Hydrogen is an excellent fuel and a difficult cargo. It carries more energy per kilogram than any other fuel, and almost none per liter unless you compress it to hundreds of bar or cool it to within about 20 degrees of absolute zero. Aluminum offers a different route: ship and store a dense, inert metal, and release the hydrogen from water only where and when you burn it. This piece explains the public chemistry, the arithmetic, why the idea has stayed on the shelf, and where our own system sits.
It is written for investors and integrators evaluating on-site hydrogen for AI compute. Every figure below is either sourced or computed in the open so you can check it. Where a statement describes our system, it describes a design, and we say so.
How Does Aluminum Produce Hydrogen?
Aluminum is a reactive metal. Given direct contact with water, it takes the oxygen and releases the hydrogen. The U.S. Department of Energy's 2008 white paper on the subject, written by John Petrovic and George Thomas as consultants to the DOE Hydrogen Program, gives the three textbook reactions (DOE, Reaction of Aluminum with Water to Produce Hydrogen, 2008):
- 2Al + 6H2O → 2Al(OH)3 + 3H2, forming aluminum hydroxide
- 2Al + 4H2O → 2AlO(OH) + 3H2, forming the oxyhydroxide boehmite
- 2Al + 3H2O → Al2O3 + 3H2, forming aluminum oxide, or alumina
The paper notes that all three are thermodynamically favorable from room temperature past aluminum's melting point, and all are highly exothermic. Which solid forms depends on conditions. Whichever path runs, the hydrogen yield per unit of aluminum is the same: two aluminum atoms release three hydrogen molecules.
That makes the yield a matter of arithmetic. Using aluminum's relative atomic mass of 26.982 (Royal Society of Chemistry) and 2.016 for H2:
| Quantity | Calculation | Result per kg of aluminum |
|---|---|---|
| Hydrogen released | (3 × 2.016) ÷ (2 × 26.982) | 0.112 kg H2 |
| Aluminum per kg of hydrogen | 1 ÷ 0.112 | 8.9 kg Al |
| Fuel energy in that hydrogen | 0.112 kg × 120 MJ/kg (lower heating value) | 13.5 MJ, or 3.7 kWh |
| Heat released by the reaction | 55.6 mol H2 × about 280 kJ/mol H2 | about 15.6 MJ, or 4.3 kWh |
The 120 MJ/kg figure is DOE's (DOE, Hydrogen Storage). The heat figure uses the white paper's thermodynamic table, which puts the reaction enthalpy for the hydroxide path at roughly 277 to 291 kJ released per mole of hydrogen between 0 and 200 °C. The DOE paper's own round number, about 9 kg of aluminum per kilogram of hydrogen, matches our 8.9.
The heat row is the one to notice. Per kilogram of aluminum, the reaction releases about as much energy as heat as it releases in the hydrogen. A system that throws that heat away is discarding roughly half of what the aluminum carried in.
Why Hasn't Aluminum-Water Hydrogen Been Used at Scale?
Three reasons, in the order they bite.
The oxide film stops the reaction
Aluminum forms a microscopically thin oxide layer on contact with air, the same layer that keeps it from corroding. The DOE white paper states that a piece of aluminum dropped into water will not react at room temperature, “or even with boiling water”, because this coherent, adherent layer keeps water away from the metal. Keeping the reaction going means continually removing or disrupting that film.
Every way around the film has carried a cost
The literature groups activation approaches into physical methods (fine powders, milling, elevated temperature and pressure) and chemical ones (promoters, additives, alloying), as summarized in a 2025 review in the International Journal of Thermofluids (“A review of hydrogen generation methods via aluminum-water reactions”). The best-known recent work is MIT's, where Douglas Hart's group used a gallium-indium alloy that is liquid at room temperature, which permeates the metal's grain boundaries and exposes fresh aluminum (MIT News, August 2021). Gallium and indium are expensive, and MIT's 2024 follow-up focused on recovering the alloy after reaction because, in the team's words, that was required for the idea to be “cost-effective and sustainable” (MIT News, July 2024). As of 2008, DOE's assessment was that none of the approaches it reviewed had proven commercially viable.
The metal is the fuel cost
DOE evaluated the idea mainly for on-board vehicle storage and concluded it could not meet the program's weight, volume, flow-rate or cost targets. On cost, it found the price of the hydrogen “is dictated by the cost of aluminum metal”. It also concluded that aluminum-water hydrogen “might have utility for non-vehicular applications”. A stationary power plant, where weight does not matter and the heat and the solid product both have uses, is one.
There is a fourth point that is not an obstacle so much as a boundary condition. Making aluminum takes a great deal of electricity. The same DOE paper used a North American average of 15.552 kWh per kilogram for smelting in 2005 and calculated that hydrogen made from freshly smelted aluminum would have an overall energy efficiency of about 24%. Aluminum is an energy carrier, not an energy source. The argument for it is logistics, storability and co-product value, not free energy, and the economics depend on the feedstock and on getting paid for everything the reaction produces.
Why Is Aluminum a Good Way to Store Hydrogen?
Because it is dense, it is stable at ambient conditions, and it does not need a pressure vessel or a cryogenic tank until it becomes hydrogen.
Aluminum's density is 2.70 g/cm³ (RSC). Multiply by the 0.112 yield and one liter of solid metal carries about 300 g of releasable hydrogen. For comparison, DOE gives liquid hydrogen an energy density of 8 MJ/L (DOE), which at 120 MJ/kg is about 67 g/L, and it lists a 700-bar Type IV compressed tank system at 0.024 kg of hydrogen per liter of system volume (DOE, Physical Hydrogen Storage). The MIT team put the gain at storing hydrogen “at a density that's 10 times greater” than compressed gas.
Read the 300 g/L figure honestly. It counts the metal only. The water that supplies the hydrogen, the reactor, and the space for the solid product are all excluded. Once DOE included the water for the hydroxide reaction, its figure fell to 3.7% hydrogen by weight and 46 g of hydrogen per liter.
For a vehicle, that gap is decisive. For a fixed site with water recovery, the relevant comparison is what you have to ship and store, and there the dense solid is the whole point.
The other half of the argument is what aluminum does not need. Hydrogen gas is stored at 350 to 700 bar, and liquid hydrogen at −252.8 °C, its boiling point at one atmosphere (DOE). Aluminum is stored on a pad or in a building, does not boil off, and does not lose pressure over months.
Evaluating on-site hydrogen for a site you own or are developing? See how our energy system is supplied and integrated for other operators, not only our own campuses.
On-Site Hydrogen Power →What Comes Out Besides Hydrogen?
Two things, and both matter to the economics.
- Heat. About 15.6 MJ per kilogram of aluminum, as computed above. Useful heat, if the system is designed to use it.
- A solid aluminum oxide or hydroxide. By mass it outweighs the metal that went in, because it has taken on oxygen and, for the hydroxide, hydrogen: about 1.89 kg of alumina or 2.89 kg of aluminum hydroxide per kilogram of aluminum, by the same molar-mass arithmetic.
The solid is not waste. The DOE white paper points out that the hydroxide from these reactions is essentially the same material as the hydrate made from bauxite in conventional alumina refining. Alumina is a traded industrial material, and in our design it is a co-product sold as a second revenue line. We describe it as high-value alumina, grade subject to independent assay, and we do not state a grade until an independent assay does.
Water is the input to watch. By stoichiometry, the oxide path consumes about 1.0 kg of water per kilogram of aluminum and the hydroxide path about 2.0 kg, which works out to roughly 9 to 18 kg of water per kilogram of hydrogen. Electrolysis consumes about 8.9 kg, by the same arithmetic. So an aluminum reactor is not a way to avoid water. It is a process that uses water, which is why our reactor design recovers and recirculates it. That is a different claim from the one our cooling makes: the compute floor uses no evaporative cooling water, covered in our post on the self-powered data center.
How Does It Compare With Other Ways to Supply Hydrogen to a Data Center?
A site that wants hydrogen has four practical routes. The table is generic on purpose. It describes the routes, not any supplier.
| Route | Storage form on site | What you have to build | Energy and grid dependency | Co-products | Main constraint |
|---|---|---|---|---|---|
| Delivered compressed H2 | Gas at high pressure. Tube trailers are limited to 250 bar by U.S. DOT rules unless exempted. | Receiving and high-pressure storage, plus a delivery contract. | Compression energy upstream. None on site beyond handling. | None | Volume. A steel tube trailer carries about 380 kg; composite trailers 560–900 kg (DOE). |
| Delivered liquid H2 | Cryogenic liquid at −252.8 °C. | Insulated storage, vaporization, delivery contract. | Liquefaction upstream consumes more than 30% of the hydrogen's energy content (DOE). | None | Boil-off: some stored hydrogen is lost to evaporation (DOE). |
| On-site electrolysis | Usually compressed gas, as a buffer. | Electrolyzers, water treatment, compression and storage. | Heavy. DOE uses about 55.2 kWh of electricity per kg of hydrogen for a PEM system at beginning of life (DOE Program Record 24005). | Oxygen | It needs large amounts of electricity, which is the very thing a power-constrained site is short of. |
| On-site from aluminum | Solid metal at ambient conditions; hydrogen generated as consumed. | Reactor, water recovery, solids handling. | Low on site. The energy was spent upstream, in smelting the metal. | Heat and alumina | Aluminum supply and its cost, and co-product handling at scale. |
The size of the logistics problem is easiest to see with an illustrative number. Assume a generator converts 40% of hydrogen's lower heating value to electricity. That is an assumption for this example, not our specification. One megawatt-hour of electricity then needs 3,600 MJ ÷ 0.40 ÷ 120 MJ/kg = 75 kg of hydrogen. A 10 MW load running for 24 hours needs about 18 tonnes a day. Delivered by steel tube trailer, that is roughly 47 trailers a day; by composite trailer, 20 to 32. From aluminum, it is about 160 tonnes of metal a day. Move the efficiency to 35% or 45% and the aluminum figure runs from about 143 to 183 tonnes. None of these is small. The difference is that bulk metal can be stockpiled for months without pressure, refrigeration or loss.
How Does Aluminum Power a Data Center?
In our architecture, the aluminum does not power anything directly. It is one of two fuel paths into a single electrical bus.
- Pipeline natural gas feeds modular, fuel-flexible linear generators. This is the bridge that runs from day one, before the hydrogen path is at full output.
- Aluminum goes into a proprietary AJ Projects reactor. One reaction gives three outputs: hydrogen, alumina and heat. The hydrogen fuels the same generators, on site.
- Heat out, fuel in. Generator exhaust heat is routed back to drive the reactor, so heat a generator would otherwise reject becomes an input to making the next unit of fuel.
- The alumina leaves the site as a product, grade subject to independent assay.
Because the generators are fuel-flexible and share one bus, the mix between gas and hydrogen is an operating decision rather than a rebuild. The array is designed N+1, and five-nines availability is a design target. The unit of deployment is a repeatable 10 MW self-powered pod; a 73 MW self-generating campus in North Texas is in development and pre-construction. The site is designed to run without the grid. On a modelled basis, the operating cost of self-generated power (fuel plus O&M, running mainly on pipeline gas) is about $0.044/kWh, against a Texas commercial grid rate of roughly $0.070/kWh. That excludes generator capital, and it is a model, not a metered result.
We deliberately do not describe how the reactor overcomes the oxide film, what goes into it, or how it is run. That is the proprietary part. What we can say is that it is a separate question from the one the published research answers. The chemistry is public; the engineering to run it continuously, recover the water and deliver a saleable solid is what has to be proven.
The energy system is also not reserved for our own compute. Other developers and operators can integrate it at their sites on a non-exclusive basis: we supply and integrate the energy system, and they own or operate the compute. For the per-rack arithmetic that sets how much power a site needs in the first place, see how much power an AI data center needs.
Is Aluminum-to-Hydrogen Safe?
The hazards are real and well characterized, and they sit in different places than they do for delivered hydrogen.
- Bulk aluminum is stable. The same oxide film that frustrates the chemist protects the stockpile. DOE's paper is explicit that ordinary aluminum does not react with water, even boiling water, without its film being disrupted.
- Fine aluminum dust is a combustion hazard. OSHA notes that materials that do not burn in larger pieces, “such as aluminum or iron”, can be explosible in dust form (OSHA, Combustible Dust). Any process that produces or handles fine particles has to be designed and managed as a combustible-dust operation.
- Hydrogen is still hydrogen. DOE's safety fact sheet gives a flammability range of 4–74% in air and a very low ignition energy, and notes that hydrogen disperses quickly when released into open air (DOE, Hydrogen Safety fact sheet). Generating hydrogen as it is consumed is intended to keep the on-site inventory small, but the reactor, piping and generators are hydrogen systems and are engineered as such.
- The reaction is exothermic. Heat management is a safety requirement as well as an efficiency opportunity.
The fair summary: aluminum shifts the hazard from storing a large volume of pressurized or cryogenic gas to handling a solid safely and controlling a reaction. It is not hazard-free, and anyone who tells you otherwise has not built one.
What Is Still Unproven?
A pilot reactor is running, with instrumented reaction and water-recovery trains. Here is what it does and does not establish.
| Question | Where it stands |
|---|---|
| Does the reaction produce hydrogen from aluminum? | Settled public chemistry, and running in our pilot. |
| Can the water be recovered and recirculated? | The pilot has an instrumented water-recovery train. Recovery performance at campus scale is not yet demonstrated. |
| What grade is the alumina? | Not yet proven. We will not state a grade until an independent assay does. |
| Does it hold at campus scale? | Not yet proven. The pilot does not demonstrate campus-scale output, and the 73 MW campus is in pre-construction. |
| Is the aluminum supply secure and priced to work? | A site-level question. As DOE found, the metal sets the fuel cost, so feedstock sourcing is part of every project's model. |
| What is the carbon footprint? | It depends on how the aluminum was made. Smelting is electricity-intensive, and its footprint follows its power source. |
That is the honest position of the field, not only of our project. The literature has shown the reaction many ways. What has been missing is a stationary system built to use every output, the hydrogen, the heat and the solid, at a scale where the logistics advantage matters. That is what we are building, and what a developer evaluating it should ask to see measured.
Frequently Asked Questions
Aluminum reacts with water, taking the oxygen and releasing hydrogen gas, heat, and a solid aluminum hydroxide or aluminum oxide. Two aluminum atoms release three hydrogen molecules, so by stoichiometry one kilogram of aluminum yields about 0.112 kg of hydrogen, and about 8.9 kg of aluminum makes one kilogram of hydrogen. The reaction is thermodynamically favorable and highly exothermic, but it does not proceed on ordinary aluminum because a thin oxide film on the metal keeps water away. Every practical system is, at its core, a way of dealing with that film.
Three reasons. First, the oxide film stops the reaction, and DOE's 2008 white paper notes that aluminum will not react even with boiling water unless the film is disrupted. Second, the published ways around it, from fine powders to promoters to liquid-metal alloys such as the gallium-indium approach studied at MIT, have each carried a cost; as of 2008, DOE found none had proven commercially viable. Third, DOE concluded that the cost of the hydrogen is dictated by the cost of the aluminum. DOE judged it unsuitable for on-board vehicle storage but noted it might have utility for non-vehicular applications.
Because it is dense and stable. At 2.70 g/cm³ and a yield of 0.112 kg of hydrogen per kilogram, one liter of solid aluminum carries about 300 g of releasable hydrogen, against about 67 g/L for liquid hydrogen based on DOE's 8 MJ/L figure. That count excludes the water, the reactor and the solid product; with the water included for the hydroxide reaction, DOE's figure is 46 g/L. The practical advantage for a fixed site is that aluminum needs no pressure vessel, no cryogenic tank and suffers no boil-off, so it can be shipped and stockpiled like any bulk metal.
Heat and a solid. The heat is about 15.6 MJ per kilogram of aluminum, roughly as much energy as the hydrogen itself carries. The solid is aluminum hydroxide or aluminum oxide, depending on conditions, and it weighs more than the aluminum that went in: about 1.89 kg of alumina or 2.89 kg of hydroxide per kilogram of metal. In our design, the alumina is sold as a co-product, described as high-value alumina with its grade subject to independent assay. Water is consumed by the reaction, which is why a well-designed system recovers and recirculates it.
In our architecture, aluminum goes into a proprietary reactor that produces hydrogen, alumina and heat. The hydrogen fuels modular, fuel-flexible linear generators on the same electrical bus as a pipeline natural gas path, which is the bridge that runs from day one. Generator exhaust heat is routed back to drive the reactor, which we describe as heat out, fuel in. The unit of deployment is a repeatable 10 MW self-powered pod designed to run without the grid, and the system is available to other developers and operators on a non-exclusive basis.
The hazards are well characterized and they move rather than disappear. Bulk aluminum is stable because of its oxide film, so the stockpile itself is not the concern. Fine aluminum dust is: OSHA notes that metals such as aluminum can be explosible in dust form, so any fine-particle handling must be run as a combustible-dust operation. Hydrogen has a flammability range of 4 to 74 percent in air according to DOE, and the reactor, piping and generators are engineered as hydrogen systems. Generating hydrogen as it is consumed is intended to keep the on-site inventory of stored gas small.
Scale and co-product grade. Our pilot reactor is running with instrumented reaction and water-recovery trains, but it does not yet prove the grade of the alumina or output at campus scale, and the 73 MW North Texas campus is in development and pre-construction. The aluminum supply and its price are a site-level question, because the metal sets the fuel cost. The carbon footprint depends on how the aluminum was made, since smelting is electricity-intensive. We would expect any serious developer to ask for measured results on each of those before underwriting a site.