Thanks for publishing this, Rich. For some clarity, regarding the comments above:

-This bunker is a hardened box atop a vast groundwater cooling resource. Gives us a PUE of under 1.1, scalable up to 15MW. Power costs aren’t everything. Efficiency is. We cool exclusively with an open-loop groundwater system, scalable to more than 10 million gallons per day. No chillers, no compressors, no need for free airside economizers.

-We don’t really care about nukes. But the government that built our bunker – for $30 million in today’s dollars – included hydrogeology in their site-selection criteria 50 years ago, for cooling and air filtration needs. Using groundwater to chill our IT critical load is so leading edge that we find comfort in over 35 years of daily logs, demonstrating beyond any doubt that we get water volumes at the temperatures we require (under 50 degrees F), year ’round.

-Plus we get a nice, low-cost 64,000 sq ft box with 14′ ceilings, which our structural engineers say will last another 500 years, at a small fraction of the cost of building new shell stock. Almost 30,000 square feet of new N+1 co-lo space is in the process of being refit downstairs, with all facilities in matching footprints on the level above. This has the added benefit of not only keeping IT and facilities functions truly distinct, but gives us a 20′ heat sink above IT in a catastrophic event – enough to have a couple minutes to manually fire up our gensets if all else fails (and eliminating the need for UPS in our cooling array, if any given co-lo client chooses to forego it).

-Actually, the entire setup was indeed a closed system, like a submarine. Power plants were (and some still are) indoors. Of course we’re deploying new backup diesels outside the building in the new build – why waste precious co-lo space for that? In the age of distributed computing, there should be no Achilles heel, as above. You just need site diversity.

-We acquired some nicely engineered shell stock at very low cost, with useful, diverse high-voltage power and fiber optic infrastructure nearby. The site made sense to us. But the clincher was the aquifer system. Avoiding the capital cost of chiller plant saves us huge money, which enables us to be more price competitive than other co-lo providers. I’m building power densities of over 250 watts / square foot at under $500 / square foot. No CRAC units either – only CRAHs.

-The greenhouse image you removed from your post was indeed related to our site. We’re feeding local agricultural research / production with our free heat / warm water, with a greenhouse directly atop our underground plant. Heat, after all, rises. Most data centers blow it into the atmosphere. We reuse it.

With Canada’s long, cold winters, it’s beneficial to be in a business that generates abundant free heat. We believe we’re demonstrating something useful about the ways data centers can serve their communities. The nearby agricultural college likes it. Our agricultural endeavors also give us bona fide farm status, which lower our property taxes, and other related overheads (including diesel fuel costs). Try saying that about your server farm. I’d rather see produce grown locally – fueled by free heat, a common data center waste product – than see refrigerated trucks ship it thousands of miles from California and Mexico. + More carbon offsets.

-I don’t think it takes one-cent kilowatts to justify site-selection criteria. We’re paying about seven cents now, and as we scale into higher purchasing volumes our tiered average will be in the five cent range. With a PUE like ours – if we compare it to *your* data center – that’s an effective rate of a third to less than half what we all pay – with nice carbon offsets to boot.

Come by and visit, guys. We’ll give you some real PUE metrics to sink your teeth into once we’ve had a couple years’ operating history, post upfits.

Yours in Dataville,

Anton E. Self
Chief Executive
Bastionhost Ltd.

Yes, it appears you’re right. I’ve removed the photo, which Bastionhost is displaying on its web site but clearly has no relation to its data center projects.

Oh yeah, you can survive a nuclear strike. If that becomes a serious shopping point for colocation I think it is time to get out of the business.

–chuck