The Future of Utility Management: Shifting from Reactive Engineering to AI-Driven Decision Making
On June 29, 2026, Delhi’s grid hit 8,748 MW, the highest peak demand in the city’s history. Most coverage stopped there. The grid held. Crisis averted.
What assumptions were sitting underneath that number, quietly not holding? Delhi’s record isn’t a story about one grid in one night. It’s a symptom of something structural: we are still measuring power reliability with frameworks built for a climate that no longer exists.
Resilience360 platform was designed using multi-science, Machine Learning (ML) and Artificial Intelligence (AI) to solve the gap between what metrics driven reliability and resilient reliability. What actually happens on the ground when extreme weather hits decides the course of 90-days action, 365 days planning and 3-5 year climate adaptation investment for Energy and Utility.
The assumption baked into the grid
The old model was simple: extreme weather is the exception. A storm hits, something breaks, crews respond, power returns, life resumes. That assumption shaped decades of grid design and investment.
It doesn’t hold anymore. India’s peak demand was 135 GW in 2013. By May 2026 it had crossed 270 GW, nearly double in thirteen years. Most of the transformers and lines carrying that load was specified almost a decade ago, against demand curves and climate patterns from a different era. Domestic electricity consumption rose 71% over roughly the same window. The load grew. The climate shifted. The hardware didn’t.
A pattern, not a series of accidents
Widen the lens and Delhi stops looking unusual. In the United States of America, black out days due to storm, hurricane, urban flood has increased to more than 40 days. Texas A&M University’s 2025 research found that weather-related outages in the United States of America have gotten 20% worse every year since 2019. In Iraq, temperatures near 50°C took down two transmission lines in August 2025, dropping 6,000 MW instantly and triggering a nationwide blackout. In Jamaica, Hurricane Melissa cut power to 530,000 people, some for over three weeks, not because restoration was slow, but because infrastructure had to be rebuilt, not switched back on. Spain and Portugal suffered Europe’s
largest outage since 2003, in minutes. In France’s Finistère region, a single heat-linked transformer failure knocked out power to 68,000 households.
This isn’t bad luck. The Union of Concerned Scientists found extreme weather behind every one of the ten worst US grid outages in the past decade. Research in Nature Scientific Reports shows heatwaves increasing outage frequency and duration, with outages projected to climb meaningfully by 2050 even under moderate warming. Demand keeps rising too: air-conditioning alone could add over 100 GW of peak demand in India within the decade.
A hardware crisis arriving at the same time
There’s a second story underneath the first, and it may be more dangerous because it’s slower to see. The grid itself is old, and the supply chain to replace it can’t keep up.
Forty percent of Europe’s distribution network has run for over 40 years. In the US, most transmission lines date to the 1950s and 60s, and more than half of distribution transformers are already past 33 years old, near or beyond their designed lifespan. Demand for replacements has surged, but lead times now stretch past two years, sometimes four, with prices rising sharply alongside them.
Put the two crises together: the infrastructure most exposed to a changing climate is also the infrastructure least able to be replaced quickly. When a transformer fails mid-heatwave, there’s no fast fix. The failure is immediate. The solution is not.
Why the metrics themselves are the problem
This is the part I think the industry hasn’t confronted. A utility can report strong system-wide reliability while specific communities absorb repeated outages every time temperatures spike. That’s not a data error, it’s what happens when performance is measured at the system level while failure happens at the asset level.
A city doesn’t lose power in silos. One non-operating transformer and a single damaged feeder causes loss of revenue for the distributor (operator). The system-wide average smooths all of that away and hides exactly what matters.
It also flattens outcomes that aren’t equivalent. A five-hour outage on a mild evening and a five-hour outage at 46°C look identical in a spreadsheet. They are not the same event. Then Europe’s June 2026 heatwave arrived, and researchers confirmed it would have been virtually impossible without human-driven climate change. Over 1,300 excess deaths followed. WHO reported 150 million people living under extreme heat simultaneously in June 2026, with deaths concentrated among the elderly,
the poor, and those without access to cooling, populations that rarely register distinctly in aggregate reliability reporting.
Resilience is a capital allocation problem
No utility can harden every asset at once. Resources are finite against an enormous exposure map. So, the useful question isn’t “where should we spend more.” It’s: which assets are likely to fail first, which failures carry the greatest consequences, which communities bear the most impact, and which intervention reduces the most risk per dollar spent.
Traditional reliability metrics were never built to answer that. They tell you how the system performed yesterday. Resilience asks how it will perform under stress tomorrow, before the stress arrives.
Delhi’s 2024 record lasted less than two years. I don’t expect June 2027’s record to last much longer either. Within a few summers, we’ll likely stop calling heatwaves like Europe’s “exceptional.” They’ll just be summer.
The utilities that come out ahead won’t be the ones that build the most or restore power fastest, though both still matter. They’ll be the ones that know, before the heat arrives, exactly where the system will break and where the next dollar does the best.
Reliability used to be an engineering problem: circuits, transformers, crews, trucks. It’s becoming a decision-making problem. The climate changed faster than our metrics and frameworks did. It’s time the frameworks caught up and metrics are resilience inclusive.
AUTHOR
