Anuraag Saxena 警告：地缘政治可能拖累印度 AI 雄心

An Indian AI company could build a competitive model, secure funding and find customers, only to discover that a government decision has put a critical part of its business out of reach.

That is the geopolitical risk Anuraag Saxena, a public policy leader, believes Indian AI companies are underestimating.

Speaking at Cypher 2026 in Bengaluru, Saxena argued that sovereignty is not about building everything domestically. It is about ensuring that no external power can unilaterally disrupt a company's ability to operate.

Sovereignty, he said, means that “no one else can unilaterally switch you off.”

For AI companies, this could mean losing access to advanced computing infrastructure, critical materials, skilled workers or important markets because of decisions taken far beyond their control.

Geopolitics is becoming a business risk

Saxena said technology had historically helped overcome geographical constraints. That advantage is now being challenged as access to natural resources increasingly determines what countries and companies can build.

“Natural resources have become a natural limitation again,” he said, describing it as a fundamental macroeconomic trend that has not received enough attention.

India's dependence on China for rare-earth magnets illustrates the problem. A September 2026 report by Boston Consulting Group and the Automotive Component Manufacturers Association of India, reported by Autocar Professional, said India sourced nearly 85% of its rare-earth magnets from China. China's export curbs, introduced in October 2025, were set to expire on November 10, 2026.

The same vulnerability extends to AI infrastructure, where access to advanced chips and computing capacity can depend on export controls and a concentrated supplier base.

Talent presents another risk. The Trump administration's October 8, 2026 decision to suspend labour certification applications for eight technology companies, including TCS and Infosys, highlighted how immigration policy can disrupt workforce planning and long-term employee mobility.

Saxena described the contradiction between companies seeking skilled engineers and voters demanding tighter immigration controls.

“Immigration is an electoral issue,” he said, pointing to a situation where businesses want more coders while political pressure favours keeping foreign workers out. For technology companies operating across borders, this creates “competing regulations”.

“Very often the building of a wall comes from an electoral objective,” he said.

His point was that political decisions need not target technology directly to affect a company's ability to operate.

AI companies need geopolitical insurance

Saxena urged AI founders to look beyond conventional questions about funding, computing costs, hiring and customer acquisition. They must also ask whether governments can restrict access to any of the resources or markets their businesses depend on.

A sudden policy change, he warned, could turn off a tap and bring an enterprise to a halt.

“Think geopolitically from day one,” he advised, arguing that redundancy should be treated as geopolitical insurance. Companies should “hedge for jurisdictions” and avoid unnecessary dependence on a single provider or vendor.

For AI businesses, that could mean diversifying compute providers, assessing exposure to export controls and ensuring that operations can withstand regulatory changes in key markets.

The objective is not to eliminate international dependencies, which would be difficult for any country to achieve. It is to identify critical dependencies and build alternatives before a disruption occurs.

Saxena also questioned whether India could rely on domestic consumption alone to develop its AI industry. That strategy would work, he said, if other countries viewed India primarily as a consumer market. But the US and China see AI as an instrument of technological dominance.

“Shameless national interest should come first,” he said.

Recalling India's computer industry in the 1990s, Saxena said domestic companies such as Zenith and HCL had sold 286 computers at roughly half the price of Dell and HP. Yet India did not establish lasting leadership in hardware.

“We lost the hardware war, then the software war, then the enterprise solutions war, then the social media war,” he said. “The last thing I want to see is that we lose out on the AI war.’

For Saxena, the next decade will not be won simply by companies building the best models. It will be won by those that understand the geopolitical conditions under which their AI businesses must operate.

For India, that means building domestic capabilities while recognising that sovereignty is not the same as self-sufficiency. The real test is whether Indian AI companies can keep building, deploying, and selling their technology when geopolitical priorities change.