SwitchOn Raises $8 Million to Take Its Factory-Floor 'Physical AI' Global
Bengaluru''s SwitchOn has raised $8 million in a pre-Series B round led by IvyCap Ventures to scale its edge-based computer-vision systems, which catch manufacturing defects at production-line speed on four continents.
Manik Gupta
Founder and editor of DeepTech India. Manik writes about India's frontier technology ecosystem — AI, semiconductors, space, quantum, robotics and biotech — translating research and policy into clear, reliable reporting.
The Round
SwitchOn, the Bengaluru-based "physical AI" startup formally known as Abee Research Labs, has raised $8 million (about ₹78 crore) in a pre-Series B round led by IvyCap Ventures, with participation from SIG Tattva and Trifecta Capital. Founded by Aniruddha Banerjee and Avra Banerjee, the company said the fresh capital will fund international expansion, deepen its research into next-generation physical AI, and scale its go-to-market operations.
The raise follows a $1.1 million seed and a $4.2 million Series A, and reflects a steady, if unflashy, thesis: that some of the most immediately useful AI is not a chatbot but a camera on a production line that never blinks.
What It Actually Does
SwitchOn embeds artificial intelligence directly into manufacturing equipment to automate quality inspection through an edge-native platform it calls VisionAI. Its flagship product, DeepInspect, uses edge-based computer vision to spot surface defects with sub-150-micron precision at line speeds exceeding 1,200 products a minute — fast enough to keep pace with high-throughput plants where human inspectors simply cannot.
"Edge-native" is the operative phrase. Rather than shipping every frame to the cloud for analysis — which adds latency, cost and a dependence on connectivity — SwitchOn runs its models on hardware sitting next to the line. For a factory, that means decisions in milliseconds and defect data that never has to leave the building.
The Traction
Since deploying its first production systems in 2021, the company says it now powers AI-driven inspection across more than 170 production lines in over 60 manufacturing facilities on four continents. Its customer list reads like a roll-call of industrial heavyweights — Unilever, Bosch, Maruti Suzuki and ALPA among them — spanning consumer packaged goods, electronics, automotive and pharmaceuticals.
That breadth matters. Quality inspection is one of the few AI use-cases with a clean, measurable return: fewer defective units shipped, less waste, less rework. It is also a market where an Indian company selling to global manufacturers is exporting genuinely hard technology, not just services.
Why It Fits the Moment
"Physical AI" — the application of machine learning to sensors, robots and machines that act in the real world — has become one of the more durable threads in India''s deep-tech funding story, alongside larger raises for robotics and industrial-automation startups. SwitchOn''s pitch is that inspection is the wedge: a task that is repetitive, precision-critical and expensive to get wrong, and therefore an ideal first job to hand to a machine that can see.
With fresh capital and a customer base already spread across four continents, the company''s next test is scale — turning a proven product into something closer to a global standard for how factories check their own work.
Sources
Tags
More from Artificial Intelligence
TCS and NVIDIA Open a Bengaluru Lab to Drag Industrial AI Out of Pilot Purgatory
TCS opened the Autonomous Engineering Lab, Powered by NVIDIA, in Bengaluru on 15 July 2026 — billed as India's first 'physical AI' hub, built to move industrial-AI and autonomous-mobility projects from pilots into production.
Emergent Becomes India's Third AI Unicorn as Its 'Build-an-App-by-Chatting' Platform Hits $1.5 Billion
Bengaluru's Emergent, whose AI agents build apps from plain-language prompts, raised $130 million at a $1.5 billion valuation on 15 July 2026 — a fivefold jump in six months that makes it India's third homegrown AI unicorn after Krutrim and Sarvam.