Capital

9,437 Checkouts Just Learned to Spot Theft Together

Edgify raised EUR7.7 million ($9 million) to run loss-prevention AI on checkout terminals, cameras and scales retailers already own, using federated learning so store video stays on site.

CapitalBy Servola Tech Desk2026-08-114 min read

AI-assisted, edited by humans. | Editor-in-Chief: Leon Soliman | Editorial standards

9,437 Checkouts Just Learned to Spot Theft Together

Key takeaways

Edgify, a London-based edge-AI company, closed a EUR7.7 million ($9 million) Series A+ round on August 10, led by Rank Ventures and Mangrove Capital Partners, taking its total funding to $25 million and its live network to 2,042 stores and 9,437 devices, all running on hardware retailers already own.
The system uses federated learning: each store's cameras, scales and self-checkout terminals train a shared model locally, and only the model updates, not raw video, leave the building - CEO Nadav Israel calls it a fleet of machines that see, decide and learn together without a byte leaving the site.
That architecture is the real pitch, not the theft-detection accuracy: rivals that depend on dedicated server rooms can take months to install in a single store, while Edgify's software runs on tills, cameras and scales the retailer is already using, which is a budget and timeline decision an owner can act on this quarter.
The company is hardware-agnostic and already integrates with equipment from Zebra Technologies and Bizerba, and plans to push the same edge model beyond grocery into quick-service restaurants, apparel, distribution centres and manufacturing.
Global retail shrinkage is estimated at $386 billion a year against a $15.8 billion computer-vision market serving it, which is the gap Rank Ventures partner Rajan Dosanjh is betting on when he says the winners will own the point where the data is created, meaning the store floor, not a data centre.

A Round Built on Hardware Retailers Already Own

Edgify, a London-based company that sells edge AI software for retail loss prevention, said on August 10 that it had closed a EUR7.7 million ($9 million) Series A+ round led by Rank Ventures, with continued backing from Mangrove Capital Partners, which has funded the company since its 2020 seed round. The raise takes Edgify's total funding to $25 million and follows a run of growth the company says now covers 2,042 stores and 9,437 devices running its models live.

What makes the round worth a second look is not the size, which is modest next to the mega-rounds crowding AI headlines this year, but the architecture it funds. Edgify's software runs directly on the cameras, scales and self-checkout terminals a store already has installed, rather than requiring a dedicated server room or a new cloud contract. Chief executive Nadav Israel described the pitch to TheNextWeb this way: 'A store is a fleet of machines that can see, decide and learn together, without a single byte leaving the building. Retail is just the beginning.'

Why the Model Trains Locally Instead of in the Cloud

The mechanism behind that claim is federated learning. Instead of streaming video from every store to a central server for analysis, each store's devices train a shared detection model on their own local footage, then send only the model's learned parameters back for aggregation, never the underlying video itself. The practical effect is that a retailer's camera and till footage stays on site rather than being uploaded to a third-party cloud for processing.

For an EU or UK retailer, that distinction is not academic. In-store cameras and point-of-sale systems already sit close to personal-data rules under GDPR, and any architecture that keeps raw footage local rather than centralizing it in an external cloud changes what a retailer has to account for when a data-protection officer asks where customer and staff footage actually goes. Edgify is not a compliance product, but the design choice narrows one part of that conversation before it starts.

The Edge Is Deployment Speed, Not Detection Accuracy

Loss-prevention computer vision is not a new category; large retailers have run centralized video-analytics systems for years. What has kept it out of reach for mid-market grocers, convenience chains and independent operators is installation: a server-dependent system typically means new rack hardware, a network redesign and a rollout that can take months per site, which only a chain with a dedicated IT budget can absorb.

Edgify's pitch inverts that cost. Because the software runs on tills, cameras and scales the retailer already owns, the company is positioning itself against that installation timeline rather than against a rival's detection algorithm. That is a genuinely different sales argument, and it is the one an owner evaluating loss-prevention tools should actually be pricing: not 'how accurate is the theft detection' but 'how many months of downtime and capital does the rollout cost me.'

What This Means Beyond the Supermarket Aisle

Edgify says it plans to extend the same edge-native model beyond grocery into quick-service restaurants, apparel retail, distribution centres and manufacturing, and it is already hardware-agnostic, integrating with point-of-sale and scanning equipment from Zebra Technologies and Bizerba rather than requiring its own proprietary devices. For an owner in any of those sectors already running standard checkout, camera or scanning hardware, that is the detail worth noting: the barrier to trying this category of tool is no longer a server-room investment.

The market context helps explain why investors are willing to back a company at this stage: the retail computer-vision market is put at roughly $15.8 billion, against a global retail shrinkage problem estimated at $386 billion a year. Rank Ventures partner Rajan Dosanjh framed the bet plainly: 'The winners will own the point where data is created.' For Edgify, that point is the checkout lane, not a data centre.

Frequently asked questions

What does Edgify actually sell?

A software subscription that runs loss-prevention and theft-detection models directly on a store's existing cameras, scales and self-checkout terminals, rather than a hardware product or a centralized video-analytics service.

Does federated learning mean no store data ever leaves the building?

It means the raw video stays local while each store trains a shared model on-site; only the model's learned updates are sent back for aggregation, not the underlying footage - a materially different data flow than uploading video to a central cloud.

Who backed the round and is this Edgify's first raise?

Rank Ventures led the EUR7.7 million ($9 million) Series A+, with continued participation from Mangrove Capital Partners, which has backed Edgify since its 2020 seed round; total funding now stands at $25 million.

The theft-detection model itself is not the scarce part of this deal; retailers have been able to buy computer-vision loss prevention for years. What Edgify is actually selling is the removal of the server room from that purchase decision, and investors just paid $9 million to bet that the store floor, not a data centre, is where retail AI actually gets deployed.

EdgifyRank VenturesMangrove Capital Partnersretail AIloss preventionedge computing

Boeing Takes 16.5% of Archer, Exits Air Taxis

Archer Aviation is buying Wisk Aero, SkyGrid and Insitu from Boeing in an all-stock deal leaving Boeing with roughly 16.5 percent of Archer - three years after Wisk sued Archer for stealing its technology.

4 min read

Capital

Intel Raises $15 Billion, First Stock Sale Since 1971

Intel is raising $15 billion in new stock, its first public share sale since 1971, to fund AI compute capacity even a doubled share price could not cover alone.

3 min read

Decisions

The Third Pivot Was the Warning Magic Leap Missed

Magic Leap filed a WARN notice for 193 layoffs effective 1 October 2026, its fourth strategic pivot after $4 billion raised. The lesson is the pattern, not the pivot.