The Story
Four Indian organisations have been selected for the inaugural Google DeepMind Accelerator: AI for the Planet, covering Asia-Pacific. Terrastack, Varaha Climate, Farmers for Forests and Climitra Carbon join a cohort of 16 startups, nonprofits and research groups from across the region.
Participants receive access to Google's AI stack, including specialised frontier models, along with technical support and mentorship over three months. The programme opened with a week-long bootcamp in Singapore from 7 to 11 September and concludes with a Demo Day in December.
The four Indian entrants apply artificial intelligence, satellite intelligence, remote sensing and drones across agriculture, agroforestry, carbon removal and biodiversity.
Terrastack is building a geospatial data platform for rural land intelligence and transactions, combining satellite and agronomic data to deliver plot-level insight for smallholder farmers. It sits in the programme's sustainable agriculture track.
Varaha Climate enables smallholder farmers to earn carbon credits by verifying regenerative agriculture and carbon removal through remote sensing and AI.
Farmers for Forests uses a payment for ecosystem services model to incentivise farmers and communities to convert degraded land into biodiverse forests, across agroforestry, forest protection and restoration, urban biodiversity, human-elephant conflict prevention and climate education.
Climitra Carbon converts waste biomass into industrial-grade bio-coal and biochar, with its bio-coal positioned as a one-for-one replacement for fossil coal in steelmaking. It is deploying Geo-AI to verify the removal of invasive species and their conversion into biochar.
Other cohort members include New Zealand's 800 Trust, Listening Lab and Wildlife.ai, Singapore's Kumi Analytics, South Korea's TelePIX and Indonesia's Yayasan Ekosistem Lestari. Google announced four further initiatives deploying AI, geospatial data and infrastructure investment towards India's clean energy transition alongside the cohort.
Why It Matters
Accelerator cohorts are announced constantly and most say little. What makes this list readable is that all four Indian entrants are solving versions of the same problem, and it is a measurement problem rather than a climate one.
Terrastack combines satellite and agronomic data into plot-level land intelligence. Varaha verifies regenerative agriculture and carbon removal through remote sensing. Climitra applies Geo-AI to confirm that invasive species were removed and converted into biochar. Farmers for Forests pays farmers to restore degraded land, which requires establishing that the restoration happened.
None of these companies is primarily generating energy or capturing carbon. They are establishing, at scale and at low cost, that something took place on a piece of land. That is the bottleneck in climate finance, and it explains why frontier model access is worth three months of a founder's attention.
It also explains the selection. Google is not offering capital here, and the programme is equity-free. What it offers is compute and models, which matters specifically to companies whose core cost is processing satellite imagery across millions of small plots. For a carbon verification business, inference cost is close to cost of goods sold.
The fit is genuine rather than decorative, which is more than can usually be said for a cohort announcement.
The Strategic Read
The business model underneath three of these four is worth understanding, because it is unusual and fragile in the same breath.
Varaha, Farmers for Forests and Climitra all convert environmental outcomes into saleable instruments. Carbon credits, ecosystem service payments, verified removals. The customer is not the farmer whose land is restored. It is a company somewhere else buying an offset, and the product being sold is a claim that something measurable happened on a specific piece of ground.
That makes verification the entire business. A carbon credit is worth what a buyer believes it represents, and voluntary carbon markets have spent several years being told by journalists and auditors that a meaningful share of credits represented less than claimed. Prices fell accordingly.
Which is precisely why satellite intelligence and remote sensing keep appearing in these descriptions. If you can demonstrate from orbit that a field was farmed a particular way, or that a stand of invasive species was actually cleared, the credit becomes harder to dispute. The AI is not there to grow trees. It is there to make the resulting claim auditable at a cost per hectare that allows smallholder land to enter the market at all.
That last point is the Indian specificity. Indian agricultural holdings are small and fragmented, which has historically kept them out of carbon markets, because manual verification costs more than the credits are worth. Remote sensing changes the unit economics of verification, and only then does the model work here.
The caution is the same one that applies to any accelerator. Three months of compute access and mentorship is a real advantage and it is not capital. The programme is equity-free, and what these four still need is a buyer willing to pay a price that makes restoration worth doing.
For daily, sharp analysis of the biggest moves in the Indian business and startup ecosystem, follow StartupFox.