One hundred hours. That is the storage duration Noon Energy claims its battery can hold — enough to carry an AI data center through four windless, sunless days without firing a gas turbine. The chemistry works. Scale is the open question.
The Palo Alto company spent eight years building a battery that stores energy in carbon-based media rather than lithium. Two deals this year turned the engineering into a commercial pipeline: a 100 GWh reservation from Meta in April, and a joint venture with Sabanci Renewables in August to co-develop up to 1 GW for AI infrastructure.
Noon's signal is contractual, not chemical: Meta's 100 GWh reservation and Sabanci's 1 GW venture are off-take arrangements, not research grants.
The near-term test is a single 25 MW project due by 2028, a reversible solid-oxide supply chain that has never run at scale, and a cost curve nobody has audited.
The gap lithium cannot close
Grid batteries have become a commodity. Lithium iron phosphate packs ship by the gigawatt-hour, cost a fraction of their 2020 price, and respond in milliseconds. Their weakness is duration. A typical utility battery discharges for two to four hours, then needs hours to refill.
Four hours is enough to move solar output into the evening peak. It is not enough to cover a multi-day lull in wind and sun. Data centers draw flat, near-constant load, and hyperscalers want clean power around the clock, not a fossil bridge for the days when the wind quits.
Above ten hours, the terrain gets contested. Iron-air, flow, thermal, and compressed-air designs all claim the long-duration niche. Few run at gigawatt scale, and almost none carry a signed contract with a hyperscaler.
Ultra-long-duration storage on reservation
Meta's April 2026 agreement reserves up to 1 GW / 100 GWh from Noon, beginning with a 25 MW / 2.5 GWh project due by 2028. · Noon Energy / pv magazine, 2026
Noon's pitch is duration measured in hundreds of hours, paired with a cost structure that does not balloon as storage hours are added.
It is early.
The first full-scale project under the Meta agreement is 25 MW and 2.5 GWh, due by 2028. The 1 GW / 100 GWh headline is a reservation, not an order book, and the commercial phase depends on that first build working.
That reservation sits inside a broader power problem. Hyperscalers are signing renewable PPAs faster than grids can connect them, and every additional wind farm adds hours of surplus generation that batteries cannot absorb. Curtailment is the flip side of the AI build-out: clean electricity produced when nobody needs it, and missing when the load peaks.
A battery that splits power from energy
Noon's machine is a reversible solid oxide fuel cell dressed as a battery. It has three parts. A power block converts electricity into stored chemical energy and back again. A charge tank holds carbon-based media that has absorbed the energy, venting oxygen to the air. A discharge tank takes oxygen back in to release stored energy as electricity.
The architecture is the interesting part. In a conventional battery, more duration means more cells, more electrodes, more of everything. Noon separates power from energy: the power block sets charge and discharge speed, the tank sets how long the system runs. Need more power? Add power blocks. Need more hours? Build a bigger tank.
That split bends the cost curve. Noon says its containerized pilot has run for thousands of hours and reached more than 200 hours of duration. It also claims the design is 20 to 200 times smaller than flow batteries and pumped hydro, and two to three times smaller than a lithium-ion system of comparable output. The storage medium is carbon-based, which keeps mined metals such as lithium, nickel, and cobalt out of the bill of materials.
Why not just stack more lithium cells?
A short detour into the cap table. Noon raised a $28 million Series A in December 2022, led by At One Ventures and Clean Energy Ventures, with Aramco Ventures among the backers. Nearly three years passed before the technology reached commercial agreements — a useful reminder of how slowly hardware crosses from demonstration to deployment.
The contract is the moat
Noon is not selling hardware into a spot market. It is selling firm power, wrapped in the contracts that make long-duration storage financeable. The Sabanci agreement pairs Noon's tanks with Sabanci Renewables' solar and wind portfolio, and the projects are structured as power purchase agreements.
Capacity under the August co-development deal
Noon and Sabanci Renewables will co-develop up to 1 GW of ultra-LDES projects, with commercial deployment possible from 2027. · pv magazine, 2026
The structure is the point. A developer holding an off-take contract can raise debt against it. A developer holding a demonstration unit cannot. Meta's reservation does the same work a utility contract does: it converts an engineering claim into a revenue line.
Sabanci Climate Ventures, an investor in Noon and a sister company to Sabanci Renewables, arranged the collaboration. The group is targeting 3 GW of US renewables within five years, which hands Noon a captive pipeline rather than a cold sales list.
Partnering with an established and forward-thinking company like Sabanci Renewables marks a critical step in accelerating the path to commercialization for our ultra-LDES technology.— Chris Graves, co-founder and CEO, Noon Energy
As we wrote in September, the contract is becoming the asset in energy storage. Noon's two deals carry that idea down to the technology layer: the system that wins may be the one that arrives with an off-take already attached.
Aric Saunders, Noon's executive vice president of commercialization, frames the demand the same way. Battery storage, he said, is what secures stable energy for data centers, and a 100-hour design is what makes the 24/7 load of the next AI campus reachable.
What has to go right
Three variables decide whether ultra-long-duration storage becomes an asset class or a footnote.
The first is the first build. A 25 MW / 2.5 GWh project is small next to the gigawatts promised, and it is not due until 2028. Until it runs, every number about unit cost remains a projection.
The second is efficiency. Reversible solid oxide systems trade some round-trip efficiency for duration, and Noon has not published a figure. If the losses are large enough, cheap tanks will not rescue the levelized cost of energy — the average cost per unit of power over a plant's lifetime.
The third is competition. Iron-air, sodium-ion, and conventional lithium systems are all improving, and they are chasing the same hyperscaler contracts. Duration is only one axis of the sale.
| Parameter | Noon ultra-LDES | Lithium-ion BESS | Flow battery |
|---|---|---|---|
| Typical duration | ✔ 100+ hours (claimed) | ◐ 2–4 hours | ◐ 8–12 hours |
| Footprint | ✔ 20–200× smaller than flow (claimed) | ◐ 2–3× larger than Noon (claimed) | ✗ Largest of the three |
| Storage medium | ✔ Carbon-based media | ✗ Lithium, nickel, cobalt | ✗ Vanadium or zinc |
| Track record at scale | ✗ Pilot only | ✔ Terawatt-hours shipped | ◐ Niche deployments |
Noon performance claims: energy-storage.news, 2026. Duration and material ranges are typical industry figures, not audited comparisons.
Noon says the system can deliver round-the-clock renewable power at "cents per kWh." That number will be settled by the first contracted project, not by the pilot. The company's footprint and material claims point to the same conclusion: the engineering case is plausible, the commercial case is unproven.
Does 100-hour storage reach gigawatt scale before 2030?
Probability: 55% — the off-take partners are credible and the first build is small enough to finance, but the reversible solid-oxide supply chain and undisclosed round-trip efficiency are untested at scale.
✅ Arguments for
Duration economics improve as tank size grows, unlike lithium systems that scale cells and power together.
The carbon-based medium sidesteps lithium, nickel, and cobalt supply bottlenecks.
Confirmation criteria: the 25 MW project reaches mechanical completion by 2028 and a second project above 100 MW is contracted by 2029.
❌ Arguments against
Iron-air and sodium-ion competitors are financed and moving toward the same hyperscaler contracts.
Both 2026 announcements are reservations; neither is an order.
Disconfirmation criteria: the first project slips past 2028, round-trip efficiency lands below 40%, or no follow-on contract is signed by 2029.
Mechanical completion date of the 25 MW / 2.5 GWh Meta project
Any disclosed round-trip efficiency figure for the reversible solid oxide block
A second off-take contract above 100 MW
Tank manufacturing cost per kWh-hour at volume
Development scenarios
🟢 Optimistic scenario (35%)
Implications: Ultra-LDES becomes a financeable asset class, and Noon moves from a Series A company to a strategic acquisition target for a utility or an OEM.
🟡 Base-case scenario (45%)
Implications: Noon stays private and well funded but does not break out; the storage market keeps its lithium backbone.
🔴 Pessimistic scenario (20%)
Implications: Noon becomes a licensing story or is absorbed for its patents, and ultra-LDES yields the niche to iron-air and sodium-ion.