The IndiaAI mission aims to reduce dependence on American and Chinese AI tech while pushing the startup, research, and development ecosystem back home. But the reliance on American hardware demands that India recalibrate the priorities in its mission and bring hardware technologies into focus.

Developing indigenous large language models (LLMs) to reduce dependency on open-source alternatives from other nations remains the focus of the ₹10,000+ crore IndiaAI mission. It is a step in the right direction to address possible geopolitical friction that could restrict access to powerful AI models in the future, poised to bring a shift in scientific discovery, defence, and public infrastructure.

If potential software restrictions justify building homegrown models, what guarantees continued hardware access?

Even if the IndiaAI mission achieves a state-of-the-art model, it raises questions about whether the project can continue to be upgraded if the country does not have access to leading-edge GPUs.

While the US and other chip-producing nations maintain their competitive advantage through hardware control, India’s quest for AI independence has to deal with this paradox, which may keep it dependent, for hardware if not for software. 

Volatile Geopolitics

Balaraman Ravindran, who heads the Wadhwani School of AI at the Indian Institute of Technology, Madras (IIT-M), echoed this sentiment last month. “What if a president [of a country] suddenly decides to change his mind and say — no GPUs for India. What will happen then?”

The US, at least for now, permits India to purchase, with the necessary licence, some of the leading GPUs, unlike its trade with China, riddled with export controls. Today, India has access to over 34,000 GPUs for the IndiaAI mission. All of these GPUs are made by companies in the US, like Intel, AMD, and NVIDIA. 

Under the US AI Action plan, the country aims to continue exporting AI hardware and systems to nations with healthy relations. But the question remains whether India will continue to have that purchasing power. 

“Today, GPUs and other AI accelerators are under export control and are already in place in certain countries [adversaries]. As model capabilities continue to increase, there is a world where those controls could be extended further,” said Myron Xie, an analyst from SemiAnalysis, in an interaction with AIM. 

And in such an event, Xie said that companies like NVIDIA would have to obey, despite the effects it may have on its business. “However, as is the case with China export controls, a black market would be created where unscrupulous actors could illicitly re-export GPUs.”

American Export Controls

Besides the GPU market, the US holds a strong position in the software and hardware for designing and developing these chips.

For instance, Dutch company ASML builds systems for manufacturing and designing GPUs, but the US largely controls exports. US restrictions on advanced components and software development for these high-end systems make it a challenge to have the GPUs assembled elsewhere through joint ventures or partnerships. 

Besides, the US continues to safeguard some of the most sensitive technologies through export controls. Under the International Traffic in Arms Regulations (ITAR), most items on the US Munitions List, ranging from fighter-jet parts to advanced sensors, cannot be exported without prior authorisation from the State Department, except in a handful of narrowly defined exemption cases (certain transfers to Canada or NATO allies).

Another example is the country’s nuclear submarine technology; the US keeps reactor designs and other technology on the Excluded Technology List, and even countries like Australia must navigate cumbersome waivers despite a defence pact. 

It wouldn’t be difficult to envisage such regulations and controls if AI hardware were regarded in the same way as defence technology, given its capabilities. In the AI action plan, the US has already mentioned that “Advanced AI compute is essential to the AI era, enabling both economic dynamism and novel military capabilities.” 

Semiconductor Mission Needs to Pay Off

On the political front, fostering healthy diplomatic relationships is one way to secure India’s AI future, but a more reliable approach would be to establish a sovereign semiconductor manufacturing industry. At the heart of it lies the country’s semiconductor mission announced in 2021 with an outlay of ₹76,000 crores. 

On the other hand, China, already facing restrictions on AI hardware, has made significant progress with its homegrown company, Huawei. For example, the company’s CloudMatric 384 system is widely regarded as on par with NVIDIA’s powerful Blackwell-based GB200 NVL72 rack-scale hardware. 

Currently, India does not yet mass-produce chips at cutting-edge nodes; its first domestically made chip, of 28nm, is expected to be rolled out at the end of this year. For context, NVIDIA’s A100 GPUs, widely used to train AI models, are based on a more advanced 4nm manufacturing process.

While manufacturing chips at that level still remains outside India’s capabilities, projects to design 3nm chips have been launched in the country, recently in Noida and Bengaluru. This marks India’s first attempt to work on a 3nm chip design. 

“We’ve done 7nm and 5nm earlier, but this marks a new frontier,” said Ashwini Vaishnav, the minister of electronics and information technology. 

While various challenges loom over India’s semiconductor manufacturing capabilities, Vaishnaw remained optimistic in saying that India will have its own GPUs in 3-5 years.

Furthermore, recent events add a new sense of urgency to building sovereign infrastructure, with Trump announcing an unprecedented amount of tariffs on India. 

“Every few years, the world reminds us of our place. A threat here, a tariff there. But the message is the same: stay in your lane, India,” said Deepinder Goyal, the founder of Eternal. 

“Global powers will always bully us unless we take our destiny into our own hands. And the only way to do that is if we collectively decide to become the world’s largest, most unapologetic superpower in the world. In economy, in technology, in defence, and most importantly, in ambition,” he added.

The post The Paradox in India’s AI Mission appeared first on Analytics India Magazine.