Dynamo Bangalore is building a membrane-driven indirect evaporative cooling system that uses 75% less electricity than vapour-compression AC — and produces clean water as a byproduct of cooling, instead of consuming it. Designed for India's data centres, large buildings, and any facility that needs cheap cooling without draining the city.
Cool with water — and drain a water-stressed city. Or cool with electricity — and burn through the grid. Every operator is forced to pick a side. There is no zero-tradeoff option in the market today.
of water per kWh of compute
A 100 MW hyperscale facility drinks up to 1.5 million litres a day. Bengaluru, Hyderabad and Chennai are already under tanker rationing. Hyperscalers face hard water caps from 2027 onwards.
of total DC electricity goes to cooling
Indian DCs run a PUE of 1.6–1.8 versus 1.1 best-in-class. Mechanical chillers degrade fast at 40°C+ ambient. R-32 and R-410A still dominate, with mounting regulatory pressure.
Conventional evaporative cooling evaporates water into the atmosphere. We invert it — we extract water vapour from incoming air using a vacuum membrane, then use the dried air to cool a closed water loop. The membrane stage produces more condensate than the cooler consumes. Net output: chilled water for the facility, plus clean distilled water as a byproduct.
No compressor. No refrigerant cycle. Just fans, a small vacuum pump, and a water circulator. Total draw of 300W for 0.5 ton of cooling. COP of 5.83 versus ~3.0 for a conventional split AC.
Every kWh of compute cooled generates roughly 0.5–1.5 litres of clean water output in humid climates, rather than consuming it. At scale: a 1 MW data centre in 100% outside-air mode in a humid Indian city would produce tens of thousands of litres per day. Actual MW-scale figures depend on outside-air fraction and ambient humidity profile.
Designed keeping in mind Bengaluru pre-monsoon, Delhi peak summer (45°C), and full monsoon (90% RH). Most evaporative coolers fail in humidity. We thrive in it.
The Indian data centre market is doubling every three years, into the same cities running out of water. Hyperscalers face board-level water mandates: Microsoft, Google, Meta and AWS have all committed to water-positive operations by 2030. Indian operators are following. We are building the only cooling technology that lets a DC be water-positive by design.
| Approach | PUE | WUE (L/kWh) | Refrigerant |
|---|---|---|---|
| Vapour-compression chiller | 1.5–1.8 | 0 | R-32 / R-410A |
| Adiabatic / evaporative | 1.1–1.2 | 1.5–3.0 | None |
| Direct liquid / immersion | 1.05–1.10 | Near zero | Engineered fluids |
| Dynamo IEC-Model A | 1.10–1.20 | 0 to −3 | None |
Liquid / immersion comes close on PUE and water — but at 5–10× the capex, requires engineered coolant fluids, and only addresses chip-level cooling. We address whole-facility cooling with conventional refrigerant-free physics.
The current 0.5-ton prototype is the first build. It validates the physics, the architecture, and the supply chain. Future scaled units share the same five-stage architecture — only the dimensions change.
Hot humid outdoor air enters at 34–45°C. Coarse dust filter removes particulate.
Air passes a hydrophilic membrane under vacuum. Water vapour crosses to the low-pressure side; dry air continues.
Dry air enters the first indirect evaporative cooler. Wet pads cool the air substantially.
Re-dried air enters the second IEC, which also houses the chilled water coil for the indoor loop.
14°C chilled water to the facility. Distilled water byproduct drains to building sump.
Each step validates the physics at the next order of magnitude. The membrane and IEC stages are linearly modular — the scaling risk is integration and controls, not core thermodynamics.
1.75 kW cooling. Lab validation. Physics proof. Engineering brief complete, BOM costed at ₹2.18 lakh (parts). Two design-and-build firms quoting full turnkey delivery at ₹25–50 lakh in 4–6 months.
17.5 kW cooling. First commercial pilot — Bangalore office building or small data hall. ₹1.2 Cr R&D phase. Customer-funded.
175 kW cooling. First data centre pilot — edge facility or Tier 2 colocation. Series A territory. Operational validation in DC environment.
Modular array, retrofit-friendly. Hyperscale colocation and AI inference cluster cooling. Series B and growth-stage.
Most evaporative coolers fail in humidity. The IEC-Model A was designed from the start for three distinct Indian conditions — pre-monsoon heat, Delhi's dry extreme, and full monsoon. All three deliver supply air at 14–15°C.
A split AC uses a compressor to pump refrigerant at high pressure — that's the dominant electricity draw. The IEC-Model A has no compressor. It cools by evaporating water (a physics process that costs almost nothing) and removing humidity with a small vacuum pump.
300W total — fans, pumps, vacuum.The only electrical loads are a main fan, a vacuum pump, two water circulation pumps, the chilled water pump, and an indoor fan. Combined: 300W for a 0.5-ton unit, versus 1,200W for a comparable split AC.
Our membrane stage extracts water vapour directly from incoming humid air using a vacuum differential. This vapour is then condensed back to liquid water in a downstream condenser. The IEC stages do consume some of this recovered water through evaporation, but in most climates we extract more than we consume — leaving a net surplus of clean, distilled-quality water.
For a data centre, this is the entire pitch.A 1 MW DC running the IEC-Model A architecture in 100% outside-air mode in a humid city would produce tens of thousands of litres of clean water per day, rather than consuming 1.5 million. Real numbers depend on outside-air fraction and ambient humidity — but the principle of net-positive water from cooling is what changes the conversation with operators.
The membrane dehumidification step works on pressure differential. The more humid the outdoor air, the bigger the driving force. At 90% RH monsoon conditions, our calculations show 14.2°C supply air — essentially identical to pre-monsoon performance. Conventional swamp coolers fail above 60% RH. We improve.
Most water produced in monsoon.Recovered water reaches ~130 L/day per 0.5T unit during monsoon — the highest output of any season, because there's more moisture in the air to extract.
The five-stage architecture is linearly modular. A 1 MW unit is not 2,000× more complex than a 0.5T unit — it has 2,000× more membrane area, 2,000× more pad surface, and proportionally larger fans and pumps. The thermodynamics, control logic, and engineering principles are identical. The scaling risk is in integration, mechanical packaging, and field reliability — not in the core physics.
Each step validates the next.0.5T → 5T → 50T → 1MW+ is a structured 4-year roadmap with operational pilots at every order of magnitude.
At 45°C / 20% RH (Delhi peak summer), the IEC-Model A delivers 15.2°C supply air and 6,144 W of cooling output — the strongest performance of any scenario we've tested. The large temperature differential between ambient and wet-bulb gives the evaporative stages enormous cooling headroom. Mechanical chillers degrade as ambient rises; we accelerate.
We are raising a ₹1 Cr seed round to fund the prototype build and a small engineering team for 12 months, getting to Series A readiness.
We are raising a seed round and are open to early pilot conversations. If you operate a data centre or large commercial building in India, are an HVAC OEM, a climate-tech investor, or a hyperscale operator looking at your water exposure — please get in touch.
We are raising a ₹1 Cr seed round to fund the 0.5T prototype build and a small engineering team, with a clear 12-month path to a Series A. Deck available on request.
info@dynamobangalore.comInterested in being our first DC pilot site? We are looking for one operational data hall in India for joint design and validation of the 50T module.
info@dynamobangalore.com