However there was a catch. Every scan, roughly 100 MB in measurement, needed to be uploaded to the cloud, processed and the outcomes despatched again. This cycle was time consuming due to web connectivity points and energy outages, widespread to rural areas. It resulted in an extended line of sufferers ready for outcomes. The delayed scan stories made it tougher to isolate a affected person or begin remedy in time.
This was a reminder of the mismatch between cutting-edge AI and the bottom realities it was constructed for.
“When you’re constructing healthcare options for the International South, you need to assume the testing situations are removed from splendid. The web will probably be patchy. Electrical energy provide will probably be irregular. Machines must have longer battery lives to final at the least a day,” mentioned Prashant Warier, co-founder of Qure.ai.
So, the staff discovered a option to re-engineer their system. As an alternative of directing each scan via the cloud, they embedded the AI mannequin immediately onto native laptops connected to X-ray tools, making the diagnostics system work even in offline mode. These laptops with ‘edge AI’ chips dealt with the heavy computing wanted to analyse the scans on the spot, in beneath 20 seconds.
At present, Qure.ai companions with greater than 15 medical gadget makers to embed AI immediately into edge gadgets, making superior diagnostics sensible in locations the place steady web connectivity remains to be a luxurious. In 2025, the corporate introduced it had crossed 5 million AI-enabled chest X-rays throughout 20 international locations in Asia, Africa, the Center East and Latin America.
The shift to ‘offline-first AI’ is gaining momentum in India and discovering use circumstances past healthcare, throughout e-scooters, surveillance programs, sensible meters and authorities companies, the place latency, energy constraints or patchy networks make cloud AI unreliable.
Not like cloud-based AI, which is dependent upon distant servers to course of information, edge AI has task-specific chips to allow real-time choices—they needn’t contact distant information centres.
As an example, a surveillance digital camera in a distant space that should detect intruders immediately can’t afford to depend on importing footage to a distant information centre as a way to take fast motion. It should course of that information domestically to set off real-time alerts. Or in railway security programs, edge-enabled cameras mounted at crossings or on prepare engines can immediately detect obstructions on the tracks like stalled autos or pedestrians and set off alerts with out ready for cloud-based processing to take motion.
To be clear, edge AI is just not about changing the cloud totally. It’s about shifting crucial intelligence nearer to the issue.
Made in India
On the coronary heart of this motion is a brand new breed of Indian chip designers.
A handful of Indian founders, lots of whom have spent years in international semiconductor companies or high analysis labs, at the moment are constructing fabless chip startups—the place design is finished domestically and fabrication outsourced to foundries overseas—from the bottom up. The fabless mannequin is the worldwide business norm, adopted even by firms like Apple, which designs its chips within the US however manufactures them overseas. The mannequin permits startups to concentrate on innovation with out the heavy capital prices of operating fabrication crops.
BigEndian Semiconductors, based by IIT alumnus Sunil Kumar, is now constructing vision-processing chips for surveillance, automotive and industrial use. Shashwath T.R. and Sharan Jagathrakshakan, co-founders of Mindgrove Applied sciences, and each IIT Madras alumnus, are designing ultra-low-power system-on-chips (SoCs), meant to be used in linked home equipment, sensible meters and different cost-sensitive gadgets. Netrasemi, based by former Intel engineers, is growing a homegrown 64-bit RISC-V processor, concentrating on use circumstances in shopper electronics and embedded AI.
Collectively, these initiatives signify a rising pool of vastly skilled engineers betting that India holds an opportunity within the AI {hardware} sport, via the sting AI route. That is the place India’s semiconductor journey would possibly lastly discover a foothold.
Why it issues
The shift towards edge chips, whereas international superpowers interact in AI wars, may very well be a strategic realignment in India’s {hardware} journey.
Whereas cloud chips, corresponding to Nvidia’s A100s and H100s, are used to coach and run huge AI fashions corresponding to ChatGPT or Gemini, they’re costly and extremely depending on infrastructure. India imports almost 100% of those high-end chips and regardless of latest authorities GPU leasing initiatives, it stays far behind in constructing foundational AI infrastructure.
Edge chips, against this, are narrower in perform however extra suited to India’s speedy wants. They don’t require huge information centres and will be deployed broadly within the subject. And whereas the worldwide chip market is consolidated amongst giants corresponding to Qualcomm and NVIDIA, the sting AI house leaves room for brand new entrants fixing extremely localized issues.
“India is essentially a low-bandwidth, power-constrained, cost-sensitive market, so the standard cloud-first AI stack was by no means going to scale right here,” mentioned Vishesh Rajaram, managing accomplice on the deeptech-focused, early-stage fund Speciale Make investments. “India doesn’t simply want edge AI,” he added. “We’re among the finest sandboxes to construct it for the world.”
Final yr, Speciale Make investments co-led an $8 million funding spherical in Mindgrove Applied sciences, alongside Rocketship.vc.
Pranay Kotasthane, deputy director at The Takshashila Establishment, a assume tank, famous that India’s energy lies not in controlling uncooked chip provide chains however in its globally embedded expertise pool. By harnessing Indian design experience throughout the AI stack, notably in edge purposes, India can create dependencies which are geopolitically tough to disregard.
The Edge AI sport
We don’t have the financial muscle to construct hyperscale GPU infrastructure just like the US or China. So, we began the place India has a pure benefit: AI on the edge,” mentioned BigEndian Semiconductors’ Kumar. He had beforehand labored at Broadcom, Intel and ARM and in addition co-founded the SaaS unicorn Zenoti.
As an alternative of importing and assembling elements like many others within the ecosystem, Kumar and his staff selected to go deeper by designing their very own chips tailor-made to India’s most urgent and high-volume use circumstances in public infrastructure, together with surveillance programs, and defence and automotive purposes, the place real-time video or picture processing is crucial.
India’s video surveillance market is projected to succeed in $7.12 billion by 2030.
View Full Picture
“The chips powering surveillance cameras in India are virtually totally imported from a single Chinese language provider right this moment. We noticed a chance to not solely change them however make them smarter, safe and India-specific,” he mentioned.
BigEndian’s camera-focused chips are at present being examined by unique tools producers (OEMs) corresponding to Honeywell to be used in shopper safety programs, site visitors surveillance and industrial monitoring.
Whereas the chips are manufactured at a fabrication facility in Taiwan, the structure and design are developed in-house by BigEndian in India. Not like chips which are exported or rebranded, BigEndian’s semiconductors are shipped again to India as completed and tangible parts for native deployment. Kumar famous that the designs are particularly constructed for India’s situations to account for energy fluctuations, restricted connectivity and safety flaws generally present in overseas firmware.
“These aren’t generic processors. They’re tightly built-in stacks of safety firmware, AI engines, and sign processors shrink-wrapped right into a 5mm die,” he mentioned.
Mindgrove’s native chips
An identical precept guides Shashwath T.R., founding father of Mindgrove Applied sciences, which is constructing India’s first commercial-grade high-performance microcontroller (MCU).
“India has tens of millions of sensible meters, washing machines and mobility gadgets. These aren’t ‘sensible’ until they’ll run AI domestically. That’s what we’re fixing for,” mentioned Shashwath.
View Full Picture
Mindgrove’s chips will be domestically programmed, audited and secured—an necessary consideration within the surroundings of geopolitical chip vulnerabilities. The corporate’s first product, a general-purpose microcontroller designed totally in-house in Chennai, is now shifting into quantity manufacturing after finishing its prototyping phases final yr.
Shashwath says Mindgrove has already signed on a number of early OEM companions and the chip is predicted to indicate up in biometric authentication gadgets, sensible meters, robotics and even thermal printers over the subsequent yr.
These chips aren’t only for India, although. They’re designed round each Indian and worldwide specs, permitting them to serve related markets with shared situations like patchy energy, restricted compute and regulation-heavy gadget sectors.
In an period of rising concern over information privateness and digital sovereignty, the power to course of delicate information domestically from a sensible meter, biometric scanner or manufacturing facility sensor, can scale back technical and regulatory dangers. That’s a function India’s public infrastructure more and more desires.
Challenges inside and outdoors
Regardless of the rise of those startups, India’s {hardware} realities are rooted in systemic challenges. {Hardware} startups are up in opposition to a hostile surroundings, affected by challenges corresponding to low investor curiosity, lengthy timelines and provide chain gaps.
Final September, BigEndian Semiconductors raised $3 million in a funding spherical led by Vertex Ventures SEA & India.
“We first mapped all the worth chain,” Kumar mentioned. “Who builds what; the place the parts are sourced; what native OEMs need. Solely then did we begin designing.”
Kumar explains that not like software program, the {hardware} business leaves little room for trial and error. Errors or pivots price far more and growth could be very capital-intensive. This makes it a frightening house for each founders and traders, until there’s a transparent path to income. Consequently, many startups start with commercially viable purposes corresponding to industrial surveillance earlier than venturing into extra complicated and long-cycle domains corresponding to defence.
“Anybody attempting to truly do actual constructing is breaking their heads in opposition to an ecosystem that doesn’t exist,” mentioned Shashwath. Indian edge chip makers could make an actual distinction in the event that they promote within the International South, the place they’re at present up in opposition to bigger gamers.
However not like their international friends, in Taiwan or China, the place tight integration between chipmakers, foundries and design homes permits new iterations in as little as two weeks, Indian groups can take as much as six months to ship the identical replace.
The specialised software program and licenses required to design, take a look at and confirm chips are additionally prohibitively costly for Indian startups. Startups should rigorously ration software utilization amongst engineers (as a result of they’ll afford solely a restricted variety of licenses), even with assist from the Design Linked Incentive (DLI) Scheme, an initiative by the ministry of electronics and knowledge know-how. The scheme provides monetary and infrastructural assist to encourage home semiconductor design.
“We’re at all times rationing instruments,” Shashwath mentioned, “and it reveals in how lengthy we take to ship.”
We’re at all times rationing instruments, and it reveals in how lengthy we take to ship.
—Shashwath T.R.
Each Kumar and Shashwath consider that whereas schemes like DLI are useful, they don’t seem to be sufficient. What India’s semiconductor ecosystem actually wants is affected person capital, simpler entry to instruments and IP, extra home patrons and higher alignment between R&D and manufacturing.
The native electronics market in India is simply starting to heat as much as native chip makers. Regardless of a powerful base of electronics manufacturing in India, most OEMs at present depend on international chip suppliers.
“Many Indian OEMs nonetheless hesitate to modify from international incumbents until there’s a confirmed delta on price, efficiency or availability,” mentioned Speciale Make investments’s Rajaram. “Startups that supply a drop-in different with native assist, tighter integration and a roadmap win quicker.”
Kotasthane of The Takshashila Establishment argues that India ought to double down on fabless incentives, area of interest areas like compound semiconductors and even fund open-source AI accelerator tasks beneath the Nationwide AI Mission.
The upside is actual. If India will get edge AI proper, it could construct a sovereign know-how stack for itself, and for the world’s underserved areas.