Nvidia Just Bottled the Climate

Nvidia Just Bottled the Climate

We finally have a digital twin of Earth with five-kilometer precision, and our first instinct is to help insurance companies raise premiums faster.

Nvidia’s new climate AI, dubbed cBottle (short for “Climate in a Bottle”), compresses 50 years of global climate data 3,000-fold to simulate the planet’s future, at five-kilometer resolution. It’s a cornerstone of Nvidia’s Earth-2 platform, now adopted by research powerhouses like the Alan Turing Institute and the Max Planck Institute. The goal? To shift from blurry forecasts to decision-ready models.

cBottle builds on generative AI, running complex climate simulations in minutes instead of hours. What once cost $3M to simulate now runs for $60K. That’s not just speed, that’s scale. Companies like Spire Global already use Earth-2 to enhance forecasting speed and reduce costs by orders of magnitude. It’s not perfect, just probabilities, not certainties, but it could reshape everything from supply chains to geopolitics.

The dilemma is not whether this works, it’s how it’s used. Will governments prepare for floods, or weaponize Arctic sea lane data? Will farmers benefit, or financial markets bet on crop failures?

The tech is no longer the bottleneck. Our foresight, and our ethics—are. Precision prediction doesn’t guarantee wise action. If you had Earth’s future at five-kilometer clarity, who would you trust to use it?

Read the full article on Wall Street Journal.

----