Someone Finally Counted Every Solar Panel on Earth

Here is a strange fact about the energy transition: until very recently, nobody actually knew how many solar panels existed on Earth. Governments tracked utility-scale farms. Utilities counted grid-connected rooftops. But the full picture — every panel on every warehouse, every off-grid installation in rural India, every floating array on a Chinese reservoir — was a blank. This dataset fills that blank. Using machine learning trained on satellite imagery, researchers identified and geolocated 68 million solar photovoltaic generating units across the globe, creating the first inventory that doesn't depend on self-reporting or administrative records alone.

The findings upended several assumptions. China's installed solar capacity was 15% larger than official statistics suggested, with thousands of small-scale installations escaping provincial reporting. India's rooftop solar sector was growing faster than any government projection had anticipated. And across sub-Saharan Africa, off-grid solar was proliferating in patterns that bore no resemblance to national energy plans — following mobile phone tower networks rather than electrical grid expansion routes. The data revealed an energy transition that was messier, faster, and more bottom-up than top-down planning frameworks assumed.

What makes this inventory particularly potent is its granularity. Each entry includes not just a location but an estimated generating capacity, enabling researchers to build capacity maps that update our understanding of global solar infrastructure. With 16,000 downloads and counting, the dataset has already been incorporated into energy models, climate projections, and policy analyses at the International Energy Agency, the World Bank, and dozens of national energy ministries. It is, in effect, the first honest photograph of the solar revolution.