We'll leave the coffee to you. We'll help you mine better.☕

Exactly the kind of Bitcoin business we’d love to see more of. Nice build.

We’re excited to share our new brand, Fog Computing. AI is scaling faster than the infrastructure built to hold it. The constraint has moved off the chip and onto the site, where powered land is scarce, interconnect queues are long, and conventional build takes years. We believe how much compute gets deployed this decade depends on how fast it can be put on the ground, and that is a manufacturing and power question before it is a compute question. We tackle it through two lines of business. The first is AI infrastructure, from power to compute. We build everything between the grid and the rack, including power gear, cooling plant, and prefabricated data halls, all engineered and handed over as one build, not three purchase orders someone else has to integrate. Over the past five years, we have delivered 400MW of liquid cooling and 500MW of power capacity. The second is AI capacity as colocation. We develop AI factories with our partners and deliver them ready to run, so the site arrives finished and ready for your GPUs. We now have over 100MW of AI capacity in the pipeline. Fog Hashing continues as our Bitcoin mining infrastructure brand, and our foundation stays the same, with our own manufacturing, full system engineering, and speed as a structural advantage, now backed by a bigger team and factory. Stay tuned for new sites and products. Welcome to the future of AI infrastructure. Welcome to Fog Computing. Please visit fogcomputing.ai for more information. #AIDataCenter #AIInfrastructure #DataCenter #LiquidCooling #DirectToChip #ModularDataCenter #PowerInfrastructure #BitcoinMining #FogComputing--

Land like this was never meant for what comes next. But every open field holds a future no one has built yet. That future could be an AI data center.

Next stop, Singapore. We'll be around at Data Centre World Asia, Sept 29-30 at Sands Expo. Happy to chat about AI data centers and cooling.

Yotta 2026. Las Vegas. Sept 28-30. Let's talk AI data centers, from cooling and power to everything in between. See you at Caesars Forum.

Most of a data center's build time is lost on site, not in manufacturing. The build runs through four stages: permitting and grid connection, manufacturing, site construction, and commissioning. Permitting and grid connection follow the utility's timeline, not the builder's. On site, everything happens one step at a time. Piping and cabling wait for the structure to go up. Commissioning waits until the entire system is in place. Manufacturing is the only stage that doesn't have to wait its turn. Usually, several vendors each ship their own piece, and those pieces meet on site for the first time. When something breaks, everyone points at someone else. We design and manufacture power, cooling, and compute as one modular system in our factory, and test it end to end before it ships. The site shouldn't be the first place the system is tested as a whole. The factory should.

Starting an AI site doesn't need to be that complex, especially for those with powered land.​ RTX PRO 6000 ships in air-cooled builds, so do H100 and B300. That's what many inference, fine-tuning, and generation fleets still run on. ​FogPod Air is built for them. Up to 280 kW of IT in a 40ft container, 8 racks at 35 kW each, then paired with a power container. Fill it with RTX PRO 6000, and that's 320 GPUs across 40 nodes.​ It's the fastest way to launch a live site. What's worth more is the know-how you get: how the site runs, how the business model works, and how you scale from the first move to megawatts. ​So if you're sizing your first AI site or adding capacity at the edge, what would settle it for you, the right spec, or the right partner?

Step inside our hydro-cooled container data center in Texas. This time, Crypto LLC took the tour — breaking down exactly how the 2.4MW setup runs. We'll let the footage do the talking. Check it out 👇 youtu.be/Vq1br1yERqU?si=rzjd… #HydroCooling #BitcoinMining

Just shipped one of our upgraded hydro racks out to Arica, Chile. Set up for Antminer's S23, S21, and S19 hydro miners. Loop's fully sealed now, using less water and needing way less upkeep. The rack has 36 independent breaks too, so one miner acting up doesn't take the rest down with it. Remote monitoring's on FogPortal, meaning nobody has to babysit it on-site. If you're just starting, hit us up. #HydroCooling #Antminer

Amazing project. Ever thought about your Bitcoin miner melting the snow off your driveway this winter? Always love watching a build like this come together. Can't wait to see it running.

Watch the #Antminer launches at #WDMS. They dropped the S23e U2H and the S23 XP Hyd. First batches sold out on site. So I took a closer look. Two things are different this time. One thing is depth. This new 2U box is 600mm deep, the same as WhatsMiner and SealMiner's 2U units. Before, Antminer hydro units were always 700 to 900mm deep, whether 2U or 3U. Now everyone's lined up at 600mm. Another thing is ports. Antminer puts water and power on the front. WhatsMiner and SealMiner? Power front, water rear. Same 2U height, but the plumbing goes the opposite way. So if you're picking up these new models, you're concerned about whether your racks can take them? Fog Hashing has two racks that cover it, the HC45-M1200 and the HC40-M2400. They handle both layouts, whatever brand you end up with. No guessing which one shows up; you just drop it in and go.

Most AI data center delays don't happen inside anyone's scope. They happen between scopes. Design, equipment, construction, commissioning, each done by different firms. Each one hits their own spec. But nobody owns the seams, and the seams are where schedules slip and finger-pointing starts. That's why the industry is moving toward integration. We'd push it one step further: integrate the delivery model itself. Treat the data center as a product, not a project. Design it once, build and test it in the factory, ship it, plug and play on site. Construction becomes manufacturing. 1 or 2 years becomes months, field labor drops 80%. And one contract from power in to compute out, so when something moves, you know who picks up the phone. Delays don't live inside scopes. They live between them. Bitcoin mining went through this shift earlier, and harder. Brutal payback math forced it. That's where our 400MW deployments came from. Now we're running the similar playbook for AI. Planning an AIDC build? Comment or DM us for the AIDC product brochure.

Every megawatt has someone walking the aisles behind it. The one closing out FAT before shipment. The one running inspections on schedule, week after week. That's what uptime is made of.

Indoor install day, 2 hydro-cooled clusters, 4.8MW. Skids showed up built and tested from the factory. Set them down, hook up the dry cooler outside, power on. No struggling with complex piping on-site. And these are fully CSA certified. If you've done an indoor site in North America, you know the inspector cares a lot more than at an outdoor container. Certification makes that visit a lot shorter.

From dirt to hash. 4MW, liquid cooled. We built it in our factory. Now it's just nice to watch it run.

Guess how many NVIDIA RTX PRO 6000 GPUs will be immersed in this container ?

Nothing makes us smile more than a Fog site running strong in the field.

Not every AI deployment needs the biggest GPU. Inference demand keeps climbing. Meanwhile, a lot of real power at the edge sits underused, waiting for its AI moment. We've been talking with a few mining operators who want to move into AI. For most sites, the real opportunity isn't an AI factory. It's a right-sized turnkey AI pilot, quick to deploy while the market's still early. One pairing we like is the #RTX PRO 6000 SE in immersion. It's built for passive cooling, which is what you want in a tank. 96GB per card is enough for real LLM inference. Put the cards in one tank, group tanks into 40ft containers. A right-sized place to start. #AIInfrastructure #Inference #Immersion #LiquidCooling #BitcoinMining