There is a particular moment in every hardware programme where the spreadsheet stops helping. You have sourced the components, validated the stack, and the thing in front of you does not behave the way the model said it would. In that moment you discover whether your organisation has capability or merely procurement.
The distinction matters more than it sounds. A country can consume the most advanced technology in the world and still hold none of it. Consumption is a transaction; capability is an accumulation. It lives in the engineers who have failed at something enough times to have intuition about it, in the test rigs nobody else would have built, in the supplier who now understands your tolerance requirements because you spent two years explaining them.
Autonomy is a systems problem wearing a software costume
It is tempting to treat autonomous systems as an AI problem. In practice, perception is the part that demos well and the part that fails last. The hard problems live in the seams: power budgets, thermal behaviour, comms degradation, the failure mode you only find at hour nine of a duty cycle. Autonomy is what a full system does under stress, not what a model outputs in a notebook.
“The technologies that define the future should not only be consumed in India. They should increasingly be imagined, engineered and built here.”
Vihanga.ai
What self-reliance actually costs
Indigenous capability is expensive in the currency that matters least to markets and most to institutions: time. It requires funding a learning curve that a purchase order would let you skip. It requires accepting that your first iteration will be outperformed by something you could have simply imported, and building the second one anyway.
The return arrives later, and it is not a product. It is a set of people who know how, a supply base that can respond, and a decision you no longer have to ask anyone else's permission to make. That is the asset. Everything else is a deliverable.
