The Cleanest Air in India Costs the Most Electricity
India's largest generic pharmaceutical manufacturers now procure seventy to eighty-five per cent of their electricity from renewables — a figure that flatters the underlying problem. Pharmaceutical cleanrooms account for fifty to eighty per cent of a sterile plant's total energy and run continuously at setpoints regulators inspect without warning; changing the grid mix does not move the meter. Model predictive control of chilled-water loops is where a ten-to-twenty per cent reduction is now measurable.
On page after page of Dr. Reddy's FY2026 Form 6-K filing with the SEC, the same line keeps returning beside plant-by-plant capacity data: renewable share of power procured, 78% for FY26, up from 68% the year before. Target: 100% by 2030. Absolute Scope 1 and 2 emissions down 33% against an FY23 baseline. Dr. Reddy's is now the only Indian pharmaceutical company with an SBTi-validated Net Zero commitment out to FY2045. These are numbers the company had to defend to a validator, not vendor copy.
They are also the easy part.
Cipla's FY 2025-26 sustainability disclosures read the same way when you stop skimming. About 55 megawatts-peak of captive solar under open access, 2.7 megavolt-amperes of captive wind, 8.1 megawatts-peak on rooftops across sites, and a 30-megawatt plant on 115 acres in Maharashtra dedicated to two manufacturing units at Kurkumbh and Patalganga. A carbon-neutral India operations target set for December 2025. A 30% Scope 2 reduction by FY30 versus a 2020 baseline. All of it laid out in the filing that SEBI now requires the top 1,000 listed companies to publish.
None of this is where the interesting fight is happening. The interesting fight is inside a chilled water loop.
Fifty to eighty percent
Pharmaceutical cleanrooms are the most energy-hungry cubic metres of air in industrial India. The International Society for Pharmaceutical Engineering puts HVAC at fifty to eighty percent of total energy consumption in a sterile manufacturing facility, with air-change rates in Grade A and B environments running an order of magnitude above commercial buildings.
For an Indian generics plant producing under EU-GMP and FDA scrutiny, which is what Kurkumbh, Baddi, Vizag, Duvvada, and Pithampur all are, the HVAC plant runs 24 hours a day, 365 days a year, holding temperature bands and humidity setpoints that inspectors sample without warning. Turn it off, or let it drift, and the batch is at risk. So the chiller runs. The air handlers run. The meter runs.
Which is why the 78% renewable number and the 30-megawatt captive solar plant do a specific thing: they change what colour the electrons are, not how many electrons the plant draws. The demand curve barely moves. The Scope 2 number falls because the grid emission factor changes. The chiller does not care.
What AI is doing here
The interesting move sits on the demand side. A study published last year in Energy and Buildings tested three model predictive control strategies against the constraint that a pharmaceutical HVAC system cannot violate temperature or humidity setpoints under any load condition: predictive functional control, nonlinear MPC with particle-swarm optimisation, and an energy-efficient variant. The energy-efficient controller returned up to a 20% reduction in total energy consumption while holding the setpoints.
Take that number with the usual laboratory-versus-plant discount. A real cleanroom has fouling, sensor drift, valve wear, filter loading, and an operator who overrides the schedule when a batch behaves oddly. Fifteen percent is a more honest expectation at scale, and it is still a very good number for a system that already accounts for most of the plant's electric load.
Johnson Controls' April 2026 acquisition of Nantum AI is the market signal. Nantum runs a real-time optimisation platform for building HVAC and distributed energy resources, and it is being folded into the OpenBlue stack. When the incumbent BMS vendor writes a cheque, the pattern has crossed from pilot to procurement.
Vendor benchmarks still need reading in context. When the incumbents quote "twenty to thirty percent HVAC energy reduction," they mean it against an unrenovated baseline in an American commercial office, with weather, occupancy, and lighting all changing at once. A pharma cleanroom does not have those degrees of freedom. Occupancy is fixed. Setpoints are fixed. Air-change rates are fixed by classification. The only levers are chiller staging, chilled-water supply temperature, VFD speeds on the AHU fans, and demand-side shifting into the wind and solar hours the captive plant is actually generating. Those levers exist. They are worth ten to twenty percent of the electric bill in a well-instrumented site. They are worth far less in a site that has neither the sensors nor the historians to close the loop.
The three majors, read together
| Company | Renewable share of power | Scope 1 & 2 target | Public artifact |
|---|---|---|---|
| Dr. Reddy's | 78% (FY26) | 80% cut vs FY23 by FY30; SBTi validated | SEC Form 6-K FY2026 |
| Cipla | ~65 MW captive renewables + rooftop | Carbon-neutral India ops target Dec 2025 | Cipla sustainability disclosures |
| Serum Institute | ~85% of energy from wind (57.2 MW) | Steam from bagasse briquettes, not fuel oil | Green initiatives page |
Three different plays on the same board. Dr. Reddy's is running the SBTi playbook end-to-end, targeting coal-free and furnace-oil-free operations by the close of FY26 and validator-approved absolute cuts by FY30. Cipla is buying captive renewable capacity at open-access scale and disclosing under SEBI's BRSR framework. Serum has been on wind well before the ESG regime demanded it, and its steam plant runs on bagasse briquettes rather than heavy oil. Each of these choices carries different Scope 1 and Scope 2 consequences. None of them touches Scope 3.
Where the reporting stops
Which is where the harder story lives. Analysis published by edie puts Scope 3 at eighty to ninety percent of the pharmaceutical sector's total climate impact, driven overwhelmingly by raw-material extraction, active pharmaceutical ingredient synthesis, and the upstream chemistry that feeds the finished-dose plant. As of the last public count that piece cites, fifteen pharmaceutical companies globally were reporting Scope 3 transparently; a handful had validation from the Science Based Targets initiative. The World Economic Forum's Scope 3 playbook for India is blunt about why: supplier data doesn't exist yet, reporting standards don't line up, incentives to invest in measurement are weak, and the buyers keep buying.
Dr. Reddy's commits to a 51.6% cut in Scope 3 emissions per rupee of value added by FY30. That per-INR-value-added denominator matters. It is compatible with absolute Scope 3 rising as long as revenue rises faster. The commitment is SBTi-validated and real, but it is intensity-based, and no other major Indian pharma has yet signed the equivalent. Cipla's disclosures on Scope 3 remain thinner than its disclosures on Scopes 1 and 2. That is not unique to India; it is the state of the industry.
The ARC Advisory Group's read on industrial AI in Indian pharma makes the related point in a different register: the generics business is entering a phase where cost discipline, regulatory intensity, and product complexity all rise together, and industrial AI is the only realistic way to hold unit economics stable. The chiller-loop optimisation piece of that puzzle is real and worth doing today. The bigger piece, traceability of API inputs, a digital thread from supplier plant to finished dose, actual per-lot Scope 3 measurement, has barely started in India.
In Kurkumbh, at night
At about four in the morning at a Kurkumbh formulation plant, the chilled water loop is at its lowest ambient load of the day. The Maharashtra grid is drawing from a wind-heavy dispatch. The captive solar plant next door is dark, so the wind PPA is doing the work. Inside the plant, the Grade B cleanrooms hold 20°C and 45% relative humidity, particulate counts within limits, differential pressures where the SOP says they should be. A shift engineer looks at a screen showing chiller coefficient of performance, AHU fan speed, chilled-water supply temperature, and the meter total for the last hour. The number is six percent lower than the same hour a week ago, on similar ambient conditions. The reason is a controller that quietly reset the chilled-water supply temperature by half a degree, load-shifted a return-air fan across the wind hours, and staged the second chiller off until the outside air rises.
Nobody in the room is thinking about the BRSR. They are thinking about the meter.
Tarry Singh is the founder and CEO of Real AI, an enterprise AI advisory and deployment firm working with global enterprises on production agent systems, model risk, and AI sovereignty strategy. He also leads Earthscan for Energy AI, and is a founding contributor to the EU-funded HCAIM and PANORAIMA programmes for responsible AI education across European universities. He writes at tarrysingh.com.