Currently in many African, Asian and South American countries, weather reports may be available only every six to 12 hours and only for broad patches of land up to 15km (9.3 miles) wide.
But IBM's new tool provides reports down to more specific, 3km-wide areas.
The company says it can even predict individual thunderstorms.
The tool, announced at the CES tech show in Las Vegas and launched in partnership with The Weather Channel, uses supercomputers to crunch data from hundreds of millions of sensors around the globe.
"The scale is almost incomprehensible to people - from a compute and complexity point of view," Cameron Clayton, at IBM, told BBC News.
"A farmer in Kansas has really good weather [forecasts] today but a farmer in Kenya only gets a weather forecast once, maybe twice, a day - they'll now get it hourly."
Mr Clayton said the Global High-Resolution Atmospheric Forecasting System (Graf) had been designed to gather data from a wide variety of sensors - including millions of smartphones equipped with atmospheric pressure sensors.