We've seen it on many of the projects we've worked on, mining and AI alike. The compute shows up ready. The cooling, ready. Then the whole site stops and waits, sometimes for a year, on a transformer. Or on a switchboard stuck behind someone else's supply chain, with no firm date and no one accountable for it.
That's why we built Fog Power, our power infrastructure line for high-density compute. We're launching it today.
The choice the market forces on you
If you've bought power gear lately, you know the bind. You can have it fast, or you can have it real, and most of the time you only get one.
Go to a legacy OEM for a compliant, well-built unit and you're quoted lead times that run from many months to over a year. Go to one of the newer shops promising speed and you often find a broker: someone reselling whatever's on hand, passing the timeline and the quality risk straight back to you. A fast quote, and no one standing behind it.
We didn't accept that those were the only two options.
A cooling company that kept running into a power problem
For years, Fog has built the compute side of the data center. Immersion and hydro cooling for Bitcoin mining. Immersion and direct-to-chip cooling for AI and HPC. More than 400MW deployed across 40+ countries, for over 5,000 customers, in six climate zones.
Through all of it, the same thing held up schedules. Not the cooling. Not the load. Getting clean, compliant power to the racks, on time. So we stopped treating that as someone else's problem. Cooling and power aren't two problems. They're the same problem from two ends.
What Fog Power is
Fog Power is the power infrastructure layer of what we do. It starts with the two pieces that gate almost every site:
Transformers from 75 kVA to 500 MVA, up to 345 kV, built to ANSI/IEEE, stocked in Texas with 14-day dispatch on in-stock units.
UL 891 switchboards from 1200 to 4000 A at 65 kA SCCR, built around UL listed breakers, with roughly eight-week lead times.
Battery storage (BESS) and gas generators are next. Both are already in the works, coming online soon.
The line covers the full power chain, from the grid connection to the rack. The transformers and switchboards shipping today are built for North American compliance and backed by North America based support that runs from the first quote through after-sale.
The headline isn't the spec sheet. It's the timeline. Power gear measured in weeks, not quarters.
We don't broker boxes
Fog Power runs on a supply network we vetted ourselves. We didn't assemble a catalog of whatever was cheapest this quarter. We picked manufacturing partners for their track record, then put our own work on top: the UL and ANSI/IEEE compliance, the integration, the control intelligence, and support that answers the phone in your time zone.
That network isn't theoretical. It has more than 20,000 units in the field, across 40+ countries and 2,500+ projects. We chose it for exactly that record.
A fast lead time only means something if the partner behind it is real. Ours is, and we stand behind every unit that ships.
That layer is also where compliance lives. Every unit is built to UL, ANSI/IEEE, NEMA, and DOE standards and cleared before it ships, stocked and supported here in North America. That's the difference between a delivery date and an in-service date.
The part a transformer vendor can't tell you
We engineer the power going in and the cooling it feeds. A company that only sells transformers can't say that.
When you size a transformer or a switchboard for a high-density site, the electrical side and the thermal side are tied together. Mining rigs and GPU racks both pull hard, nonlinear loads, and we've spent years living with what they do downstream, because we build the systems that cool them. That goes straight into how we spec the power side.
We don't just deliver the power. We understand the load it feeds.
Why now
Demand for high-density compute is outrunning the grid in North America, and the equipment market hasn't caught up. That's true whether you're standing up an AI data center or expanding a mining site on new power.
The wait costs from both ends. AI infrastructure runs roughly $8 to $15 million per MW (CoinShares, Q1 2026), so every idle month is a fortune in stranded capital. Mining runs on margins thin enough that a lost month is a month you never get back. Different math, same conclusion: the winners are the ones who shrink the time between "we have power" and "we're computing."
We know that problem from the inside, from hundreds of MW deployed at the edge of the grid where a month of delay is real money. We didn't leave mining to chase AI. We build for both. Hashing or inference, the power problem underneath is the same.
Power In, Compute Out
"Power In, Compute Out" is how we've described Fog from the start. With Fog Power, the first half of that line is something we now deliver ourselves, end to end.
If you're sitting on power and racing a timeline, the power gear is the part we just made faster. See the Fog Power lineup, or build a quote and tell us what your site needs.
