Gather AI, maker of ‘curious’ warehouse drones, lands $40M led by Keith Block’s firm 

Gather AI, a startup that offers an AI level for warehouse cameras and drones, has raised a $40 cardinal Series B backing circular led by Smith Point Capital. That’s the VC steadfast founded by erstwhile Salesforce co-CEO Keith Block. 

The Gather squad archetypal met Smith Point a twelvemonth agone astatine a logistics conference, and “it took Keith and his squad 5 minutes to get what we’re doing,” co-founder and CEO Sankalp Arora told TechCrunch. 

What Gather AI is doing is unusual. The 4 founders met arsenic PhD students astatine Carnegie Mellon University, wherever they built 1 of the archetypal autonomous helicopters and tested it connected the FBI grooming grounds successful Quantico. (Block is simply a trustee for CMU.) 

In 2017, the founders took what they learned astir teaching helicopters to alert and onshore safely and launched Gather AI. Using off-the-shelf cameras placed connected strategical moving instrumentality similar forklifts, arsenic good arsenic off-the-shelf drones flying astir the warehouse, the cameras ticker on-the-floor operations and log what they find into the warehouse absorption systems.  

But the drawback is, the AI isn’t being random astir what it scans. It is being “curious,” arsenic Arora described it.

“My PhD enactment focused connected however to marque antithetic kinds of flying robots curious,” helium said. “So they’re funny astir boxes and barroom codes and workflows.” 

In summation to barcodes, they look for batch codes, text, expiration dates, lawsuit counts, damages, occupancy, and different items. The thought is that they volition observe and foretell issues similar debased inventory, misplaced stock, and workflows that whitethorn origin information issues. 

They besides enactment successful environments unfriendly to people, similar freezers and acold storage. 

Because Gather’s underlying tech was built years earlier the property of ample connection models, this is not the benignant of AI that an LLM uses. 

“They’re not end-to-end neural networks,” Arora explains. “They are classical Bayesian techniques, combined with neural networks.”  

AI imaginativeness Bayesian techniques usage probability-base methods to thatch computers however to construe ocular data. These systems let the exertion to larn by utilizing information and anterior cognition to marque decisions — meaning they don’t endure the hallucination problems of LLMs. 

Instead they “get curious,” arsenic Arora put it, to stitchery accusation (hence the startup’s name) and marque a determination connected the adjacent enactment based connected what they’ve learned. 

As old-school arsenic that sounds, Gather AI is sitting astatine the borderline of the adjacent large happening successful AI, sometimes called “embodied AI.” These are robots that interact with the existent world, arsenic opposed to an LLM interacting via machine chat oregon web app.  

To that end, successful December, the startup won the 2025 Nebius Robotics grant for Vision AI and Streaming Video Analytics. (Nebius is simply a Netherlands institution that provides AI infrastructure.)

Gather presently employs astir 60 people, Arora said, and customers see Kwik Trip, Axon, GEODIS, and NFI Industries. With this caller funding, the startup has present raised $74 cardinal total. Other investors see Bain Capital Ventures, XRC Ventures, and Hillman Investments.