Bloom Energy’s Sweet Spot: Data Center Backup?


The bloom on Bloom Energy has begun to fade, and questions are rising about the fuel cell startup’s competitiveness against existing forms of power. At $7 to $8 dollars per watt,  Bloom’s fuel cell capital costs have trouble matching those from some distributed generation systems such as solar panels, not to mention grid power — though Bloom says it’s working hard to bring its costs down further.

But what about those rare, but critical, hours when grid power goes out? While most power outages are brief, even a tiny break in electricity flow can wreak havoc with always-on facilities like hospitals, factories — and, in particular, data centers. To keep the juice flowing all the time, data centers need uninterruptible power supply (UPS) systems — and combining UPS ability with always-on power is where Bloom’s boxes might find themselves competitive, if they can prove their reliability.

That’s according to Sam Jaffe, analyst with IDC Energy Insights, who points to Bloom’s showcase customers Google (s GOOG) and eBay (s EBAY) as potential early adopters of next-generation UPS’s for their immense data center backup power needs. The fact that Bloom boxes can also deliver power 24/7 365 days a year strengthens their case, he says. In fact, it’s possible that Bloom could find data center clients to buy up all 40 megawatts of the production capacity Jaffe said that the company is planning to bring online in the near term — though they’ll have to bring down costs to expand much beyond that, he warned.

Data centers use up to 2 percent of the nation’s electricity. That power demand is expected to double by 2011, which would cost data centers some $7.4 billion in power bills, according to the Environmental Protection Agency. In an August 2007 report, EPA pointed to more efficient UPS systems — current ones can waste up to 10 percent of the power they use — as a key step to help avert a data center energy crisis. Google’s taken a crack at it with its one-battery-per-server methods, and other innovative UPS ideas are cutting inefficiencies as well.

Today’s most common UPS alternatives, Jaffe said, tend to technologies such as a warehouse of batteries or several flywheels to cover power outages for the first few moments, and diesel generators or microturbines for longer-term backup. How much that all costs is hard to calculate, since systems differ so widely from data center to data center, he said. Compared to similar-sized microturbines from Capstone Turbine (s CPST), which Jaffe estimated are selling at about $2,000 a kilowatt, Bloom’s boxes are still about four times as expensive, he noted.

But besides the fact that they’re always on — no more battery-plus-generator redundancy — fuel cells hold one distinct “green” advantage over diesel generators and natural gas-fired microturbines — both of those alternatives come with air emissions that might keep them banned from many office parks. Natural gas-powered fuel cells, which emit carbon dioxide but no other pollutants, don’t.

The big question is whether Bloom boxes can offer the 99.99-percent (“four-nines”) reliability required of UPS systems. According to GigaOm’s Pedro Hernandez (sub required), they don’t — at least not according to early trials with Google, which reported its Bloom boxes delivered 98-percent availability over 18 months of testing. How that might compare to post-testing reliability rates is hard to say. Google has also noted that it has been using Bloom boxes to power parts of its headquarters building, and not any off-site data centers.

UPS requirements, of course, stand in stark contrast to raw, everyday power. Many data centers get electricity price discounts to match their immense appetites, Jaffe noted, which makes powering them full-time on distributed systems like solar panels or Bloom boxes even tougher to pencil out. But adding UPS capability might just make it worth the high cost of that always-on power to replace grid power, he said.

Still, using fuel cells in data centers isn’t unprecedented. Fujitsu gets half its Sunnyvale, Calif. data center’s cooling power needs from a 200-kilowatt fuel cell, and Verizon is using seven fuel cells to power its Garden Island, N.Y. data and telephone center. Both projects use fuel cells from UTC, the division of United Technologies (s UTX) that started out making fuel cells for NASA some 50 years ago, which gives a sense of how long this technology has been around. Among the many fuel cell startups on Bloom Energy’s tail are several aiming at the backup power business.

By the way, Jaffe had a couple more calculations on Bloom that don’t look as good as the startup’s publicized figures. While Bloom says its fuel cells can deliver power at 8 to 10 cents per kilowatt-hour when federal and California state subsidies are taken into account, Jaffe puts the unsubsidized figure at 14 to 15 cents per kilowatt-hour. And while he hasn’t done a calculation for what they’d cost with just federal subsidies, the fact that Bloom’s claim to California incentives appears to rely on the availability of hard-to-get biogas might require customers to start doing that math on their own.

Related GigaOM Pro articles (subscription required):

Facebook’s Coal-Powered Problem
Bloom Energy and Data Centers – Perfect Together?
Report: Green Data Center Design Strategies


Bob Plugh

At 14-15 cents per kw-hr, that’s just about the going rate here in New Hampshire, and we have some of the most expensive electricity anywhere (mostly due to all the nut cases that demonstrated against nuclear power and caused the many delays in building the plant).

In any case, I can install solar panels for something like $8-10/watt right now and that number is going down. I use roughly 2000 kw-hr / month – approximately $300/month. Assuming I get 4 hrs of good sunlight per day, that’s 120 hr/month, so I could use 16 2/3 kw of solar power to meet my needs. Ok, let’s say I can be more efficient, drop that down to 15kw. That means a solar system would cost 120-150k for my house.

If I take 120k and invest it at 5%, that is 6k/yr, or $500/month. That covers my electricity bill and gives me 200/month more back.

Oh – there’s a 30% solar tax incentive, so, that solar will only cost me 84-105k. 84k invested at 5% is still 4,200, or $350/month – still enough to cover my electricity cost and that’s at 5%, a very low ROI.

All these alternative energy schemes will be GREAT, BUT, they have to be affordable and right now they’re not without HUGE government (meaning you and I pay for them!!!) subsidies.

I REALLY REALLY would like to install solar, and I REALLY REALLY would like to install a geothermal system, but the cost simply cannot be justified at todays prices.



If it succeeds, there are many benefits to Bloom’s solid oxide fuel cell technology. The distributed nature of the device could help foster adoption of electric vehicles since it could help offload some of the stress on the national grid.


Bloom’s energy server can also help reduce the amount of fresh water used to generate electricity at centralized coal-fired plants especially for arid climates such as in the southwestern US.


I think it is important that we look at the big picture of the value of technologies such as this instead of just focusing on whether its claims of GHG reduction are accurate.



I’m not sure that analysis will play out. Combined Cycle power plants can achieve 55% efficiency NOW, using the same natural gas, and they only cost $550/kiloWatt to build. I don’t think on-site fuel cells are cheaper than that, even if the server companies don’t have to pay “retail”.

Jeff St. John

Thanks, Jim. I think you’re spot-on with both comments about fuel cells and how they work. Your points get to the chief (and hopefully clear) point that I was trying to bring across — fuel cells work for data centers as BOTH distributed, always-on power AND as UPS for when the grid goes down. Given the need to run fuel cells constantly to pay back their capital costs, you’d need to use the power to cut utility bills in 99.99% of the time the grid works, and also use it for UPS when the grid goes down. Hope that helps…


No, no, no. The Bloom box is a BAD idea for data center backup.


  1. Spin-up. The box needs to be heated to 500C or so to work. That means for “instant” on, you need to burn lots of fuel just to keep it hot. Compare this with a turbine that can spin up in a few seconds.

  2. Efficiency. The Bloom boxes are very efficient. That is part of the reason to justify the cost premium. But for a backup system, efficiency is a secondary or tertiary concern. What you want is low cost and high reliability. Both turbines and diesel generators do this much better than a 500C fuel cell.

Dimitri balamotis

If possible i want to discuss with you a different application
Could you please give me your phone number?
Mine is 917 846 7701.


Comments are closed.