Build-to-suit, colocation, and powered shell differ on who builds, who owns the plant, and how long you commit. For an inference deployment that will run for a decade or more, the choice is narrower than people think.
/ Resources
Insights.
How we see the market for AI inference capacity, with the sources.
/ Insights
How we screen a site: the criteria that matter
Screened is a word every developer uses. Here is what it means when we use it, the criteria a site has to pass, and how screened differs from qualified and optioned.
Latency, measured: why almost nobody publishes a number you can check, and why we will
Most operators say low latency. Few give you a number you can check, with a destination, a method, and a date. We will measure ours and report it that way.
You own land near power. Here is what an edge data center developer can offer you
A small data center needs a few acres or an existing building, with power and fiber close by. If you own a property like that, here is how a deal works, what you get, and what to ask.
The mirror is live: what operating a Metro HA pair teaches you
Our mirrored pair is operating, with the second site 20 to 50 miles from the first, carrying live traffic in a serial path to end users. Here is what running it teaches about high availability, failover, and operations.
Why AI inference belongs close to users, not in a distant campus
Gartner projected that inference would overtake training in AI-optimized cloud infrastructure spending in 2026. Inference runs continuously, where the work happens. That is why it belongs at the edge.
The telco central office is the edge data center waiting to happen
Thousands of central offices were built for the copper network and now sit underused. They already have the building, the power, and the fiber. Retrofit is the fastest way to add edge capacity.
What Metro High Availability actually means (and why we use 20 to 50 miles)
A single data center is a single point of failure. The fix is a second site 20 to 50 miles away: far enough to survive a site loss, close enough to stay in sync.
The 1-10 MW gap: AI inference needs capacity hyperscalers won't build
Hyperscalers build 100 MW campuses that are committed years in advance. AI inference needs 1 to 10 MW close to users, and we believe that is the least served part of the market.
Talk to a specialist.
Tell us the market, the megawatts and the date. We reply within one business day.








