What is the story about?
Artificial intelligence (AI) should be leveraged to improve how governments design large welfare and infrastructure programs, especially those that now operate at a national scale. The Pradhan Mantri Ujjwala Yojana (PMUY) is a case in point. Since its launch in 2016, PMUY has provided more than 100 million LPG connections, significantly expanding access to clean cooking among low-income households. Evaluations by Indian and international agencies show that this expansion has reduced reliance on solid fuels and improved convenience and time use, particularly for women.
When a clean-cooking program reaches this scale, the policy question shifts from access to precision. As PMUY matures, the task is to sustain these welfare gains while gradually reducing marginal dependence on imported LPG and the fiscal pressures associated with global price volatility. India currently imports around 20-21 million tons of LPG each year, accounting for roughly 60-65% of total supply, with household cooking forming the largest share of demand. This reflects broader structural factors such as income growth and urbanization, but it also links clean-cooking outcomes more closely to external energy markets.
Electric cooking, deployed in a targeted manner, offers a practical pathway to reduce marginal dependence on LPG over time. AI enables targeting by identifying where electric cooking can be adopted sustainably and at the lowest system cost. Even a modest shift—reducing LPG demand by 0.5 to 1 million tons a year could avoid roughly $300–600 million in annual LPG imports, while easing longer-term subsidy exposure and household fuel costs.
Targeting where electric cooking fits
The policy task is to identify where electric cooking can be deployed reliably and at low system cost. Electricity supply quality and distribution capacity vary significantly across regions, and distribution companies (discoms) operate under tight financial constraints. Careful sequencing is therefore essential.
AI can support this by combining datasets that already exist across government systems. Discoms routinely collect information on feeder outages, transformer loading, and peak demand. Oil marketing companies maintain granular administrative data on PMUY connections and refill patterns through the Direct Benefit Transfer-LPG system, which provides insights into usage behavior. Administrative datasets across power and social-sector departments provide indicators of household electrification and vulnerability. Analyzed together, these inputs can identify areas where electric cooking can be introduced immediately, areas that would benefit from modest strengthening, and areas where it should be deferred.
This analytical work can be undertaken centrally. A small unit anchored in the Ministry of Power, working with the Ministry of Petroleum and Natural Gas and the Central Electricity Authority, can generate periodic readiness assessments and phased shortlists. Discoms can then validate local conditions and focus their operational effort where adoption is most likely to succeed.
Strengthening adoption and financing
Evidence from India and other countries shows that access to clean cooking does not automatically translate into sustained or exclusive use. Household fuel choices continue to reflect affordability, reliability, and cooking practices. This makes program design as important as program coverage.
AI can strengthen this dimension in two ways. First, it can support household-level differentiation within suitable areas. Existing data, such as refill frequency, household size, location, and electricity quality, can help distinguish households likely to adopt electric cooking for routine meals from those likely to continue fuel stacking. This allows public support to be calibrated rather than uniform, improving the effectiveness of spending.
Second, AI can enable market-based financing mechanisms that reduce reliance on new subsidies. One constraint on the adoption of electric cooking has been uncertainty among manufacturers and financiers about demand and repayment risk. AI-enabled demand forecasting at the district level can reduce this uncertainty, allowing suppliers to plan production and financiers to offer pay-as-you-save or micro-EMI products linked to expected fuel savings. India has seen the impact of such aggregation before. Under the UJALA program, assured demand contributed to an 80–85 percent reduction in LED bulb prices over time. Electric cooking does not need to replicate this scale for the economics to improve materially; even 20-30 percent cost reductions would significantly lower adoption barriers.
This approach allows existing public support to work harder. LPG continues to provide a safety net under PMUY, while incremental support can be directed toward electric cooking to reinforce sustained use and reduce LPG consumption over time.
Managing the transition on the ground
Local grid conditions will shape how electric cooking is rolled out. For distribution companies, cooking demand has attractive characteristics: it is daily, predictable, and less seasonal than many other loads. When introduced in a planned manner, it can strengthen revenues and improve load predictability without triggering disproportionate network upgrades. AI-supported analysis can help utilities anticipate cooking-related demand, identify where networks can absorb it, and sequence deployment accordingly. Central analysis can guide where to begin, while operational decisions remain with local utilities.
Adoption also depends on whether electric cooking fits everyday household needs. In practice, this means designing programs around real kitchens and meal patterns, often combining electric pressure cookers for staple dishes with induction cooktops for quick tasks, while retaining LPG for flexibility. AI can help tailor appliance bundles and early-stage support to regional cooking practices and household size, improving uptake. Practical guidance on cookware, safety, and everyday use can be delivered through existing oil-marketing distributor networks and women’s self-help groups using low-cost, local-language tools.
PMUY has demonstrated that India can deliver clean cooking at scale. The next phase is about delivering it with greater precision. AI can help sustain welfare gains while reducing incremental dependence on LPG. The lesson of the past decade is that scale matters; the lesson of the next may be that precision matters just as much.
(The author is Director of Policy at the India Energy and Climate Center at UC Berkeley and an adjunct fellow (non-resident) with the Chair on India and Emerging Asia Economics at the Center for Strategic and International Studies.)
When a clean-cooking program reaches this scale, the policy question shifts from access to precision. As PMUY matures, the task is to sustain these welfare gains while gradually reducing marginal dependence on imported LPG and the fiscal pressures associated with global price volatility. India currently imports around 20-21 million tons of LPG each year, accounting for roughly 60-65% of total supply, with household cooking forming the largest share of demand. This reflects broader structural factors such as income growth and urbanization, but it also links clean-cooking outcomes more closely to external energy markets.
Electric cooking, deployed in a targeted manner, offers a practical pathway to reduce marginal dependence on LPG over time. AI enables targeting by identifying where electric cooking can be adopted sustainably and at the lowest system cost. Even a modest shift—reducing LPG demand by 0.5 to 1 million tons a year could avoid roughly $300–600 million in annual LPG imports, while easing longer-term subsidy exposure and household fuel costs.
Targeting where electric cooking fits
The policy task is to identify where electric cooking can be deployed reliably and at low system cost. Electricity supply quality and distribution capacity vary significantly across regions, and distribution companies (discoms) operate under tight financial constraints. Careful sequencing is therefore essential.
AI can support this by combining datasets that already exist across government systems. Discoms routinely collect information on feeder outages, transformer loading, and peak demand. Oil marketing companies maintain granular administrative data on PMUY connections and refill patterns through the Direct Benefit Transfer-LPG system, which provides insights into usage behavior. Administrative datasets across power and social-sector departments provide indicators of household electrification and vulnerability. Analyzed together, these inputs can identify areas where electric cooking can be introduced immediately, areas that would benefit from modest strengthening, and areas where it should be deferred.
This analytical work can be undertaken centrally. A small unit anchored in the Ministry of Power, working with the Ministry of Petroleum and Natural Gas and the Central Electricity Authority, can generate periodic readiness assessments and phased shortlists. Discoms can then validate local conditions and focus their operational effort where adoption is most likely to succeed.
Strengthening adoption and financing
Evidence from India and other countries shows that access to clean cooking does not automatically translate into sustained or exclusive use. Household fuel choices continue to reflect affordability, reliability, and cooking practices. This makes program design as important as program coverage.
AI can strengthen this dimension in two ways. First, it can support household-level differentiation within suitable areas. Existing data, such as refill frequency, household size, location, and electricity quality, can help distinguish households likely to adopt electric cooking for routine meals from those likely to continue fuel stacking. This allows public support to be calibrated rather than uniform, improving the effectiveness of spending.
Second, AI can enable market-based financing mechanisms that reduce reliance on new subsidies. One constraint on the adoption of electric cooking has been uncertainty among manufacturers and financiers about demand and repayment risk. AI-enabled demand forecasting at the district level can reduce this uncertainty, allowing suppliers to plan production and financiers to offer pay-as-you-save or micro-EMI products linked to expected fuel savings. India has seen the impact of such aggregation before. Under the UJALA program, assured demand contributed to an 80–85 percent reduction in LED bulb prices over time. Electric cooking does not need to replicate this scale for the economics to improve materially; even 20-30 percent cost reductions would significantly lower adoption barriers.
This approach allows existing public support to work harder. LPG continues to provide a safety net under PMUY, while incremental support can be directed toward electric cooking to reinforce sustained use and reduce LPG consumption over time.
Managing the transition on the ground
Local grid conditions will shape how electric cooking is rolled out. For distribution companies, cooking demand has attractive characteristics: it is daily, predictable, and less seasonal than many other loads. When introduced in a planned manner, it can strengthen revenues and improve load predictability without triggering disproportionate network upgrades. AI-supported analysis can help utilities anticipate cooking-related demand, identify where networks can absorb it, and sequence deployment accordingly. Central analysis can guide where to begin, while operational decisions remain with local utilities.
Adoption also depends on whether electric cooking fits everyday household needs. In practice, this means designing programs around real kitchens and meal patterns, often combining electric pressure cookers for staple dishes with induction cooktops for quick tasks, while retaining LPG for flexibility. AI can help tailor appliance bundles and early-stage support to regional cooking practices and household size, improving uptake. Practical guidance on cookware, safety, and everyday use can be delivered through existing oil-marketing distributor networks and women’s self-help groups using low-cost, local-language tools.
PMUY has demonstrated that India can deliver clean cooking at scale. The next phase is about delivering it with greater precision. AI can help sustain welfare gains while reducing incremental dependence on LPG. The lesson of the past decade is that scale matters; the lesson of the next may be that precision matters just as much.
(The author is Director of Policy at the India Energy and Climate Center at UC Berkeley and an adjunct fellow (non-resident) with the Chair on India and Emerging Asia Economics at the Center for Strategic and International Studies.)
