01 / Who they are nowThe quiet giant of physical AI
Walk into a Spot AI customer site - a tire shop in Texas, a Tier-1 supplier in Ohio, a school district in Florida - and the cameras look ordinary. They are not. Somewhere on the local network, an Intelligent Video Recorder hums along, running models that flag a worker without a hard hat, a forklift driving too fast, a checkout line ballooning past three deep. By the time a human notices, Spot AI's software has already pinged a manager's phone.
That is the company in 2026: about 130 people, headquartered in San Francisco, roughly $95 million raised, deployed at more than a thousand businesses across 17 industries. It is not the loudest AI story in the Bay. It might be the most installed.
02 / The problem they sawA billion cameras, almost no intelligence
The dirty secret of physical security: nobody watches the footage. The cameras roll, the disks fill, the tapes get overwritten. Insurance companies want them. IT departments tolerate them. Almost no one looks at them - unless something goes wrong, and even then, it's usually a manager scrubbing through hours of grainy black-and-white at 3 a.m.
The world has, by industry estimates, well over a billion business cameras already installed. Most are running. Most are recording. Most are functionally blind. Spot AI saw the gap the way a plumber sees a leak: not as a tragedy, but as a fix waiting for a wrench.
03 / The founders' betThree Stanford friends, one stubborn idea
Rish Gupta, Sud Bhatija, and Tanuj Thapliyal met as students at Stanford. They didn't start by writing a manifesto. They started by asking a simpler question: why is video surveillance the worst-served slice of enterprise software? Their backgrounds - Cisco Meraki on the cloud-networking side, Samsara on the IoT side - gave them a hunch that the answer was distribution, not invention. The models were almost ready. The cameras were already there. Nobody had bothered to glue them together.
The bet, formalized in 2018: build a cloud-managed camera system that any business could install in ten minutes, then quietly stuff it with computer vision until customers realized they were running an AI workforce that happened to look like a security setup. Bessemer wrote the seed check. Redpoint led the Series A. Scale Venture Partners came in for the $40 million Series B in late 2022. By October 2024, Qualcomm Ventures had joined a $31 million extension, taking the company near $100 million raised.
The founders, briefly
Rish Gupta - Co-Founder & CEO. The face on the keynote stages. Background in software and product.
Sud Bhatija - Co-Founder & COO. Engineering and sustainability roots, an MBA detour at Stanford.
Tanuj Thapliyal - Co-Founder. The third corner of the founding triangle, with the same Stanford-to-startup arc.
04 / The productA camera, a brain, an agent
The Spot AI stack has three layers. There are the cameras themselves - NDAA-compliant, important if you want to sell into the U.S. government, a fact lost on no one. There is the Intelligent Video Recorder, an edge appliance that lets the system swallow whatever third-party cameras are already on the wall. And there is the cloud dashboard, which is where the work shows up: searchable footage, multi-site views, semantic search ("show me every time the warehouse door was propped open last quarter") and, since October 2024, Video AI Agents.
The agents are the part that changes the conversation. Earlier products surfaced incidents. Agents resolve them. A line at the cash wrap grows past four people; an agent pings the floor manager. A worker steps inside a robot's safety cage; an agent halts the cell. A loiterer paces the back lot at 2 a.m.; an agent triggers an outdoor speaker and escalates to monitoring. The cameras stopped being passive a long time ago. Now they have opinions.
AI Camera System
Cloud-managed cameras with computer vision built in. NDAA-compliant.
Intelligent Video Recorder
Edge appliance that plugs into existing camera infrastructure - no rip-and-replace.
Cloud Dashboard
Multi-site monitoring, role-based access, semantic search across feeds.
Video AI Agents
Agentic AI that watches, alerts, deters, and escalates in real time.
A short history of cameras that learned to think
05 / The proofCustomers, dollars, and the hours that piled up
It is one thing to claim a category. It is harder to staff it with paying customers. Spot AI has, by its own count, more than a thousand of them - across manufacturing, retail, education, healthcare, logistics, construction, auto services, and government facilities. The investor list reads like a who's-who of late-stage software: Scale, Redpoint, Bessemer, StepStone, Qualcomm Ventures. The hardware-meets-software thesis usually scares off pure software VCs. Spot AI's pitch deck apparently does not.
The numbers move in the right direction. Roughly $47.5 million in annual revenue, by external estimates. Roughly 130 employees. Roughly $95 million raised. One billion-plus hours of business video flowing through the platform - a number that is interesting less for its size and more for its compounding rate.
Spot AI funding, round by round
06 / The missionEvery camera, useful
The internal phrasing has shifted over the years. The earliest tagline reportedly read, with admirable bluntness, "turning dumb cameras smart." The current framing is gentler: video intelligence for safety, security, and operations. The thing being sold has not changed. The thing being promised has matured.
What that mission looks like in practice depends on the customer. A logistics yard wants to know when a truck has been idling too long. A school wants to know when an exterior door is propped open after hours. A factory wants to know - immediately - when a worker has wandered into a hazard zone. Spot AI's claim is that one platform can do all of it, because the underlying technology is the same: pixels, models, alerts, agents.
07 / Why it matters tomorrowThe operating system for things that already exist
The most interesting startups don't always build new infrastructure. Some build software on top of infrastructure that is already there - paid for, mounted to walls, plugged into outlets, mostly idle. The world's billion business cameras are an installed base in search of a brain. Spot AI's wager is that the brain is finally cheap enough, fast enough, and small enough to fit. Qualcomm's investment is, among other things, a statement about which silicon thinks that's true.
The category has competition - the Vergeses and Rhombuses of the world have their own takes - but the customers Spot AI lists are not theoretical, and the agents it has shipped are not slideware. The next twelve months will tell whether agentic video is a feature or a market. The founders, predictably, think it's a market.
08 / Back to the floorThe tire shop, two years later
Return to where we started. The tire shop in Texas. The school district in Florida. The supplier in Ohio. The cameras still look ordinary. The cameras have always looked ordinary. The difference is what happens after the shutter clicks - the model running, the agent watching, the manager getting the ping before the incident becomes an incident. Spot AI did not invent the camera. It changed the camera's job description. Sometimes that is the more interesting move.