Himanshu Gupta on where AI helps climate — and where it's overkill
CTVC (Climate Tech VC) · August 2020
Read the full interview at CTVC (Climate Tech VC)OpenIn a CTVC 'In the Spotlight' interview, Himanshu Gupta lays out why he built ClimateAi around agriculture, the three arenas where AI moves the needle on climate, the autonomous-driving trick he borrows for sparse data, his hazard–exposure–vulnerability framework, his refusal to serve fossil fuels, and a candid line on where AI is overkill.
What he said
Why did he build ClimateAi around agriculture?
Because that's where the damage lands. Gupta points out that agriculture absorbs a huge share of climate-disaster costs while sustaining most of the world's population — a mismatch that makes it the place to start.
“Last year alone, 25% of global climate disasters costs were absorbed by agriculture – which 70%+ of the population depends on for their livelihoods.”Read the full interview
Where can AI actually make a difference on climate?
In three distinct arenas. Gupta organizes AI's climate potential into fundamental science, applied science, and metrics production — a way to separate hype from where the leverage really is.
“I see 3 arenas where AI can make a difference to climate challenges: 1) fundamental science, 2) applied science, and 3) metrics production.”Read the full interview
How do you model climate when the data is sparse?
Borrow the self-driving playbook. Gupta's approach is the one autonomous-vehicle teams use when road data runs out — generate what you're missing in simulation and train on that.
“We're taking an approach similar to the autonomous driving industry; if there isn't enough data to train your algorithms on the road, you train them on a simulation.”Read the full interview
How should a company think about its climate exposure?
Through three questions. Gupta reduces climate risk assessment to a simple framework — the hazard itself, the company's exposure to it, and where it's most vulnerable.
“(1) What is the hazard (e.g., drought, flood, or wildfire)? (2) What is the company's exposure? (3) Where is the vulnerability?”Read the full interview
Are there customers ClimateAi won't take?
Yes — a hard line. Gupta says the company's work will always be climate-positive, which rules out fossil fuels, and he's already turned oil-and-gas prospects away.
“Our actions will always be climate positive, so we'll never work with the fossil fuel industry. We've turned down a couple of customers from the oil and gas industry already.”Read the full interview
Where is AI overhyped in climate?
When it's used for its own sake. Gupta is candid that the fanciest model isn't always the right one — sometimes a simple regression delivers the value a neural network is credited with.
“AI can be overkill. A simple linear regression can create enough customer value, rather than applying complex neural networks to the problem.”Read the full interview
Key takeaways
- Agriculture absorbs an outsized share of climate-disaster costs — so that's where ClimateAi starts.
- AI helps climate across three arenas: fundamental science, applied science, and metrics.
- When road data runs out, train on simulation — the autonomous-driving playbook.
- Climate risk in three questions: hazard, exposure, vulnerability.
- Climate-positive by mandate: no fossil-fuel customers.
- AI can be overkill — sometimes a linear regression is enough.