IO Founder and CEO George Slessman told me during a phone call on Tuesday that the company raised the latest money in order to capitalize on a huge opportunity. One industry group estimates that companies will spend $105 billion building and managing data centers in 2012, and Slessman thinks IO can address a significant portion of that spending, especially as it relates to costs around designing, building and generally operating facilities.

IO’s modular data centers include their own cooling units and backup power and can go pretty much anywhere a customer has room to put one. The company also has developed its own data center operating system software, which lets users analyze, automate and manage their data centers with a fine-tooth comb using a rather slick interface. The goal is to cut down on excessive data center build-outs to account for future needs, wasteful cooling practices and unintelligent resource management.

A screenshot of the IO.OS software

The approach has garnered a respectable number of big-name customers, including, most recently, Goldman Sachs. The investment bank decided to deploy IO modules as its primary means of capacity expansion in the future, meaning its days of designing and building data centers are over.

IO also has a managed hosting and colocation business, which began as traditional raised-floor space housing servers in cages but is now primarily comprised of the company’s modules. However, Slessman said, in the relatively short term, “well over half our revenue will come from product sales where people are buying it and putting it on their site.” Because of regulatory and other needs, many companies simply will never host the majority of their applications.

Slessman said about half of its latest investment will go toward R&D and expanding its footprint deep into the Asia, Australia and the Middle East (the other half will go in the bank). Already, IO is planning to launch hosting locations in numerous locations throughout Europe, South America and Asia. I was told during a visit to the company’s Phoenix headquarters in April that is has developed a manufacturing model that can be easily repeated from location to location so modules aren’t being shipped around the world.

But IO isn’t able to raise so much money just because it costs so much to invent and build modular data center technology. Everyone — including the mainstream press — is now paying attention to data center efficiency, and any technology that cuts down on the amount of energy a data center uses should be at least somewhat appealing to investors and customers. For example, Calxeda, an Austin, Texas-based startup that’s building enterprise-grade computing fabrics out of out of low-power ARM processors, announced a $55 million investment round on Monday.

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

• 3 baby steps toward greener data centers
• How a Snapshot of a Green Data Center Can Be Misleading
• The economics of clean-data-center innovation

Autodesk has actually been moving various products and services to the cloud in some form for a few years, but the new product is different, said Grant Rochelle, the company’s senior director of manufacturing industry marketing. Most importantly, it’s cheap. Rochelle said traditional simulation software can cost between $20,000 and $100,000 per user per year, and that’s not to mention the high-performance systems necessary to run it. Many employees who need it don’t get it, and many companies are shut out from purchasing it altogether.

Simulation 360, on the other hand, costs mere thousands. For $3,200 a year, users can run a total of 120 jobs. For $7,200, they get unlimited access. For a limited time, actually, unlimited use is free. And rather than needing separate products for mechanical, fluid and thermal simulation, they’re all included in the new cloud offering.

Rochelle said the goal of Simulation 360 isn’t to move big-money customers such as large aerospace companies or defense contractors to the cloud service (“they will be the last bastion of adopting cloud for most things,” he said), but to attract entirely new users building innovative new products or doing interesting work. One customer, for example, is working to make hospital operating rooms as clean as the rooms in which microprocessors are fabricated. It’s simulating the spread of airborne germs through a hospital to prevent certain illnesses that patients contract while receiving treatment for something else.

Another customer, BioLite, has developed a product that turns the excess heat from a fire into electricity so, for example, campers or individuals in third-world countries can ensure their portable electronics are always charged. Rochelle noted that green building and heating and cooling, generally, are areas where the new cloud offering looks to open up a lot of new doors.

However, while offering this technology as a cloud service might be new, the value story is as old as cloud computing itself. Give innovators without huge IT budgets access to resources at a price point never before possible and see what happens. Maybe it’s Instagram, maybe it’s a new approach to climate modeling, maybe it’s a revolution in energy-efficient HVAC systems, but chances are we’re better off because of it.

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

• As Devices Converge, Chip Vendors Girding For a Fight
• A near-term outlook for big data
• Flash analysis: the tech startup investment environment, Q3 2011

The microprocessor giant just finished a yearlong test of Green Revolution Cooling’s mineral-oil server-immersion technology and is very happy with the results. According to Mike Patterson, senior power and thermal architect at Intel, not only does the technology appear perfectly safe for server components, but it might also become the norm for anyone needing maximum computer power or building out data center capacity.

Oil immersion, you say?

For the record, Green Revolution Cooling has been in the spotlight for a few years now and has reportedly attracted some big users — GigaOM’s own Katie Fehrenbacher actually profiled it and its technology in 2010 — but Intel is the biggest company to publicly come out as one of those customers. In essence, the company’s product, called CarnotJet, houses servers in a specialized coolant oil that absorbs the heat they give off and is then sent to a radiator where it’s cooled before being recycled back into the server housing.

And the technology is incredibly effective. Patterson said that whereas traditional air-cooled server racks often operate at a Power Usage Effectiveness rating of about 1.6 (meaning cooling tacks on a 60 percent increase over the power needed power the servers’ computing workloads), Intel’s oil-immersed servers were operating at a PUE between 1.02 and 1.03. It’s possible to achieve similarly low PUE ratings with traditional air- and liquid-cooling methods, Patterson said, but getting there can require some serious engineering effort and cost a lot of money.

As for concerns over the effect of all that oil on the servers’ processors, hard drives and other components, Patterson says companies probably shouldn’t sweat it. When its test period ended, Intel sent its servers to its failure-analysis lab, which, he said, “came back with a thumbs up that a year in the oil bath had no ill effects on anything they can see.”

Designing for the future

To be clear, though, Intel isn’t about to replace all the air-cooled servers in its data centers with oil-cooled ones. Rather, Patterson said, it’s just now in the evaluation phase. “We’re doing our math to understand if we developed an oil optimized platform, what that would mean [for performance, efficiency, etc.].”

Intel’s servers racks in their oil homes.

The results might mean Intel adopts the technology for production, or might mean Intel can help its server-manufacturer partners drive oil immersion into the mainstream. The first serious users of the technology — even withinin Intel — will probably be high-performance computing departments that want to put as much power as possible toward computing and little as possible toward cooling. “I think once it has proven itself in the HPC arena,” Patterson said, “further adoption will be the next step.”

He’s right about the heat-related pain points in HPC. I recently spoke with Greg Rusu of Peer1 Hosting, who explained to me some of the economic challenges of its on-demand HPC cloud service called Zunicore. Unlike the commodity scale-out servers that populate many cloud data centers, Peer1′s HPC servers are chock full of powerful processors and GPUs and can cost tens of thousands of dollars per machine.

And then there’s the power bill of those high performance servers: “These servers easily consume anywhere between 15 to 20 times the power of a high-end Xeon server,” Rusu said. “… They’re not just a little bit hotter, they’re a lot hotter.” While immersing Peer1′s HPC servers in oil might not cut down on their sticker prices, it could signficantly reduce the cost of operating them. A handful of Green Revolution’s early customer are research centers running HPC systems.

Taking oil immersion mainstream

But not everyone is willing to retrofit air-cooled servers for a life in oil, so taking oil immersion mainstream might require letting businesses buy off-the-shelf servers designed for oil immersion. Patterson said Intel’s research into how to build oil-optimized servers could result in a reference architecture around which servers manufacturers could begin building such systems.

Most servers today follow design principles for optimal airflow, but “we could throw some of those rules out,” he said, and maybe build a better server. The obvious steps are eliminating anything to do with fans, sealing hard drives (or going to solid-state drives) and replacing any organic materials that might leech into the oil. A redesign of the heat sink probably would be in order, as would a rethinking of where things sit on the motherboard.

A motherboard sitting in oil.

Although it’s a new idea to most companies, Patterson thinks the cost savings associated with oil immersion might might make it more palatable than you’d think. “I think the C-level [acceptance] may actually happen sooner,” he said. “… The bigger hurdle might be the data center operations folks themselves who don’t pay the energy bill.”

This is especially true as companies start building out data center space and are looking to save on construction costs as well as energy bills. Oil immersion means there’s no need for chillers, raised floors or other costly measures typically required for air cooling. And, Patterson added, it’s possible the energy stored in the hot oil could be reused more easily than the warm air servers return today, thus making a data center even more efficient.

However, as with anything new, even penny-pinching CFOs will have to come to terms with an entirely new way of operating their servers. “The first time yo uhear about it, you think, ‘Oh, come on, that’s a crazy idea’ … ” Patterson said. “You just have to get past the initial reaction. I think it’s an emotional response more than anything.”

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

• Migrating media applications to the private cloud: best practices for businesses
• What cell phones can teach us about energy efficiency
• The fourth quarter of 2012 in cleantech

Granted, EMS’s boxes are nowhere near as powerful as a massive data center chock full of computing gear, but that’s kind of the point. Modular data centers are all the rage right now because they let companies grow capacity as its needed, whether that’s a rack at a time inside an IO Data Centers unit or 1,920 servers at a time inside one of eBay’s specially designed modular units.

But modular doesn’t necessarily mean easy or flexible. At the smallest, a standard shipping container unit needs 120 square feet of floor space and can weigh 100,000 pounds, while one of IO’s units needs 500 square feet.

If you just need a rack full of dense computing power that can go quite literally anywhere you have room, EMS might be your provider. The company’s biggest unit, the 42U and roughly 165-cubic-foot R.A.S.E.R. HD, can handle up to 80 kilowatts at a PUE of 1.1. Its mid-range unit — the R.A.S.E.R. DX, which is the 105-cubic-foot unit AOL uses — can support 12 kilowatts and can fit through a doorway. Its smallest product, the C3-S.P.E.A.R. comes on wheels.

Aside from their size, the micro-modular data centers (as the company calls them) are so versatile because they’re fully contained units complete with state-of-the-art cooling and fire-supression technologies. EMS CEO Bill Stockwell explained their appeal to me like this: “Can you imagine taking a gallon of milk out of your refrigerator, putting it on the counter, and cooling your whole house [from the coolness it puts off]?” That’s how many traditional data centers function, he said, with central cooling that has to handle all the racks within the facility. “We make a refrigerator unit for IT deployment.”

Tony Cole, EMS’s vice president of North American sales, told me many customers put them inside office building in lieu of building out new rooms capable of handling computing gear. Royal Caribbean puts EMS units on its cruise ships. AOL appears content putting them on concrete slabs outside its offices.

Others that have their own data centers will deploy a micro data center when they want to increase density for a certain application but their current infrastructure can’t handle tens of kilowatts per rack. That’s what happened with a missile range in New Mexico that bought a load of high-density gear but then found out it would cost $900,000 to retrofit their data center to support it. Instead, they consolidated 10 racks into two R.A.S.E.R. HD units and actually acquired more computing capacity as a result.

Because it’s liquid-cooled, “there isn’t an IT package on the market today that the R.A.S.E.R. HD couldn’t support from a cooling perspective,” Cole said.

Currently, Cole said, EMS’s micro data centers often act as disaster-recovery sites for mission-critical applications or sometimes get placed at remote offices. But EMS co-founder and Chief Technology Evangelist Simon Rohrich said he sees a shift happening as more companies get interested in owning their own cloud computing infrastructures, but don’t have the budgets or expertise to build and run their own data centers.

In this regard, AOL could turn out to be one heck of an important customer as both a use case and a cheerleader. AOL’s use case of deploying EMS units globally as part of a single cloud platform shows customers how easy it can be to make 10 racks across 5 continents look like a single location with the right software, Rohrich said. And having the company out in public touting 5 times the computing capacity for 10 percent of the cost is sure to turn some heads, too.

Already, he added, customers are starting to realize they can deploy multiple EMS units at a single location to achieve some powerful systems. Cole noted that EMS has deals in place for 13- and 16-unit arrays already, but suspects bigger arrays will come. The company is still in the early stages, he said, “so people are just kind of kicking the tires.”

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

• 3 baby steps toward greener data centers
• How a Snapshot of a Green Data Center Can Be Misleading
• How to Make Cloud Computing Greener

The Nibiru project, he explains, is a set of “incredibly game-changing” goals for transforming the way AOL’s services division carries out the work of managing the company’s infrastructure, and the newly materialized mini data centers we’re high on the list:

Our primary “Nibiru” goal was to develop and deliver a data center environment without the need of a physical building. The environment needed to require as minimal amount of physical “touch” as possible and allow us the ultimate flexibility in terms of how we delivered capacity for our products and services. We called this effort the Micro Data Center. If you think about the amount of things that need to change to evolve to this type of strategy it’s a bit mind-boggling.

Among those changes, Manos writes, are capabilities such as being able to deploy infrastructure wherever it’s needed regardless of temperature and humidity, and the “ability to fit into the power envelope of a normal office building.” The units will also be part of AOL’s automated cloud computing infrastructure, which means they’re managed as part of a greater pool of resources without the need for dedicated staff.

But these mini data centers aren’t just about a cool operations project; they could end up paying big dividends for AOL’s business lines, as well. Not only can AOL save money by taking up less space and power within a traditional colocation facility — if it decides to place a Nibiru box in such a facility at all — but the box gives AOL the flexibility to move into new geographies with ease. Among the benefits Manos points to are:

• It allows us an incredibly flexible platform for driving and addressing privacy laws, regulatory oversight, and other such concerns allowing us to respond rapidly.
• Gives us the ability to drive Edge Computing delivery to potentially bypass CDNs for certain content.
• Gives us the capability to drive ‘Community-in-a-box’ whereby we can quickly launch new products in markets, quickly expand existing footprints like Patch in a low cost, but still hyper-local platform, allow the Huffington Post a platform to rapidly partner and enter new markets with minimal cost turn ups.

Containers atop eBay’s Project Mercury data center in Phoenix.

AOL’s new boxes might be about the smallest data-center-in-a-box units around, but they’re actually part of a great move toward modular data centers across companies of all types. As I reported in April, eBay is buying custom-built containers full of thousands of servers that it can drop into (or on top of) its data centers as capacity dictates. IO Data Centers has built an entire business around modular data centers that can sit just about anywhere.

Update: AOL’s micro data centers were built by Elliptical Mobile Solutions, according to a spokesperson for that company. It has also built units, which it calls micr0-modular data centers, for NATO, the U.S. government and the Canadian Department of Defense.

All these efforts share the same goal of letting companies grow their server count when and where needed rather than trying to predict necessary capacity and relevant geographical locations years in advance.

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

• Big Data, ARM and Legal Troubles Transformed Infrastructure in Q4
• Social third-quarter 2012: analysis and outlook
• The economics of clean-data-center innovation

Dell and eBay took home the award for “Modular Data Center Deployment” for its Project Mercury data center in downtown Phoenix. That data center, which I profiled earlier this month, makes use of standardized racks of gear on the floor and of high-powered but high-efficiency data center modules on the roof. It currently has two Dell modules and two HP units in place, but is looking to add seven more as its computing needs ramp up. Because they’re so efficient, the entire facility’s Power Usage Effectiveness (PUE) rating, which gauges the percentage of power used for tasks other than computing, falls with each new module.

Facebook, for its part, won the award for “Audacious Idea,” a category that didn’t even have another finalist, for its Open Compute Project. That’s probably rightly so. The project began with Facebook open sourcing its designs for both servers and entire data centers, and now features an industry-led foundation aimed at bringing some of these highly efficient principles and practices into mainstream enterprises.

The awards will be presented on May 14 during a session and the Uptime Institute Symposium in Santa Clara, Calif.

There’s still a long way left to go in 2012, but there are already some innovative data center designs underway that could result in awards at next year’s event. EBay has plans to outdo Project Mercury with a modular data in Salt Lake City called Project Quicksilver. Facebook is building a presumably green facility in Sweden, in the same region where Google takes advantage of seawater to cool its Finnish data center. And speaking of Google, it has built a system in its Douglas County, Ga., data center that’s cooled in part by wastewater from nearby towns.

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

• The economics of clean-data-center innovation
• Migrating media applications to the private cloud: best practices for businesses
• How a Snapshot of a Green Data Center Can Be Misleading

The problem, IO Co-Founder and CEO George Slessman told me during a recent visit, is primarily one of planning. This is especially true for traditional IT departments that try to plan ahead to determine what they’ll need as far as 10 years down the road. That’s a laudable goal, but “you’re gonna be wrong,” he said.

Money spent on nothing

If companies realized they didn’t build enough capacity — something Slessman noted can happen even while a data center is still being planned and built — they’re going to pay. While the the gear that fills them drops in price with each new generation, data center prices “have been going up since the pyramids,” Slessman said. That’s not to mention the cost of energy to power them, which itself is always rising in terms of both kilowatt-hours and total capacity.

George Slessman

More often, though, Slessman said, “The real issue is the wasted resources that sit in most data centers.” Estimates vary as to how underutilized the average data center is (Slessman says most are operating at about 30 percent capacity, although some estimates put that number lower), but the end result is always far too high a price tag per megawatt actually used. Fifteen million per megawatt to build a traditional data center effectively turns into $45 million per megawatt if you’re only using a third of that capacity.

Data centers are also inflexible or, as Slessman likes to say, “thickly provisioned.” What he means is that it’s extremely difficult, if not impossible, to build a data center that adapts to changes either internally or in the greater IT world. While servers continually get more powerful and racks more dense, the amount of power available to each rack doesn’t change. And although some applications might vary in importance or new applications might come online, they all get the same resiliency. A Tier 4 data center is always a Tier 4 data center.

IO isn’t for webscale

Off the bat, you can see the differences between what IO is trying to do with its modules and what webscale operators such as eBay demand from theirs. Unlike the units from Dell and HP, which are really just lots of efficient compute power, IO’s modules are fully contained data centers including their own power supplies and cooling systems. They can handle only 250 kilowatts apiece (as opposed to eBay’s 800-kilowatt Dell module) and sport a very respectable, but not world-beating, Power Usage Effectiveness rating of 1.17.

Inside an IO.Anywhere module

But that’s plenty good enough for most of IO’s more-mainstream module customers, which aren’t aiming to run one of the web’s largest search engines. They don’t need to buy a module packed to capacity with servers (in fact, the 500-square-foot modules are roomy even when filled to capacity with 18 racks), but want something that can be sited wherever there’s room and that can be filled with gear as demand requires. They also like that IO.Anywhere units are capable of meeting Tier 4 resiliency requirements (that is, they’re designed to maintain a very high level of availability) but a single space can also house additional IO modules running at lower resiliency for less-critical applications.

Since shipping its first container in July 2011, Slessman said demand is ramping up nicely, especially among large enterprises. While about 75 percent of IO’s customer presently host their capacity at IO data centers, he expects that number to drop to 50 percent by the year’s end as more companies buy IO.Anywhere containers to put on their own premises. Business is doing so well, in fact, that IO built its own factory dedicated to constructing the modules with assembly-line-like efficiency.

Practice what you preach

Oh, and IO isn’t just the IO.Anywhere vendor, it’s also a client. About two-thirds of the company’s 538,000 square square feet of compute area in Phoenix is dedicated to modular data centers (the third is a traditional data center), and it has an entirely modular, 831,000-square-foot data center in New Jersey that recently achieved Tier 3 certification.

Just like it preaches for clients, IO only brings in new modules when it needs to. In Phoenix, it has space for 90 modules in its operational modular data center, which it calls Phase 2, but only has 33 in place. There’s additional space labeled Phase 3 that’s still sitting empty and unpowered; no use lighting or cooling something that’s not being used.

Compared with traditional data centers, IO’s modular data centers are striking in their minimalism. There are no cooling ducts and power supplies coming down from the ceiling, just open air. Standing on the three-foot-high raised floor, your eyes rarely focus above nine and a half feet, which is as high as the steely units rise. Because the modules contain their own cooling, the air in the modular data center space is markedly warmer than in the neighboring traditional space full of servers in cages.

However, the strongest evidence of IO’s commitment to efficiency might be its quest to, as Slessman told me, “actually reduce the amount of data center resources our customer have to buy.” It’s doing this by selling data center management software called IO.OS that lets customers monitor, analyze and control all their data center resources from a single interface. The software, he said, will drive hardware optimization by letting data center staff see what’s running, at what capacity, drawing how much power — you name it — and manage it all from the same screen (even if that’s on a tablet or smartphone).

Even if they’re not building scale-out cloud data centers, companies still will demand the level of automation and efficiency that cloud computing promises, and that’s only possible by complementing lower-power hardware with software-based intelligence. “And as we skate toward the cloud world,” Slessman said, “it’s all going to happen programmatically.”

All images courtesy of IO Data Centers.

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

• 3 baby steps toward greener data centers
• How a Snapshot of a Green Data Center Can Be Misleading
• How to Make Cloud Computing Greener

If eBay’s Dean Nelson has his way, that was just the beginning. His future is one of ever-greater density in data centers driving ever-greater efficiency, and he’s relying on modular data centers like the ones Dell has provided to get him there.

Sometimes the modular data centers are the standard 8-foot by 15-foot shipping containers, and sometimes they’re more unique, custom designs. But they’re always loaded to the teeth with gear. A single unit can weigh 100,000 pounds — if you drop one in, you instantly have a whole lotta computing power in a relatively small, wholly weather-proof package.

Add a thousand servers, lower the PUE

Nelson, who serves as eBay’s senior director of global foundation services (read “he runs its data center operations”), is so excited about what his team has accomplished with the new data center that he couldn’t help but take to the whiteboard in the conference room and break into a mini-lecture on how the project came to be. Many of the nitty-gritty details on how Project Mercury shaped up — including specifications on how it’s cooled and powered –  are available in a whitepaper by the Green Grid organization. But here’s the takeaway: total cost of ownership (TCO) drives everything.

If Nelson is going to buy it, it’s going to be flexible enough to change with future generations of gear and it’s going to hit the perfect blend of density, efficiency and performance. Done right, containers and modular data centers fit the bill perfectly, and Nelson will buy them from whichever vendor is best able to meet his requirements when it’s time to load up on more capacity (and when he needs more than 1,000 servers at a time). Already, he told me, “We’ve doubled our capacity and my [operational] budget has stayed flat.”

Two HP containers, and the two Dell units in one enclosure.

Among that added capacity was 50 petabytes worth of Hadoop drives split among two 1,008-node clusters. One cluster resides at the SuperNAP data center in Las Vegas, the other in an HP container atop eBay’s new data center.

Nelson has been able to maintain a steady operational budget because it’s difficult to beat what Nelson’s containers are doing in terms of efficiency. And you can’t talk about efficiency without talking about power usage effectiveness, or PUE. It’s an industry standard for determining how much energy a data center uses for tasks other than computing (e.g., cooling or lighting). The ratio is simple enough, you divide the total power usage by the power used for computing, with 1.0 being perfect. The world’s most-efficient data centers from Google, Facebook and Yahoo top out at about 1.08.

Project Mercury gets free cooling year round, even in the heat of summer. On Aug. 23, 2011 — a 119-degree day — one of eBay’s Dell units had a partial-PUE score 0f 1.044 while drawing 520 kilowatts of power. On January 17, 2012, while drawing 1 megawatt, the same unit had consistent partial PUE of 1.018 while the rest of the data center was doing between 1.26 and 1.35. Project Mercury has room for 11 modular data center units on its roof, and every one drives down the PUE of the entire facility’s PUE. Nelson realistically expects an entire facility PUE of less than 1.2.

Big thing, small package

Given all the computing power it’s capable of consuming, though, one of the most striking things about eBay’s Project Mercury data center is its size. eBay has an existing data center next door that’s 43,000 square feet and can handle 6 megawatts of power. Project Mercury has about 14,000 square feet of computing space (including the roof), but is designed to handle 12 megawatts of power (it’s currently drawing 4 megawatts). It also costs half as much as the larger space to fill with gear and to operate.

Project Mercury has a maximum power capacity of three times its current draw for two major reasons: 1). it’s not yet full (there’s still space for seven more units on the roof alone) and 2). Nelson demands higher performance with each new generation of gear he buys. He’s even pushing modularity on the first floor that looks more like a traditional data center, which means standard-size racks with ever-more power crammed into the exact same amount of space.

It’s a strategy Nelson calls “rack and roll.” Vendors with the winning bids deliver 48U racks chock full of gear that eBay only has to plug in. When it’s time to replace racks or add new ones, his team just rolls them in and out as needed.

Right now, the HPC racks that power eBay’s search engine contain 96 servers and pull 28.8 kilowatts of power. They weigh in at 3,500 pounds apiece. Its Hadoop racks contain 48 servers, each stuffed with 12 2TB drives. The first-floor space can fit 220 of these power-packed racks, while each rooftop container can fit 20.

But they keep getting more powerful. The 28.8-kilowatt racks seem downright Herculean next to their legacy 8-kilowatt neighbors on the raised floor. When future generations surpass 40 kilowatts per rack, Nelson has equipped the facility for liquid cooling direct to the chips. On the roof, Dell boosted its performance 800 kilowatts max on the second-generation modular data center up from 550 kilowatts on the first-generation model.

At webscale, (almost) everyone’s doing it

To hear Dell’s Data Center Solutions division tell it, modular data centers aren’t exclusive to eBay. Steve Cumings, the division’s executive director, told me Dell has a good business in selling modular data centers to webscale (and other hyperscale) customers. And they like them for the sames reasons eBay does: they’re super-efficient and super-dense. Given the right cooling setup, the units can sit wherever they have access to power and act like their own little data centers.

Although Cumings isn’t allowed to give the names of most DCS customers, or even the size of their deployments, he did note that Microsoft uses Dell Modular Data Centers to power Bing Maps. He also said deployments range from eBay’s scale to “many, many, many modules.” Elsewhere, everyone from the U.S. Army to Amazon is going modular.

Better, perhaps, but will it be Tron-themed?

But eBay’s Nelson isn’t about to let anyone steal his thunder, at least when it comes to driving the most bang for his buck out of modular data centers. eBay is about to break ground on a new facility called “Project Quicksilver” in Salt Lake City, Utah, that he says is even bigger and better than Project Mercury. It hasn’t yet released details on what will be inside, but the RFP calls for a per-rack performance of up to 40 kilowatts and overall facility scalability from 4 megawatts to 30 megawatts.

And Nelson — always hungry for more performance per watt — has a two-word message for server vendors that want to win eBay’s business on future buildouts: liquid cooling. For servers, it’s like the difference between fanning yourself and jumping in a pool, and he says it will be necessary as his racks achieve more power density.

“That will be the competitive advantage for a vendor,” he said “because we will buy it.”

All images courtesy of eBay.

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

• How tomorrow’s mobile-centric data centers will look
• Cloud computing infrastructure: 2012 and beyond
• A near-term outlook for big data

With these dramatic growth rates, it makes one wonder what 2021 will look like, and at what point our insatiable desire for more data, be it video, or search results or CRM databases, will make it even more critical to make data centers sustainable.

Other stories that Adam is reading and thinking about this week include:

• Renewable power trumps fossil fuels for first time: Renewable energy ($187 billion) got more new investment than fossil fuels ($157 billion) for the first time.
• Ford trots out electric focus, promises its hybrids will trump Toyota, GM: Ford is preparing to get into the hybrid and EV game with its all electric Focus and its C-Max Hybrid and C-Max Energi Plug-in Hybrid. The Energi Plug-In will go about 20-30 miles on a charge and can go another 500 miles on the tank of gas. Ford is seeing sales increase in the compact car market.
• Wind, biodiesel subsidies promoted: Concern over expiration of the dollar a gallon subsidy for biofuels is mounting as the industry could get hit, along with the wind power industry.
• Zipcar upping stake in Spanish car-sharing firm: Zipcar’s European expansion began in the U.K., but now it has its eyes on Spain.

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

• How to leverage the web-sharing economy now
• What cell phones can teach us about energy efficiency
• 3 baby steps toward greener data centers

To read the rest of his weekly update check out GigaOM Pro.

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

• Why the future of smart meters is data, not hardware
• How to leverage the web-sharing economy now
• What cell phones can teach us about energy efficiency