NEW DELHI: Walk the floor at Aero India and you will hear the same litany. Engine. Stealth. Weapons. Timeline.Yet, the real battlefield transformation is happening elsewhere, driven by India’s Military AI. Hardware is easy to photograph. Code is not. That is the problem.
The Strategic Imperative of India’s Military AI
Earlier this year, a DRDO official said the quiet part out loud, noting India cannot outsource military AI. The press filed it under “ChatGPT bad.” That missed the point entirely.
AMCA, Tejas Mk2, and the drones meant to fly with them will not be won in a wind tunnel. They will be won in a server room. A jet today is a flying data centre. It must fuse its own radar with a wingman’s feed, a satellite tip, a ground radar track, decide what is real, jam what is hostile, and hand the pilot one clean choice before the missile arrives. That is not an avionics upgrade. That is combat.
Why AMCA Requires Indigenous Software Sovereignty
We keep debating AMCA as if it is an engine deal with wings. It is not. The engine matters. So does the shaping. But the jet lives or dies on its mission computer, its sensor fusion, and its decision aids. Without those, stealth is just expensive geometry.
The drone wave makes this unavoidable. HAL’s CATS Warrior, America’s CCA, Europe’s remote carriers, China’s loyal wingmen, all bet on the same idea: a pilot leading a small pack of machines. Conceptually neat. Operationally brutal.
No human can manage that traffic alone when links flicker and jamming climbs. You need software that throws away 99 percent of the noise, ranks the remaining one percent, and tells the pilot “shoot this, ignore that” in under a second. If that code fails, your loyal wingman becomes an expensive decoy.
Building a Robust Ecosystem for India’s Military AI
Delhi is waking up. DRDO now puts AI, machine learning and swarming at the centre of its roadmaps. ADA has to integrate it into AMCA from day one. HAL is pushing CATS. BEL and Data Patterns are scaling mission systems. Outside the tent, NewSpace, Grene Robotics, Raphe mPhibr are actually building autonomy stacks instead of PowerPoints.
Good start. Not enough.
The comforting rebuttal you hear in South Block is that fighters do not run on foreign chatbots, they run on certified boxes. True. Also irrelevant. Strategic dependence is not about a single chip. It is about the pipeline behind it. Who trains the models on Indian terrain data. Who owns the simulation environments. Who can re-validate the code after an adversary spoofs your seekers over eastern Ladakh. Who can push an update in weeks, not years, without calling a vendor abroad.
Civilian AI can hallucinate and apologize. Military AI gets people killed when GPS is denied and sensors lie. That gap demands test ranges, red teams, cybersecurity baked in, and a cadre of engineers cleared to work on classified data for a career, not a contract.
We have seen this movie with jet engines. We built airframes while waiting decades for propulsion sovereignty. We cannot afford a software repeat.
Because the next air war will not be platform versus platform. It will be network versus network. The jet that processes faster, shares cleaner, and learns quicker will dictate the fight. Speed will matter less than sense-making.
Make in India fixed the metal problem. It will not fix the brain problem. That needs a different policy muscle: long funding for software, not just prototypes; procurement that buys code iterations, not just boxes; a certification regime that moves at software speed; and a domestic data and compute base the services actually trust.
Otherwise we will do what we have done before. We will paint the roundel on a superb Indian airframe, then quietly rent the intelligence that flies it.
The tail will be ours. The question is whether the mind inside it will be.
