Boost Robotics Is Sending Actual Robots Into Data Centers Because Humans Can't Keep Up

The Macro: Data Centers Are Growing Faster Than the People Who Maintain Them

I have been watching the data center buildout accelerate for two years now and the staffing problem is getting worse, not better. Every hyperscaler, every AI lab, every cloud provider is racing to deploy more compute. Microsoft alone plans to spend $80 billion on data center capacity this year. Google, Amazon, Meta, all the same story. The physical infrastructure is scaling at a pace that the labor market simply cannot match.

The inspection problem is particularly ugly. Data centers need constant monitoring. Thermal hotspots, cable integrity, hardware failures, environmental conditions. The industry standard is manual walkthroughs. A technician with a clipboard, or maybe a tablet, walking rows of server racks and checking things by hand. This worked fine when data centers were smaller. It does not work when your facility is the size of a small town and you need to check every rack every 48 hours to maintain your SLA commitments.

When inspections slip, two expensive things happen. First, insurance premiums go up. Underwriters price risk based on inspection frequency, and if you can’t prove you’re maintaining coverage, you pay more. Second, SLA violations pile up. Missed thermal events turn into hardware failures, which turn into downtime, which turns into penalty payments that can run into the millions. The math is straightforward: inspect more often, pay less in losses. The constraint is that there aren’t enough humans available to do the inspecting.

Robots are the obvious answer. But “obvious” and “easy” are very different words in the robotics industry. Fixed sensors and basic monitoring drones have existed for years. What hasn’t existed is a robot that can move through a data center autonomously, manipulate physical objects, and actually do something useful beyond taking pictures.

The Micro: Two CMU Robotics Grads Who Met on a Dance Team

Boost Robotics is building autonomous mobile manipulation robots for data centers. Not just rovers with cameras. Robots that can physically interact with infrastructure. The pitch is that these machines handle inspection and maintenance tasks that currently require a human technician to walk the floor, increasing inspection frequency from whatever your current schedule allows to something approaching continuous.

The founding team is Hardik Singh and Hans Kumar. They met at Carnegie Mellon in 2015, not in a robotics lab but on a competitive dance team, which is a detail I find genuinely charming. Both went deep into robotics after that. Hardik earned his MS in Robotic Systems Development from CMU’s Robotics Institute and led robotics at an agtech startup. Hans spent time as a Staff Software Engineer at Boston Dynamics, specifically on computer vision for the Spot robot, and also built GPS-denied drone autonomy systems. Between them, they’ve been working on autonomous systems for a decade.

The Boston Dynamics connection matters here. Hans didn’t just study robotics in a lab. He shipped production autonomy on one of the most well-known commercial robots in the world. That’s a very different resume line than “built a research prototype that worked in simulation.” He knows what it takes to make a robot function reliably in messy real-world environments, which is exactly the hard part of this problem.

They’re backed by NVIDIA’s Inception Program, which signals that the GPU side of the industry sees value in what they’re building. The team is currently two people in Boston, part of YC’s Spring 2025 batch.

The product is early. The website is deliberately sparse. There’s no pricing page, no feature matrix, no customer logos. What you get is a clear statement of what they’re building and evidence that the team has the right background to build it. For a hardware robotics company at this stage, that’s about what I’d expect. Nobody ships a polished marketing site while they’re still building the robot.

The Verdict

I think Boost Robotics is positioned at a genuinely interesting intersection. The demand side is enormous and growing. Data center operators need more inspection capacity and the labor market can’t provide it. The technology side is finally mature enough to make autonomous mobile manipulation viable outside of research labs. And the team has exactly the right credentials for this specific problem.

The competition in data center robotics includes companies like Gecko Robotics, which does infrastructure inspections with wall-climbing robots, and Spot from Boston Dynamics, which is already deployed in some facilities for basic monitoring. Symbotic and Locus Robotics dominate warehouse automation but haven’t crossed into data center territory. The opening for Boost is the manipulation piece. Most existing solutions are sensors on wheels. If Boost can build a robot that physically interacts with server hardware, that’s a step function improvement over the current options.

The risk is execution timeline. Hardware companies burn cash while iterating on physical products. Two people and seed funding is enough to build a prototype, but manufacturing, testing, and deploying robots in critical infrastructure environments takes time, capital, and patience. Data center operators are also extremely conservative buyers. They don’t let unproven hardware near their production environments without months of validation.

Thirty days from now, I’d want to see a working prototype doing real inspections in a test environment. Sixty days, a pilot agreement with an actual data center operator. Ninety days, the question is whether the robot can operate unsupervised for extended periods without breaking something expensive. The market is there. The team is credible. Now they have to build the thing.