The Underground Climate Machine

We spend a lot of time looking for new solutions to climate change. Sometimes, though, nature gets there first.

Plants have been pulling carbon dioxide out of the atmosphere for hundreds of millions of years. They take CO₂ from the air, turn it into leaves, stems and roots, and send some of that carbon underground. There, roots interact with vast communities of microbes and fungi, moving carbon into the soil where some of it can remain long after the plant itself has disappeared. It is an extraordinary piece of natural engineering.

The question is: can we make plants store even more carbon underground than they already do? That is the idea behind a new $18 million grant from the Bezos Earth Fund to the Salk Institute.

How we got here 

Climate change is often discussed in terms of what happens in the air around us. But some of the most promising opportunities may lie beneath our feet. At the Bezos Earth Fund, we're interested in innovative solutions that can make a meaningful difference at scale. Agriculture stands out because it already spans billions of acres around the world. If crops can be developed to store more carbon underground while helping farmers grow resilient harvests, the potential impact could be enormous.

That’s why, six years ago, we made an early bet on Salk’s Harnessing Plants Initiative. The idea was audaciously simple: rather than invent an entirely new machine to remove carbon dioxide from the atmosphere, could we redesign one of nature’s existing machines —a crop plant like soybean? More specifically, could we change its roots?

For thousands of years, we have bred crops for what happens above ground. Bigger grains. Sweeter fruit. Higher yields. Resistance to pests and diseases.

Roots have received less attention.

But roots are responsible for finding water and nutrients, anchoring the plant, interacting with the extraordinary microbial world beneath our feet and moving some of that carbon into the soil.

They are basically the plant’s underground headquarters.

And during the first phase of our work with Salk, scientists began opening that black box.

They have now developed more than 40 experimental “Salk Ideal Plant” lines and assembled an unprecedented dataset connecting genes to different root characteristics: how deep roots grow, how much they branch, their architecture, their chemistry and how much biomass they produce. The results are starting to get interesting.

In early soybean field trials, changing a single target gene produced plants with 30% to 150% more roots than standard plants.

One gene.

From promising to unstoppable 

This next phase, which our new grant will help support, is about turning this exciting scientific result into something that could ultimately operate at enormous scale. That requires finding the right genetic changes faster and more precisely than ever before. Fortunately, we now have tools that simply did not exist when plant breeding began.

Precision-breeding technologies, like CRISPR, allow scientists to make extraordinarily targeted changes to plants. Rather than spending decades crossing plants and hoping the right combination of traits emerges, we can increasingly identify the genetic instructions we want and work directly with them.

In addition, AI could accelerate the process even further. 

Through new funding, Salk will develop RootGPT, an open-access AI system trained on the huge amounts of genetic, root, soil-carbon and fungal data generated through the program. Instead of researchers investigating potential genes largely one by one, the ambition is to use AI to predict which genes, or combinations of genes, are most likely to produce the roots we want, test them, learn from the results and repeat.

Think of it as putting plant breeding on fast-forward.

But speed of discovery is only one part of the challenge.

The team also needs to prove that additional carbon entering the soil actually stays there.

Salk will extend field experiments, including an existing trial in Illinois from three to seven years, and use sophisticated carbon-tracing techniques to follow carbon as it moves from the atmosphere, through the plant and into different forms in the soil. The goal is to prove that these crops can move at least half a ton of additional CO₂ per acre from the atmosphere into the soil each year. Minimum. 

That number matters because agriculture has something most carbon-removal technologies do not: scale already built in. That is what makes this opportunity so compelling to the Bezos Earth Fund. These crops could be deployed through agricultural systems that already exist. If they store more carbon while helping farmers cope with drought and other stresses, climate impact and economic value begin to reinforce one another. Those are the kinds of solutions we love and that we know can create real, lasting impact. 

If just one-quarter of growers of five major crops across the world eventually planted varieties with these enhanced root systems, the climate impact could approach 200 million tons of additional CO₂ removed each year, roughly equivalent to the emissions from 43 million American cars.

We are still a ways away from being able to make that claim with scientific rigor. Biology will have its say, different crops and soils will behave differently, and carbon permanence has to be demonstrated rather than assumed. But it explains why we are excited.

The farmer has to win

There is another reason I think this approach could be powerful.

For a farmer, the climate benefit is not the most interesting thing about these plants. More roots mean more capacity to search the soil for water and nutrients. In a good year, that could help a crop make better use of the resources available to it. In a bad year, particularly a dry one, a larger and deeper root system could be the difference between a plant that keeps growing and one that fails.

This next phase will test whether these plants can better withstand drought and disease, while producing at least 15% more yield when conditions are tough. This flips the proposition around. 

We are interested in these crops because they could mitigate climate change. A farmer may be interested because they help adapt to it. And that is exactly how I want climate solutions to work.

The best solutions should not require billions of people to do something purely for the benefit of the planet. They should make economic and practical sense anyway.

Nature invented the machinery. Modern genetics lets us understand its controls. AI now helps us learn how to use them at extraordinary speed.

After thousands of years breeding the half of the plant we can see, perhaps some of the biggest opportunities are in the half we can’t.

Time to get to the root of it.