Who keeps the systems running

Everyone's Debating AI. Nobody's Talking About the People Keeping It Running.

I recently returned from the Photonics Research Conference at Stanford, where much of the discussion focused on optical interconnects and how photonics may help scale the next generation of AI infrastructure.

The consensus was clear: AI's growth will require new approaches to moving data, reducing power consumption, and increasing compute density.

But as I listened to those conversations, I found myself thinking about a different scaling challenge.

We spend enormous time discussing GPUs, power grids, cooling systems, and model architectures.

We spend far less time talking about the people who keep those systems running.

Hard drives fail. Power supplies fail. Cooling systems fail. Network equipment fails.

At hyperscale, these aren't occasional events. They're daily operational realities.
Someone has to identify the issue, locate the equipment, perform the repair, validate the fix, and get systems back online.

While policymakers, researchers, investors, and industry observers debate how quickly AI should advance, operators face a more immediate question:


Can the industry recruit, train, and support enough skilled people to keep AI infrastructure running as demand continues to accelerate?


Without them, AI doesn't scale.


Behind every new model release is an army of technicians performing rack-level repairs and component replacements.

Behind every hyperscale expansion are engineers maintaining power, cooling, networking, and physical infrastructure.

Behind every uptime metric are operations teams monitoring thousands of systems around the clock.

Many of these teams are already working at full capacity.

Full disclosure: I'm the CEO of Optaric, so I'm not a neutral observer. But this is exactly the problem we built the company to solve.

Smart glasses matter because they bridge the gap between frontline technicians and remote experts. Instead of relying on a radio call and a printed runbook, technicians can receive real-time guidance and hands-free access to procedures while experts see exactly what they see and help resolve issues faster.

The AI conversation usually begins with GPUs, power grids, and breakthrough models.

It should also include the people keeping the physical layer standing.

Because the next breakthrough in AI may not come from a larger model.

It may come from helping the people who keep the infrastructure running at scale.

My thanks to Armand Niederberger for organizing such a thought-provoking conference, and to DP Prakash for the conversations that helped shape these ideas.dsgfaeryusrg