Whalesbook
India's AI Hardware Boom Fuels Foreign Companies
India is investing heavily in artificial intelligence infrastructure like data centers and hardware. However, this push inadvertently benefits foreign tech giants because the crucial AI intelligence—the core engines and foundational models—remains largely owned by overseas companies. This setup creates a "Token Tax," where small payments for each AI use, from coding help to personalized suggestions, leave India. Global tech leaders are expected to earn trillions by 2026, with a large share coming from fast-adopting markets like India that don't own the underlying AI models.
The 'Compute Deficit': India's AI Power Shortage
A major concern is India's "Compute Deficit," meaning a significant lack of AI computing power. Nvidia executives noted that India has less than 2% of global AI compute, while the US and China together hold nearly 60%. This scarcity limits India's own AI research and forces reliance on foreign resources. Unlike visible trade deficits in oil or gold, this digital outflow is hard to track but has major long-term economic consequences. India's traditional IT success, based on skilled labor and software licenses, is challenged by this new model. It operates like a rental service where the most profitable part—fees for AI intelligence—is paid to foreign providers.
Nations Build Their Own AI Capabilities
Several countries are actively developing their own AI to prevent dependency. France, with strong government support, backs Mistral AI as a European rival to US tech giants, ensuring its military data stays under French control. The UAE is heavily investing in its own Falcon AI models to diversify its economy beyond oil and build independent AI power. China has focused on insulating its AI sector, prioritizing self-reliance in chips, hardware, and algorithms to create an "independently controllable" AI industry.
UPI Success Offers AI Blueprint
India has previously overcome similar dependency challenges. The Unified Payments Interface (UPI) is a key example. Instead of letting global companies control payment infrastructure, India built its own digital system, kept ownership of the core technology, and allowed local companies to build on it. This strategy boosted domestic growth. A similar approach could work for AI, with the government supporting consortia that build national AI models and compute power, rather than just subsidizing foreign hardware.
AI Reliance Creates Economic Pitfalls
Relying on rented AI intelligence poses a major risk for India's AI startups. The financial model is unsustainable: thin profit margins in India, with low average revenue per user, cannot cover per-use fees in dollars paid to foreign AI providers. This traps products in a "pilot purgatory," where they function but can't scale profitably, hindering innovation. Globally, companies controlling key AI layers—like advanced chips, large-scale computing, and AI models—capture most of the economic benefits. These areas are dominated by US firms. Without owning these critical layers, India risks exporting labor and hardware but not the valuable intelligence itself.
India's Path to Owning Its AI Future
To address this, India needs to shift from simply using AI to controlling it. This means developing its own AI models and compute infrastructure. Companies should look to adopt Small Language Models (SLMs) that can be trained on their own data and run locally, reducing reliance on foreign gatekeepers and improving privacy. While AI could add hundreds of billions to India's GDP by 2035, this potential can only be met by investing in domestic AI capabilities, not just hardware. The choice is stark: keep subsidizing hardware and sending profits abroad, or do the harder work of building and owning India's AI future. If India doesn't act by 2030, its most costly import might not be oil or electronics, but intelligence.