India's Startups Make Their Move on Physical AI Robots

For the past two years, India's artificial-intelligence story has played out almost entirely on screens — chatbots, language models and software copilots. A quieter shift is now underway. A cluster of Indian startups is betting that the next phase of AI belongs in the physical world: machines that can see, reason and act, from industrial robot arms to humanoids and quadrupeds. The market for these "physical AI" systems is projected to grow into a $100 billion-plus opportunity by 2030, and Indian companies want a share of it.

The ambition is real, but so is the gap. India has world-class software talent and a growing robotics research base, yet it has historically lagged in the hardware, precision manufacturing and capital intensity that physical AI demands. What is changing is the combination of maturing local engineering, fresh venture capital and an unexpected geopolitical opening.

The companies leading the push

Several names are emerging at the front of this wave. Bengaluru-based Ati Robotics, backed largely by Silicon Valley investors and having raised around $37 million, is building physical AI-powered robots and has floated plans to set up a manufacturing plant in Michigan to serve the US market early next year — an unusually outbound strategy for an Indian robotics startup.

Addverb Technologies, in which Reliance Retail Ventures holds a majority stake of roughly 55%, is testing physical AI robot prototypes aimed at both industrial and defence applications and is readying them for commercial deployment. Addverb already has a track record in warehouse automation, giving it a base of real-world deployments to build on.

Gurugram's Novus High-Tech Robotic is working on humanoids and quadrupeds for industrial use, with a deliberate emphasis on software that is entirely "made in India." The company has signalled that its robots will move into pilot deployments soon.

Together, these firms sketch the shape of India's early physical AI sector: a mix of venture-funded newcomers and better-capitalised players attached to large industrial groups, spanning robot arms, humanoids and four-legged machines.

Money is starting to follow

The capital picture is improving, if still modest by global standards. Startups in India's physical AI ecosystem have raised roughly $155 million across about 31 deals in 2026 so far, according to funding trackers — a small figure next to the billions flowing into the category worldwide, but a meaningful jump for a segment that barely registered a couple of years ago.

Much of that money is arriving at the pre-seed and early stages, funding prototypes and pilots rather than mass production. That reflects where India sits on the curve: proving that its robots work in the field, before proving they can be built and sold at scale. The financing gap that plagues Indian deep tech more broadly — plenty of early capital, far less at the growth stage — will test these companies as they try to move from demonstrations to factories.

The China factor

An external development has handed Indian robotics an opening. Recent moves in the United States to restrict imports of humanoids and quadruped robots from China on security grounds have created space for suppliers from other countries. For Indian firms, that is both a market opportunity and a positioning advantage: a robot whose hardware and, crucially, whose software are built outside China can be an easier sell to Western customers wary of supply-chain and data-security risks.

Novus's stress on fully India-made software and Ati's plan to manufacture in the US both read as bets on this shift. In a sector where trust and provenance increasingly matter as much as raw capability, "not made in China" has become a genuine differentiator.

What still has to be proven

The obstacles are substantial. Physical AI is brutally hard: robots must perceive messy real-world environments, manipulate objects reliably and operate safely around people, all while being cheap enough to justify their cost. India's manufacturing ecosystem for precision components, actuators and sensors is still developing, and much of the high-end hardware is imported.

Talent is another constraint. Building capable physical AI systems requires a blend of machine learning, mechanical engineering, controls and systems integration that few teams anywhere have mastered. India has strength in the software half of that equation and is racing to build the rest.

Then there is the question of customers. Demonstrations impress, but sustainable businesses need buyers willing to deploy robots at scale — in warehouses, factories, defence and beyond. Addverb's industrial base and Reliance's backing give it an advantage here; younger startups will have to win that trust from scratch.

A frontier worth watching

None of this guarantees that India will produce a physical AI champion. But the ingredients are assembling: capable startups, rising capital, a supportive policy environment for deep tech, and a global market that is actively looking for alternatives to Chinese robotics. After years of India's AI ambitions being confined to software, a serious attempt to build machines that act in the physical world is now underway — and the next couple of years of pilots and production decisions will show how far it can go.

Sources

• Business Standard — "Indian startups bet on physical AI robots as global market takes shape": https://www.business-standard.com/companies/start-ups/indian-startups-bet-on-physical-ai-robots-as-global-market-takes-shape-126081700773_1.html
• Entrackr — "Indian Physical AI attracts $155 Mn in 2026 as robotics, real-world data gain traction": https://entrackr.com/report/indian-physical-ai-attracts-155-mn-in-2026-as-robotics-real-world-data-gain-traction-12210766