Energy · Deep dive

Voya Energy

A one-year-old California startup that raised $35M Series A in August 2026 (on top of a $13M November 2025 seed) to convert low-grade scrap aluminum into onsite 2 MW backup power for data centers via a metal-air electrochemical generator — reviving a category, aluminum-fuel electricity, that has been publicly promised and publicly missed for the better part of half a century.

emerging

The question that decides it: Does aluminum-water/aluminum-air electrochemistry actually beat lithium-ion batteries plus gas peakers plus diesel gensets on delivered $/kWh at the 4-hour backup duration data centers actually underwrite — once you charge full cost for scrap-aluminum feedstock volatility, the parasitic hydrogen evolution and passivation losses that have crippled every prior aluminum-fuel programme (Alcan/Alupower 1980-95, Phinergy 2014-25), the round-trip energy cost of recycling aluminum trihydrate back to metal, and the reactor-plus-container capex — and can Voya prove it inside the 2027 pilot window before diesel-plus-gas turbines eat the data-center backup TAM anyway?

My take ★★ Very interesting ★ Interesting − Maybe ✗ Not interesting

Snapshot

Voya Energy is a California startup — founded in 2025, out of stealth November 12, 2025 — that on August 26, 2026 announced a $35M Series A led by Energy Impact Partners with John Doerr, Mantis VC, StepStone, Founders Fund, Overmatch and Seven Stars, on top of a $13M seed in November 2025 led by the same lead. Its product is a metal-fuels energy system: scrap aluminum, processed into high-density pellets, is metered into a metal-air electrochemical cell with a proprietary liquid electrolyte, air and water; the reaction produces electricity, waste heat and inert aluminum trihydrate powder, with no combustion and no point-of-use emissions. The industrial-scale unit is a 20-foot shipping container rated at up to 2 MW. Voya says pilots begin in 2027 with manufacturing scale-up in 2028. The pitch is precise and topical: replace the diesel genset behind every data center in America. The category, aluminum-fuel electricity, has been publicly promised and publicly missed by better-capitalised teams for almost half a century, and no honest read of this company can ignore that.

Founding story

The three co-founders are the story. Richard Wang founded Cuberg in 2015 out of his Stanford PhD, ran a lithium-metal battery startup to ~225 people, and sold it to Northvolt in March 2021 — a genuinely good outcome for a hard-tech founder, and one of the reasons EIP is willing to lead two rounds inside nine months. Wang then ran North American operations for Northvolt and stood up its Silicon Valley advanced-technology centre, exiting as Northvolt’s core European operation ran into trouble.

Matt Horton is the commercial spine. 25 years across Propel Fuels (CEO), Proterra (CCO), Rivian (EVP Energy & Charging, built the Adventure Network and Waypoints businesses) and most recently CEO of Voltera, the fleet-charging infrastructure company, where he secured >$300M of customer contracts and raised >$150M of project debt and grants. He is exactly the profile Voya needs: someone who has actually sold and financed physical energy infrastructure at hyperscaler and fleet scale.

Steven Kaye is the electrochemist. Battery-materials R&D at Wildcat Discovery, then Apple’s Special Projects Group (the car programme), then CTO at Our Next Energy from February 2021 until his resignation in February 2024. The ONE departure is worth naming — ONE’s Aries and Anthem cell programmes have been repeatedly delayed and the company has laid off staff since — because it shows Kaye’s Voya bet is a deliberate escape from a lithium roadmap he ran, into a metal that is not lithium.

There is no romantic origin anecdote in the coverage. What connects the three is a shared conclusion, arrived at from three different sides of the battery industry: that lithium-based chemistries cannot cost-effectively deliver the multi-hour, deployable, off-grid firm power that data centers need in the volumes they now need it. So they picked a different metal.

How it works

Voya’s system has two parts. First, the fuel: low-grade scrap aluminum — the industry’s dross, turnings, used beverage cans, off-spec castings — is processed into dense, engineered pellets designed for shipping-container storage. The company says the pellets can be stored indefinitely without capacity degradation, which is the single largest operational advantage over lithium (which self-discharges) or hydrogen (which leaks and embrittles steel).

Second, the generator. Inside a standard 20-foot ISO container Voya houses a stack of metal-air electrochemical cells. Pellets are metered in and react with air and water in the presence of a proprietary liquid electrolyte. The cell converts the chemical potential of aluminum directly into electricity at a low temperature — Voya’s language is “no combustion, no noise, no atmospheric emissions at point of use.” The byproduct is aluminum trihydrate (ATH), an inert white powder used commercially in flame retardants, water treatment, ceramics and antacids. Voya claims a single 20-foot container will deliver up to 2 MW; the company’s marketing further claims ~100 MW of nameplate and ~10 GWh of stored energy per acre at scale.

The mechanics matter because they define the honest failure modes. Aluminum is famously energy-dense per unit mass — the top of the metal-fuel chart — but only if the passivation layer that instantly forms on any exposed aluminum surface is removed and kept off. Every aluminum-air programme of the last 45 years — Alcan/Alupower 1980-95, Phinergy 2014-25 — has had to fight three specific parasitic losses: passivation (aluminum hydroxide forms a barrier on the anode and blocks further reaction), aluminum self-corrosion (the metal reacts with water without producing current, so fuel disappears into hydrogen and waste heat), and CO2 contamination of the electrolyte (which fouls the air cathode). Voya’s technical claim is that its proprietary electrolyte and cell architecture manage all three. That is the exact claim every predecessor has made.

Product and business overview

The commercial offer is not fully public, but three components are visible.

The fuel. Aluminum pellets, produced from scrap. This is a consumable, not a battery — pellets are shipped in, ATH is trucked out, and there is presumably a closed loop (or a partnership with a smelter) that converts ATH back to aluminum metal. The energy cost of that reduction step is the hidden $/kWh in the whole model (see below).

The generator. A containerised 2 MW electrochemical unit, sold or leased to a site. Voya positions it as “grid-independent” and modular — data centers, industrial manufacturing, heavy transportation depots, utilities, real-estate, emergency response.

The service layer. Voya has said it has “nearly a dozen” pilot partners across data centers, industrial manufacturing, transportation, utilities, real estate, emergency response and energy infrastructure — none named publicly as of August 2026. The 2027 pilot cadence suggests Voya is selling early sites at a subsidised price or as an EPC-plus-fuel service, with pellets billed on a $/kWh delivered basis rather than a hardware sale.

Business model and pricing

No public price sheet. What is public is the goal — displace diesel — which anchors the arithmetic. Diesel genset backup for data centers is typically bought at roughly $300-500/kW installed (Cummins/MTU/Caterpillar Tier 4 units, per multiple 2026 industry procurement guides), fuelled at diesel rack prices of ~$3.50-4.50/gal, and used for a few tens of hours a year for backup — implying a lifetime $/kWh dominated by capex, not fuel. A 2 MW diesel unit is roughly $600K-1M installed.

Voya must therefore either come in at a comparable capex-per-kW while producing dramatically cheaper permitting, air-quality and noise outcomes (a plausible pitch, because California and Northern Virginia data-center permitting is now diesel-constrained), or come in at higher capex but win on Scope 1 emissions and displaced fuel cost for customers running gensets more than a couple of hundred hours a year. The company has not published either number.

The uncomfortable arithmetic is on the fuel side. Primary aluminum smelting via Hall-Héroult costs ~13,000-17,000 kWh of electricity per tonne (International Aluminium Institute; ACEEE 2003; multiple 2025 reviews). One tonne of aluminum, fully oxidised, releases roughly 8.6 MWh of chemical energy — which sets a hard theoretical ceiling of ~0.5 MWh out per MWh of smelting electricity in even before you count cell efficiency, parasitic hydrogen losses, balance-of-plant and ATH-to-aluminum recycling. If Voya uses scrap aluminum, secondary aluminum melting requires only 700-800 kWh per tonne (5-8% of primary), which is a genuine differentiator — but scrap aluminum has a real market price ($0.80-1.20/lb in mid-2026 LME-linked pricing) and a growing recycled-content bid from the beverage-can, automotive and construction industries. Voya must either lock in feedstock supply at industrial prices, or scale up its own closed-loop ATH-to-Al recycling, which reintroduces the primary-smelting energy problem. Neither is impossible; neither has yet been publicly resolved.

Traction over time

Two honest observations. First, this is nine months of company history from seed to Series A — the round cadence is that of a founder profile investors want to underwrite, not that of a company with demonstrated field data. Second, no MW has been deployed at a customer site as of the Series A announcement.

Market analysis

The addressable pond is deep and growing. The data-center backup-generator market was estimated at ~$8.57B in 2026 (MarketsandMarkets, 2026-31 report), with diesel alone ~$5.79B; broader industry estimates put the total at $7.6-10.3B for 2025-26 (Fortune Business Insights, Grand View, MarknTel). The data-center fuel-cell and alternative-backup segment is projected at ~19.6% CAGR through 2032 (MarketsandMarkets 2026). BloombergNEF projected in 2025-26 that the US will still face a ~19 GW data-center power shortfall by 2035 even assuming aggressive grid expansion plus hyperscaler-owned gas turbines. Latitude Media has separately documented the “data-center boom is a diesel-genset boom” reality, with delivery lead times for large gensets stretching past 24 months as of 2026.

The structural forces are all in Voya’s direction: (i) Northern Virginia, California, Washington State and the Dublin/Ireland cluster are now permitting-constrained on new diesel; (ii) grid interconnect timelines for new data centers regularly exceed 5 years; (iii) hyperscalers have published Scope 1 net-zero commitments diesel gensets directly threaten; (iv) gas-turbine lead times are now 3-5 years for the biggest frames. Any technology that ships onsite, non-combusting, containerised, non-permitting-triggering firm power at MW scale has a real buyer.

The counterforces are equally structural: (i) the diesel and turbine incumbents (Caterpillar, Cummins, MTU, GE Vernova, Siemens Energy, Mitsubishi Power) have decades-deep OEM channels and known reliability; (ii) Bloom already has a 1 GW purchase order from AEP; (iii) Form Energy already has a 30 GWh Google/Xcel purchase order; (iv) large lithium-ion BESS deployments now clear at ~$100-150/kWh installed and are the industry-default 4-hour block.

Competitive intel

The honest competitive frame has four rings.

Ring 1: other aluminum-fuel startups. Phinergy (aluminum-air, Israel, since 2008; Rosendin data-center partnership targeting first megawatt-scale unit mid-2026) is the direct peer. Found Energy / Found Industries (aluminum-water, MIT-spinout, Peter Godart, ~$12M seed 2023) is running the largest aluminum-water reactor announced to date. Alumina Energy (thermal storage in aluminum) and Trolysis (aluminum-water hydrogen for fuel cells) sit adjacent. None has meaningful MW-scale deployed field data as of August 2026. That is either a category opportunity or a category verdict.

Ring 2: other clean firm/backup electrochemical. Bloom Energy (SOFC; $7.65B in disclosed data-center deals through 2026; 1 GW AEP procurement November 2024) and Plug Power (PEM; late-2025 data-center push) are the incumbents in electrochemical onsite backup, running on natural gas or hydrogen. They have purchase orders Voya does not.

Ring 3: other long-duration/firm players. Form Energy (iron-air, 100-hour; $1.2B+ raised; 30 GWh Google/Xcel Minnesota deployment announced February 2026, plus 100 MW California). Antora ($550M Series C July 2026; carbon-block thermal + TPV heat-to-power). Rondo (thermal). These do not compete on the exact same duration profile as Voya’s 2 MW pitch, but they compete for the same “firm clean onsite” capital and the same “beyond lithium” investor thesis.

Ring 4: the actual incumbent. Diesel gensets and natural-gas turbines. Cummins, Caterpillar, MTU, Kohler, Generac at the diesel end; GE Vernova, Siemens Energy, Mitsubishi Power, Solar Turbines at the turbine end. This is the ring that will decide whether Voya matters — because if hyperscalers can hold their nose on Scope 1 and buy diesel for another decade while grid interconnects catch up, no metal-fuel startup gets to industrial scale.

History and evolution

• 1980-1995: Alcan subsidiary Alupower runs the first major aluminum-air research programme post-1973 oil embargo. Programme discontinued. The reason repeated in the academic literature: passivation and self-corrosion could not be economically managed at production scale.
• 2008-2014: Israeli startup Phinergy is founded; aluminum-air demo car with Alcoa announced 2014, projecting ~1,000-mile range.
• 2014-2017: Phinergy signs Renault-Nissan partnership targeting 2017 production. Timeline missed; company pivots away from EV toward stationary/backup applications.
• 2022: Peter Godart spins Found Energy out of MIT for aluminum-water reactors delivering heat and hydrogen; $12M seed follows in 2023.
• 2024-25: Rosendin (US electrical contractor with heavy data-center exposure) partners with Phinergy to deliver first megawatt-scale aluminum-air backup at a US data-center client site targeted for mid-2026.
• 2025 (early): Voya Energy incorporated in California by Wang, Horton and Kaye.
• Nov 12, 2025: Voya Energy exits stealth with $13M seed led by Energy Impact Partners.
• Aug 26, 2026: Voya Energy announces $35M Series A led by EIP with John Doerr, Mantis VC, StepStone, Founders Fund, Overmatch, Seven Stars; discloses “nearly a dozen” pilot partners and unveils the 2 MW / 20-foot-container industrial-scale reference design.

What people say

The case for. Investor conviction is genuine and unusually concentrated. Energy Impact Partners led both rounds inside nine months, a signal that the strategic LPs behind EIP — mostly major North American investor-owned utilities — see this as a defensive bet on being early to whatever replaces diesel. John Doerr writing a personal cheque alongside Founders Fund and Mantis is a founder-quality signal (Doerr almost never does personal energy angel deals). Latitude Media’s August 2026 Catalyst episode with Richard Wang is the most substantive public discussion; the reception in the climatetech press (Bloomberg, ESG Today, Mining.com, Canary Media adjacent coverage, Mercom) has been broadly favourable and unusually specific about the “diesel-genset boom is a data-center boom” backdrop that gives Voya its wedge. The founder set — Wang shipped and sold Cuberg; Horton has actually financed hundreds of millions of dollars of physical energy infrastructure at Voltera and Rivian; Kaye is one of the more credentialed battery-materials chemists in Silicon Valley — is at the 90th percentile for a pre-pilot Series A.

The complaints. Almost none of the skepticism is in the recent press coverage, but it is very much in the academic literature and the historical record. Aluminum-air and aluminum-water chemistries have three unsolved parasitic loss modes — passivation, self-corrosion (hydrogen evolution), and CO2 fouling of air cathodes — that Alcan’s Alupower could not economically manage in fifteen years of programme spend from 1980-1995, and that a 2026 Frontiers review still identifies as the principal barriers to commercialisation. Phinergy’s public trajectory is instructive: 2014 range-extender demo car with Alcoa, 2015 Renault-Nissan partnership targeting 2017 production, no shipped product, pivot to stationary backup, first commercial megawatt unit still not on a site as of August 2026 — 18 years and counting. The founder-of-record on aluminum-fuel-cell technology, Phinergy’s Aviv Tzidon, has openly acknowledged the “stigma” of the previous 20-year failure cycle. The lifecycle-economics complaint is the sharpest and rarely stated in Voya coverage: aluminum is a battery of the electricity that smelted it (~13-17 MWh/tonne primary; ~0.7-0.8 MWh/tonne secondary from scrap), which means the true $/kWh depends entirely on whether you win on scrap-price arbitrage plus a favourable ATH offtake, and whether your electrolyte and cell design actually suppress the corrosion side reactions in the field, not just in the lab. Neither has been demonstrated at pilot scale in public. Voya has raised $48M in nine months from investors who have a strong prior on the team; it has yet to install a single MW in a customer’s yard.

Outlook: the open question

The open question resolves in the 2027 pilot window, and it resolves specifically on three numbers Voya has not published: cell-level round-trip efficiency, delivered $/kWh net of ATH recycling, and time-to-first-fill after container commissioning. Everything else is downstream. If Voya can show even one hyperscaler that a 2 MW containerised unit ships onsite in weeks (not the 24-month lead time on Tier 4 diesel), operates without a Title V or Bay Area Air Quality permit fight, delivers firm power for a duration ceiling in the 8-24 hour range at less than $0.30-0.50/kWh delivered, and does it without the passivation degradation that killed Alupower, then the category is real and Voya’s founder profile plus EIP’s utility LP set gives it distribution the earlier aluminum-fuel programmes never had. If any one of those numbers misses — cell efficiency runs below 50% wall-to-wall, ATH offtake pricing collapses, cell degradation looks anything like Phinergy’s, or the 2027 pilots slip 12 months — Voya joins a very long list of aluminum-fuel companies whose main asset was a decade of investor patience. The bear case is not that the science is wrong; it is that the science is a decade from commercial and the data-center backup market will have bought diesel plus BESS plus SOFC before then.

How to attack it

Skip aluminum entirely and attack the wedge directly. Voya’s wedge is not aluminum; it is “onsite, non-combusting, non-permitting-triggering, containerised, firm power at MW scale, deliverable in weeks not years.” A funded attacker who takes that wedge seriously does not necessarily need a novel chemistry — the fastest path is a modular hybrid of already-shipping components: an oversized lithium-iron-phosphate BESS block (currently clearing at ~$100-150/kWh installed) sized for four hours of runtime, paired with a hydrogen or ammonia PEM/SOFC fuel-cell top-up for the long tail, plus a proprietary power-electronics and controls layer that presents a single 2 MW containerised UPS-plus-genset replacement. That system exists today in components; the wedge is packaging and permitting. A challenger who ships the container in 2026 while Voya is still building 2027 pilots eats the entire “diesel replacement in permit-constrained markets” segment before aluminum can prove itself. The specific weaknesses in Voya’s current position that a challenger exploits: (i) zero deployed MW as of August 2026; (ii) chemistry with 45 years of scarred history and specifically-named unsolved parasitic-loss modes (passivation, hydrogen evolution, CO2 fouling per multiple 2025-26 reviews); (iii) unresolved feedstock strategy — scrap aluminum has an active LME-linked commodity price and rising recycled-content demand from beverage-can, auto and construction; (iv) unpublished cell efficiency and $/kWh delivered; (v) a customer set (data centers) that buys the incumbent (diesel) because it is what their operations, insurance and permitting teams already model. A well-capitalised attacker also lifts the packaging idea (20-ft ISO container, drop-in), the hyperscaler go-to-market (Horton’s Voltera and Rivian playbook is now public), and the “firm clean onsite” investor narrative that has funded Voya, Form, Antora, Bloom and Base Power in parallel — and skips the metal-fuel science project.

Adjacent-segment play

The aluminum-pellet-plus-metal-air-cell stack has more natural adjacencies than the diesel-replacement pitch does. Three axes are attractive. First, mobile heavy transport: aluminum’s mass-energy density (~8.7 kWh/kg theoretical; ~1-4 kWh/kg practical) is materially higher than lithium-ion (~0.16-0.30 kWh/kg) and would suit off-grid heavy trucking, mining haul trucks, and marine applications where energy per kg matters — Phinergy has spent a decade on this and Trolysis is doing the aluminum-water-to-H2-fuel-cell version, but the buyer is different and the willingness to accept a consumable-fuel model is much higher. Second, disaster response and expeditionary power: shipping-container units that store pellets indefinitely, ship dry, do not leak, do not require a fuel-supply chain and produce inert waste are structurally attractive to FEMA, DOD, humanitarian logistics — the IQT participation in the seed round is a tell that this vector is already being scoped. Third, industrial process heat: Alumina Energy already exists in the thermal-storage-in-aluminum niche, and Found Energy has already pivoted toward industrial heat plus hydrogen — the same reaction Voya is trying to convert to electrons is arguably a better product delivered as heat, because it skips the electrochemical efficiency tax entirely. Adjacent-segment risk: none of these buyers underwrites the $10B+ TAM the data-center pitch does, which is precisely why the founders chose data centers. If the data-center pitch is what raises the money, mobile and industrial-heat is what actually clears a first pilot.

Sources and further reading

• Bloomberg — Voya Energy raises $35M to develop aluminum power for data centers (August 26, 2026)
• Voya Energy — Company press release: Series A $35M and industrial-scale metal-fuels system (August 26, 2026)
• ESG Today — Voya Energy raises $13M seed to turn metal into carbon-free fuels (November 14, 2025)
• Voya Energy — Launch press release (November 12, 2025)
• Interesting Engineering — Aluminum fuel converts scrap metal into 2 MW of clean firm power (August 2026)
• Mining.com — Clean tech startup scores $35M for aluminum-based energy system (August 2026)
• Latitude Media — Catalyst episode: the rise of metal fuels with Richard Wang (August 2026)
• Latitude Media — The data center boom is a diesel generator boom (2026)
• MIT News — Found Industries aims to strengthen America’s industrial supply chains (May 3, 2026)
• MIT Technology Review — Startup about to conduct biggest real-world test of aluminum as a zero-carbon fuel (October 23, 2025)
• Rosendin — Partners with Phinergy to roll out aluminum-air backup for US data center market (2024-25)
• pv magazine USA — Google to deploy world’s largest iron-air battery for Minnesota data center (February 24, 2026)
• Business Wire — Bloom Energy announces gigawatt fuel cell procurement agreement with AEP (November 14, 2024)
• ESG Today — Antora raises $550M Series C for thermal batteries (July 30, 2026)
• Frontiers in Batteries & Electrochemistry — Challenges and strategies for aluminum-air batteries (2026)
• Lithium Horizons — A Short History of Aluminum Batteries (Alcan/Alupower, 1980-95)
• International Aluminium Institute — Aluminium recycling saves 95% of the energy of primary production