The Tech Blog

There has been a lot of hue and cry around AI data centers. I think a good part of it is valid, because big ones can stress the local power grid and water systems. The energy side is harder to solve as every calculation needs electricity. The water side is more interesting, because water is often used to move heat. Let’s discuss the latest tech trying to reduce that.

From the basics, almost every watt entering a server becomes heat. A 100kW AI rack is close to a 100kW heater. If you remove heat using air, you need airflow, fans, cold aisles and chilled water. Traditional cooling often uses an open or semi-open evaporative path. Water takes heat from the chiller loop and some of it is intentionally evaporated in a cooling tower. This works because evaporation carries a lot of heat. Water demand can peak on hot days when cities and farms also need water.

In a closed-loop system, a water-glycol coolant flows through cold plates on the chips, picks up heat, goes to a coolant distribution unit, then to a dry cooler that works like a radiator. The same coolant keeps circulating. NVIDIA’s reference talks about 45°C coolant going in and roughly 55°C coming out. That higher temperature loop matters because warmer coolant creates a larger temperature difference with ambient air, improving heat transfer in dry coolers and reducing the need for water-intensive cooling towers.

Microsoft, Google and Meta have been pushing liquid cooling and higher temperature operation for the same reason. Recent lifecycle work suggests advanced cooling can cut total water use by roughly 31-52% versus traditional air cooling. An NVIDIA white paper says cooling-tower-based systems can use about 10 Million litres per MW-year, while warm liquid loops with dry coolers can bring that close to zero in good climates. So good from the local nature PoV.

I think this should be considered by every new data center being planned now. A site already built around air cooling or evaporative towers may carry that architecture for years, but new builds still have the chance to choose a lower-water heat path from day one. I hope India follows the newer tech.

There has been a lot of discussion around E20 fuel being rolled out across India. One side says it is cleaner and better for emissions. Another side says it hurts mileage. I wanted to look at the cleaner-fuel claim from a science point of view just for my understanding as there is lot of misinformation around.

The useful question is simple. Cleaner in what sense?

E20 can be cleaner if you are talking about some local pollutants from the exhaust. It looks less dramatic if you are talking about tailpipe CO₂ per km. It can look better again if you use lifecycle carbon accounting, where farming, fertilizer, irrigation, distillation, transport, and fuel production are included.

E20 means petrol blended with 20% ethanol by volume. Ethanol still burns to produce CO₂ and water. It has less carbon per litre than petrol, but it also has less energy per litre. A good rule of thumb is that ethanol carries about 65% of petrol’s energy per litre. So E20 is 20% ethanol by volume, while ethanol contributes only about 14% of the fuel energy.

This is where many claims become confusing. If a certain ethanol pathway has 50% lower lifecycle emissions than petrol, the full E20 blend does not become 50% lower carbon. Only a fraction of the fuel energy is coming from ethanol, so the blend-level climate benefit becomes much smaller. For tailpipe CO₂ per km, the saving is also limited because lower carbon per litre is partly offset by lower energy per litre. Mileage drops.

The stronger case for E20 is local pollution. Ethanol has oxygen in the molecule, so it can help reduce carbon monoxide, soot & particulate matter. The tradeoff is that acetaldehyde can increase, especially during cold start, and NOₓ depends on engine design and catalyst behaviour.

Based on all of this, it is surprisingly easy to make E20 look either extremely good or completely pointless depending on which metric you choose to highlight. So before accepting any claim about E20 being cleaner or dirtier, the first question should be: cleaner in which unit, and measured across which boundary?

Once you ask that question, most of the confusion around E20 starts to disappear and you can take an informed decision.