapatnam and has more employees in India than anywhere outside of the United States. Microsoft is investing $17.5 billion in cloud and AI infrastructure in India over four years centered around a large AI data center in Hyderabad, and building on an earlier $3 billion in funds. Amazon Web Services (AWS), which has 100,000 employees in India and hosts many elements of India's digital public infrastructure, has committed another $35 billion outlay on artificial intelligence infrastructure in India by 2030 on top of an existing, broader $40 billion investment. Meta has announced a partnership with Reliance to build a 186-megawatt datacenter in Gujarat. Meanwhile, firms like Intel, Qualcomm, and AMD have chosen to partner with Indian conglomerate Tata in different areas of technology manufacturing and the India Semiconductor Mission has notched some key successes, including with Micron. Several of these efforts have benefited from recent improvements in India's business environment, including tax breaks, large subsidies, infrastructure upgrades, and increased energy availability.
Official avenues for cooperation on emerging technology have also endured. India has joined Pax Silica, a U.S.-led effort to secure AI supply chains and the two countries signed a U.S.-India AI Opportunity Partnership in February as India hosted the AI Impact Summit. Yet just this week, the Trump administration’s struggle to align policies intended to export the American AI stack and model services while retaining export control safeguards has manifested in policy disruption. Earlier this month, India was among the first countries to receive expanded access to Anthropic's Claude Mythos large language model, but the White House decision to block all foreign access to Anthropic’s new Mythos and Fable models via export control elevates concerns regarding reliability of U.S. partnership and AI dependency.
While public- and private-sector cooperation between the United States and India on AI infrastructure development continues, several challenges remain. It does not help that the international environment is laying adverse conditions. U.S. tariffs on many countries — including India — have injected uncertainty, which has only been compounded by global supply chain crunches (including on semiconductors) and the Iran War. But while the Indian government will naturally prioritize home-grown solutions, resilience, domestic manufacturing, and economic development, it can continue to reduce barriers to accelerating India's digital transformation through external partnerships, alleviating ease of doing business factors, and further improving core infrastructure.
One set of challenges has emerged as AI, digital, or data sovereignty — a priority for the Indian government — gets translated into legislation and policy. While digital sovereignty is a natural impulse — and is by no means unique to India — how its policies are formulated, interpreted, and imposed can have important ramifications for tech companies' abilities to conduct business. The fact is that even the United States is not truly sovereign and is dependent on other actors for many aspects of its AI stack, from the manufacturing of semiconductors to talent and data. As a result, India will need to take this into account when considering local content requirements for global firms with fine-tuned advanced technology supply chains. At the same time, India’s 21-year tax moratorium on data centers announced in February is a tangible financial commitment.
Second, India's import licenses and certifications remain uncertain, with necessary authorizations by the Bureau of Industry Standards (BIS), Directorate General of Foreign Trade (DGFT), and Department of Information Technology taking as long as 4 or 5 months — a long-time particularly in a fast-moving sector such as AI, where CPUs and GPUs are often replaced regularly. For some badly needed infrastructure, such as high-speed undersea cables and connectivity, approvals from even more Indian government agencies are required, lengthening wait times to 8 months, on top of initial project approvals that can take years. The effects of import duties and uncertainty affect a variety of necessary hardware imports, including data server trays and racks and refurbished or repaired semiconductors. Logistic delays, which have improved significantly in India, can also be further ameliorated. From a private sector perspective, India remains in competition for technology manufacturing and AI infrastructure with states like Vietnam, Malaysia, Thailand, and Singapore that have addressed these issues more fully.
Finally, there remain assorted barriers to entry to the Indian market which inhibit long-term investment. Certain newly introduced policies, even if well intentioned, are not necessarily feasible, requiring constant regulatory clarifications and tweaks. Notifications from the Ministry of Electronics and Information Technology (MeitY), Ministry of Home Affairs (MHA), and other relevant Indian government agencies are not always harmonized. Although U.S. tech firms stand to benefit from long-term tax incentives, steps can be taken to sequence them in a manner that can feasibly enable the manufacturing and exports of complex hardware.
The U.S. and Indian governments could prioritize addressing these challenges in a manner that serves both countries' interests: accelerating U.S. AI exports and investment to India and advancing India's digital transformation while allaying Indian concerns about sovereignty and security. The resurrection of a bilateral high technology dialogue that addresses digital trade, export controls, and cybersecurity notifications — and involving the relevant government agencies in both countries — would help. Indian participants would include some combination of MeitY, MHA, External Affairs (MEA), Information and Broadcasting, Telecommunications, and Commerce and Industry, and the Prime Minister’s Office. On the U.S. side, it would require the Departments of State, Commerce, and (on some issues) War, as well as the Cybersecurity and Infrastructure Security Agency (CISA) and the White House. Discussions through such a dialogue or other platforms on how to define and apply data localization and sovereign requirements, better align cybersecurity standards, and formulate tax regimes to incentivize cooperation on AI-related hardware would also be beneficial to both countries.
Dhruva Jaishankar is Executive Director and Jeffrey D. Bean is a Fellow for the Technology Policy Program and Editor at ORF America.