'You will miss the boat because it's coming': How the dawn of lights-out fabs will transform facilities operations

The UltraFacility 2025 Conference gave a glimpse into some of the behind-the-scenes work that is developing the next-generation lights-out facility – one where robots and agentic AI make operations more repeatable and less manual.

In 2026. Intel and suppliers will give a progress update in a pre-conference workshop ‘Autonomous Operations: The Future of Lights-Out Facilities.’

It’s a space that is changing fast, with the scale of incoming change not fully realized by most of the supply chain. As Ryan Kovalchick, Operational Technology Architect and Snr. Global Manager at Intel warned in 2025, 'you will miss the boat, because it's coming.'

Here is a recap of the 5 key takeaways from last year's workshop to prepare you for UltraFacility 2026 (16–18 September in Phoenix, Arizona).

1. AI won't be everywhere

Let's throw AI at everything? No. Intel was explicit: put the right technology in the right place, judged by ROI and operational impact. Robots and sensors cost money, so the first decision isn't "where do we add AI", it’s about which assets are worth instrumenting how you match fixed sensors, mobile robots and people to each.

The issue isn't too much data, but how little gets used. SEEQ put it at around 85% of data going untouched. Operational data is scattered across the historians and control systems that log every sensor, and a raw reading on its own means little. The real work is contextualising it.

2. Facilities are a focus area for semiconductor digital transformation

The fab gets the attention, but the facility systems underneath it are where much of the early lights-out work is happening.

The payoff areas are utility usage optimization, and predictive maintenance to reduce downtime.

Here are 3 applications where this technology is already practical:

• UPW was one of the first areas Intel looked at. An early Spot robot deployment from Boston Dynamics ran in the ultrapure water space. The need to reduce maintenance and repetitive rounds with many similar assets suits the robot and AI combo.
• Robots listening for compressed air and gas leaks. Spot’s acoustic payload – an array of subsonic microphones – listens for the specific frequencies of a leak on its rounds and flags it before it becomes a safety or a loss issue.
• Vibration anomalies on bearings and rotating equipment. Continuous listening can catch an upcoming bearing failure while there's still time to plan around it.
• Thermal imaging on pumps and motors. Temperature trends that tell you a work order is due before an unplanned fault.

3. It is a response to a shortage of people on unscalable, manual work

The whole program is a response to a workforce problem. Technician roles aren't being backfilled, and the knowledge retiring outsizes what's coming in. The available workforce in five years’ time may be a third of today's while the workload doubles, and you cannot hire across that gap.

• The tasks first in line are the repetitive, unscalable rounds: degassing, sampling, calibration and cleaning, alongside the daily gauge and temperature reads.
• The aim is to upskill technicians, not remove them: the same people will run the automation, maintain the robots, and use AI to work proactively rather than reactively.

Boston Dynamics' Dan Zuba made the core case for the robot: manual inspection rounds are inherently inconsistent, because each technician brings a slightly different read on when a motor is running warm or a bearing sounds off. Robots turn a subjective assessment into a consistent record. The constraint, as Zuba framed it, is rarely a lack of data – it is whether that data is good, and consistency is what makes it good.

4. It connects teams across fab and facility

One interesting result came from breaking silos. Facilities data has traditionally sat apart from fab data. SEEQ's example is process cooling water: once you can see fab tool-load demand, you can make far better decisions about the cooling-water loops that serve it, e.g., how to run the system, and when maintenance is genuinely due, instead of running facilities blind to what the fab is doing.

5. Call to action for vendors

Intel's key message was aimed at the supply chain. A fab cannot go lights-out on equipment that needs a person to manually get the data out of it.

• Choosing the right sensor for the job - instrumenting for the context-rich decisions you want to make is important.
• Plug and play - Systems should be instantly online, connected and flowing. Vendor-to-vendor incompatibility, with every supplier exposing its data differently, or not at all, is a direct blocker. In a lights-out facility you can't accept equipment that only works if you hire people to tend it.
• Physics-based principles - Physics-based principles are more important than having AI-integrated equipment, so that end-users can easily integrate equipment into their own digital architecture.
• Clean, procedures and failure codes - Not a mystery buried in a manual. This is what lets analytics (and eventual all-purpose robots) act on a fault without a person translating it first.

Amidst the industry’s largest buildout, understanding how the next generation of facilities will look very different to today’s is critical for your company’s strategy. Find out more at the workshop on September 16 2026 in Phoenix. Book your conference place for access.