A technology abandoned by American policy could make server farms entirely self sufficient. The question is whether anyone will demand it.
By Staff Writer, South Carolina Bulletin
I’ve written two previous articles on this for the South Carolina Bulletin “The Fuel Cell Revolution” (July 29) and “The Other Half of the Flame” (July 31) both available on the Bulletin’s site. The first covers the trigeneration opportunity that Oracle’s 2.8 gigawatt fuel cell deployment opens up. How the 800 degree Celsius exhaust those solid oxide fuel cells produce could run absorption chillers to cool the server racks for free instead of being vented into the atmosphere while the data center burns additional grid power for compressor based cooling. The second covers the fuel cell revolution itself. How the technology works, why Japan has over three hundred thousand residential units and South Korea is building toward fifteen gigawatts while the United States has exactly zero homes running on solid oxide fuel cells and the structural reasons American policy was written to subsidize solar panels while penalizing self sufficient generation. What follows connects both pieces to the question that should be asked at every zoning hearing and county commission meeting from here forward. Why are we still approving data centers that burden local grids and water tables when the technology to make them fully self sufficient already exists, is manufactured domestically and is being deployed at gigawatt scale by the largest companies in the world?
The American grid is straining. Data centers are the villain of the moment, blamed for blackout warnings, rate hikes and transmission corridors rammed through unwilling communities. But the fight over who gets the electrons misses a deeper truth. The technology to make data centers fully self sufficient, generating their own power and their own cooling from a single fuel line without drawing a single watt from the grid, already exists. It is deployed at gigawatt scale. It is manufactured in the United States. And it is being ignored.
The Two Halves of the Flame: Oracle is currently deploying 2.8 gigawatts of Bloom Energy solid oxide fuel cells across its American data center footprint. The first 1.2 gigawatts are already going in. These fuel cells generate electricity at 60 percent efficiency with no combustion, no turbine, and not a single moving part. The technology is 125 years old, rooted in Walther Nernst’s 1899 discovery that zirconia doped with yttria conducts oxygen ions at high temperatures. K.R. Sridhar commercialized it through Bloom Energy after NASA canceled the Mars mission his original solid oxide device was built for. A quarter of the Fortune 100 now runs on Bloom servers. Google was the first customer in 2008. eBay followed, saving a hundred thousand dollars in electricity costs in nine months. Apple, Walmart, FedEx, AT&T, Bank of America and now Oracle at a scale that exceeds the generating capacity of many small nations.
The numbers are staggering. But they only tell half the story.
Every watt of electricity those fuel cells produce comes with roughly 1.3 watts of high grade heat at 800 degrees Celsius. In the standard American deployment, nearly all of that thermal energy will be vented into the atmosphere while those same data centers burn additional electricity to run compressor chillers keeping the server racks from melting. The same fuel that makes the power could make the cold. The technology has existed for over a century. The rest of the developed world uses it at scale. The United States, once again, is choosing to burn fuel twice.
The Cooling That Runs on Exhaust: Absorption chilling is not a prototype. Albert Einstein and his former student Leo Szilard patented a version of it in 1930, a refrigerator with no moving parts driven entirely by a heat source. The technology powered RV refrigerators and hotel minibars for decades. At commercial scale, it is standard infrastructure in hospitals, university campuses and district cooling plants throughout Japan, South Korea, China, and the Middle East.
The dominant system uses a lithium bromide water cycle. Water is the refrigerant. Lithium bromide is the absorbent. The entire loop runs under vacuum. In the evaporator, pressure is kept low enough that water boils at roughly 4 degrees Celsius. Liquid water sprayed over the evaporator tubes flashes into vapor, pulling heat from water circulating through those tubes. That chilled water, now around 7 degrees Celsius, gets pumped through fan coil units or radiant panels to cool the building.
The water vapor drifts into the absorber, where a spray of concentrated lithium bromide captures it. Lithium bromide has an aggressive chemical affinity for water molecules. It pulls vapor out of the chamber continuously, maintaining the vacuum that keeps the evaporator boiling. The now dilute solution is pumped to the generator. This is where the fuel cell’s waste heat enters the picture. The generator boils the water out of the solution, leaving behind concentrated lithium bromide that flows back to the absorber. The water vapor rises to the condenser, returns to liquid, and drains back to the evaporator. The cycle repeats indefinitely.
The only electricity required runs a small solution pump and a cooling tower fan, roughly 5 to 10 percent of what an equivalent compressor chiller would draw. Everything else runs on heat that would otherwise be dumped into the sky.
Critics will point to the coefficient of performance. A modern electric chiller achieves a COP of 5 to 7, meaning one unit of electricity moves five to seven units of heat. A single effect absorption chiller manages 0.6 to 0.8. A double effect unit reaches 1.0 to 1.3. On paper, absorption looks like a toy.
On paper, the comparison is a fraud.
The electric chiller’s COP ignores where the electricity came from. A combined cycle gas plant runs at roughly 45 percent efficiency. Transmission losses eat another 5 to 7 percent. The true system COP from fuel burned to heat moved drops to between 1.5 and 2.5. The absorption chiller’s fuel input is zero. It runs on exhaust. Its marginal COP is infinite. The only honest comparison is total system efficiency from fuel input to useful output, and on that metric absorption paired with a solid oxide fuel cell destroys electric compression paired with the grid.
Solid oxide fuel cells produce exhaust at 800 degrees Celsius. Single effect chillers need 80 to 120 degrees. Double effect need 150 to 200 degrees. Triple effect, still rare but commercially available, need 200 to 260 degrees. The fuel cell exhaust is hot enough to run all three in sequence. First pass through a triple effect chiller for maximum cooling. Second pass through a single effect unit. Third pass through a hot water heat exchanger for domestic or process use. Only then does whatever residual heat remains get vented. The combined efficiency, electricity plus heating plus cooling, can exceed 90 percent from a single fuel input. That is not a laboratory number. That is the operating reality of trigeneration plants across Asia.
The United States does not do trigeneration. The reasons are the same structural failures that keep residential fuel cells out of American homes while Japan installs its five hundred thousandth Ene Farm unit.
American electricity was kept artificially cheap for decades through a regulatory model that socialized infrastructure costs and externalized environmental damage. When a compressor chiller costs pennies per kilowatt hour to run, the capital premium for an absorption system looks like a bad investment. The utility bill never reflected the true cost of generation, so the efficiency case never closed on paper.
American capital budgeting is pathologically biased toward first cost. An absorption chiller costs more to purchase and install than an equivalent electric chiller. The fact that it pays for itself in three years through avoided electricity costs does not matter when the capital budget is siloed from the operating budget and the two departments answer to different vice presidents with incompatible incentives. This institutional failure is well documented and never corrected.
The American grid was built for centralized generation pushing power outward to passive consumers. Distributed cogeneration does not fit the utility’s business model. The utility cannot meter the heat, cannot bill for the heat and cannot control the heat. Interconnection rules, rate structures, standby charges, and exit fees are all designed to penalize anyone who generates their own power.
The trades have atrophied. Absorption chillers run under vacuum and use lithium bromide, which is corrosive and crystallizes if mishandled. In Japan and Korea, an entire ecosystem of engineers and technicians understands these systems because the technology is standard. In the United States, an HVAC contractor who can rebuild a Trane or Carrier chiller in his sleep will stare blankly at an absorption unit. The knowledge gap raises perceived risk, which inflates installation quotes, which kills projects before they start.
And nobody lobbied for it. Solar has the Solar Energy Industries Association. Wind has the American Clean Power Association. Batteries have the Energy Storage Association. Absorption chilling has no trade group, no lobbyists, no foundation funded white papers and no presence in the policy conversation. The HVAC industry makes its money on compressor based systems and has no incentive to promote a technology with fewer moving parts, lower service revenue and a smaller replacement parts pipeline.
The federal Investment Tax Credit offers a 30 percent break on fuel cell installations but the residential credit is capped at five hundred dollars per half kilowatt. On a small home system, that returns perhaps fifteen hundred dollars. The solar credit, by contrast, is uncapped and easily returns five to ten thousand dollars on a typical rooftop installation. The incentive structure was written for solar from the start. Fuel cells were an afterthought in the tax code, not because they do not work, but because no lobbying network pushed for parity.
The implications for data center siting are transformative. Under the current model, every new data center announcement triggers local opposition because the math is zero sum. The facility draws gigawatts from the regional grid, straining capacity and driving rate hikes. Evaporative cooling consumes millions of gallons from local aquifers. New high voltage transmission corridors get rammed through farmland and neighborhoods. Diesel backup generators and cooling towers produce noise. Residents absorb the costs while the developer captures the revenue.
A trigeneration solid oxide fuel cell data center running absorption chilling removes itself from every one of those conflicts. Zero grid draw. Closed loop cooling with minimal makeup water. No new transmission corridors. Silent operation with no moving parts in the generators. No impact on local electricity rates because the facility is not a utility customer.
The political calculus flips completely. Under the current model, a data center looks like a parasite from the host community’s perspective. Under the trigeneration model, a data center is invisible to the local grid, local water table and local ratepayer. It is a quiet box connected to an existing underground gas line generating its own everything. The local opposition evaporates because there is nothing to oppose.
County commissioners, zoning boards and economic development authorities face a relentless tension. They want the jobs and tax base that data centers bring. They dread the angry residents who fill hearing rooms when grid impact studies come back with nine figure transmission upgrade requirements. A trigeneration data center resolves that tension entirely. The revenue arrives. The angry residents do not. That is a deal any local official will take.
The long term play closes the carbon loop entirely. Bloom Energy manufactures solid oxide electrolyzers that are essentially their fuel cells running in reverse. Feed them electricity and steam and they split water into hydrogen and oxygen at 80 to 90 percent efficiency, far better than the 60 to 70 percent achieved by conventional proton exchange membrane electrolyzers.
Nuclear coupled hydrogen production is the most efficient pathway at scale. Nuclear reactors produce enormous quantities of both steady baseload electricity and high grade waste heat. Both inputs feed solid oxide electrolyzers at their optimal operating temperature without consuming additional energy to reach it. A single reactor coupled to electrolyzer banks could produce hydrogen sufficient to feed gigawatts of fuel cell capacity, with the fuel cells sited at the point of use to eliminate transmission losses.
When those fuel cells run on pure hydrogen, the only byproduct is steam. When their waste heat runs absorption chillers, the cooling is produced with zero additional emissions. When the electrolyzer is powered by nuclear or excess renewables, the entire chain from primary energy to electricity to cooling runs carbon free. The natural gas reforming that currently feeds most solid oxide fuel cell deployments is a bridge, not the destination.
The Oracle deal may finally force the conversation that fifty years of engineering logic could not. A 2.8 gigawatt deployment of solid oxide fuel cells is the largest single commitment to on site fuel cell generation in history. Data centers spend 30 to 40 percent of their total electricity budget on cooling. Every watt of that cooling load that shifts from electric compression to absorption is a watt of fuel cell output freed up for compute, a watt not drawn from the grid and a watt of operating cost eliminated.
If even a fraction of those installations include absorption chilling, the demonstration effect will be impossible to ignore. The economics will be documented at scale. The supply chain will develop. The knowledge barrier will begin to erode. The financing models will catch up. Once a few high profile data center deployments prove the trigeneration case, the same logic cascades to hospitals that need power, steam, and chilled water around the clock. To hotels with predictable thermal and electrical loads. To food processors that need electricity, process steam, and refrigeration simultaneously. To cold storage warehouses. To district energy systems serving entire commercial districts. To the residential market, where a home fuel cell producing electricity, hot water, and air conditioning from a single gas connection makes the solar plus battery model look like the half measure it has always been.
Japan has installed over three hundred thousand residential solid oxide fuel cell units through its Ene Farm program, with a target of more than five million units by 2030, roughly ten percent of all Japanese households. The Japanese government treats fuel cells as essential household infrastructure and subsidizes them accordingly. South Korea has deployed over four hundred megawatts of Bloom fuel cells and has a government roadmap targeting fifteen thousand megawatts by 2040. In the United States, not a single home has one.
The technology is proven, deployed at gigawatt scale, and already heating water in hundreds of thousands of Japanese homes. What is missing is not engineering. It is political will.
And political will is the one form of energy ordinary people can generate without a fuel cell. Show up at zoning hearings when a data center is proposed and ask why the developer is not using on site trigeneration with absorption chilling when the technology exists and is deployed at scale. Demand that county commissions make grid independence a condition of site plan approval. Call your state public utilities commission and ask why interconnection rules are designed to penalize self sufficient generation rather than reward it. Call your congressional representative and ask why the Investment Tax Credit caps fuel cells at five hundred dollars per half kilowatt while solar enjoys no cap at all.
The fossil fuel lobby did not build its power overnight and neither will the people who want to break free of it. But the math is on their side. A technology that runs 24 hours a day at 60 percent efficiency on fuel already piped to the site, with no moving parts, no grid dependence and cooling produced from exhaust heat that would otherwise be wasted, is an existential threat to a utility model built on metered dependence. Japan aimed for millions of residential units. South Korea is building toward fifteen gigawatts. The United States is at zero. That gap is not a technology gap. It is a policy gap. And policy gaps close when enough people stop waiting for permission and start demanding it.
Sources: The titles are:
“The Other Half of the Flame: Why Data Centers Are About to Get Cooling for Free” South Carolina Bulletin, July 31, 2026 https://samueleburns.substack.com/p/the-other-half-of-the-flame
“The Fuel Cell Revolution: How Data Centers Could Solve the Grid Crisis Instead of Breaking It” South Carolina Bulletin, July 29, 2026 https://samueleburns.substack.com/p/the-fuel-cell-revolution
A search on the South Carolina Bulletin site or a general web search with those exact titles and dates should pull them up.
