US Expands Strategic Controls on AI Technology
The United States is increasingly treating advanced artificial intelligence capabilities as strategic assets, expanding its export controls beyond semiconductors to include frontier AI models. Initially aimed at containing rivals with restrictions on advanced chips in place since 2022, these policies now signal that access to cutting-edge AI hardware and software is a geopolitical concern, not merely a commercial one.
Recent restrictions on models like Anthropic's Fable 5 and Mythos 5 for non-US individuals highlight this shift. Experts suggest these controls are a "wake-up call" for nations like India, which are heavily reliant on foreign technologies across the entire AI stack, from semiconductors to software frameworks.
India's Deepening Reliance on Foreign AI Infrastructure
Despite India's ambitious push for sovereign AI capabilities through initiatives like the IndiaAI Mission, a structural dependence on foreign technology remains. The mission aims to establish a national compute infrastructure with over 38,000 GPUs to support startups and research. However, the vast majority of these GPUs come from US companies such as NVIDIA, AMD, and Intel.
This reliance extends beyond hardware. Even indigenous efforts, exemplified by Bengaluru-based Sarvam AI, which develops foundation models tailored for India, utilize foreign frameworks. Sarvam's leading models have been trained using NVIDIA's Megatron-LM and NeMo frameworks on thousands of imported H100 GPUs. While intellectual property and end-user applications may be domestic, the underlying compute infrastructure and software ecosystem are overwhelmingly American.
Risks of Software Lock-in and Future Access
The risks associated with this dependence are multifaceted. Beyond existing hardware, the greater concern lies in future access to advanced chips, model updates, cloud services, and technical support. Software ecosystems, with their integrated libraries and developer training, create a deep lock-in that makes substitution slow, costly, and operationally risky.
Losing access to critical AI services like ChatGPT, Claude, or Gemini could immediately disrupt Indian startups, enterprises, and researchers, leading to higher costs and slower innovation. Furthermore, such restrictions could damage India's reputation as a global hub for AI work, impacting both IT service providers and global capability centers.
Pathways to Reducing Dependence
Experts caution that complete technological independence is neither practical nor economically viable. Instead, India's strategy should focus on reducing dependence in areas of highest risk where alternatives are attainable. Key priorities include:
- Achieving inference sovereignty
- Ensuring model portability
- Building infrastructure resilience
- Developing robust Indian-language AI capabilities
- Implementing disciplined procurement practices
The immediate focus should be on making AI workloads portable, establishing fallback options for critical systems, and securing clear audit and exit rights within contracts. Ultimately, while India strives to build its own AI future, it must ensure its ambitions are anchored in capabilities and ecosystems it can reliably access, shape, and leverage.