The carbon cost is worse than the financial cost. Google's total greenhouse gas emissions rose by more than 50% in 2024 compared to five years earlier, even as the company worked harder than any of its peers to source clean power. In Armstrong County, Texas, the company is working with developer Crusoe Energy on a nearly gigawatt-scale natural gas plant to power its Goodnight data center campus.
But here is something else to consider: In 2024, Google ran a 17-week trial on 2,400 transatlantic American Airlines flights using a system designed to predict and avoid the formation of contrails. Contrails are the ice crystal trails left by jet engines that account for roughly a third of aviation's total warming impact — more than the impact of the fuel burning. The AI model rerouted flights slightly to avoid the atmospheric conditions that produce persistent contrails, and in doing so, cut contrail formation by 62% without any meaningful increase in fuel burn.
Around the same time, Microsoft used its Azure Quantum Elements platform to sift through 32 million possible chemical candidates for new battery chemistries, and in 80 hours narrowed the field to a handful of promising compounds that could reduce the amount of lithium required by as much as 70%. Meta, working with Georgia Tech, built one of the largest open-source datasets for discovering better sorbent materials for direct air capture, the process of pulling carbon dioxide directly from the atmosphere. Researchers ran nearly 40 million quantum mechanics calculations across 8,400 candidate materials, looking for those that could grab CO2 efficiently without also absorbing water from the air.
All of these things happened in the past two years. They were largely invisible to consumers. Yet they produced meaningful, even transformative climate benefits. Crucially, they cost almost nothing compared to the AI infrastructure buildout. They were side projects, pursued by teams whose quarterly numbers did not depend on their product’s success.
I argued two years ago in an interview with Heatmap that the steep financial and carbon costs of the AI buildout are worth it, and that if we stick with it, the power of AI will quickly yield innovative solutions to address climate change. But the opposition to data centers and AI deployment has created a frustrating paradox. A sector that has spent years describing a technology primarily as a threat — to the grid, to society, to humanity itself — will not, at the end of that time, be in a strong position to invest in what that technology can build. The sector wrote itself into the role of the regulator and critic at precisely the moment it should have been adopting the role of the main customer.
In the first half of 2026 alone, investors put $407 billion into AI startups, Pitchbook calculated. Climate tech, over the same stretch, did fine: $26.1 billion, according to CTVC’s insights report, up 55% year-over-year, the strongest first-half investment numbers since 2022. But low-carbon data centers alone took 34% of it, and two of the sector’s biggest deals were both for data center infrastructure. We essentially took an historic year of climate tech investment and used it to become AI's electricity supplier.
This is definitely a net positive, and critical to a clean hyperscale movement. But there’s more to be done.
Roughly one climate venture dollar in five went to something AI-enabled in 2025, which is a real increase from previous years. But out of $40 billion total climate tech investment last year, that amounts to only about $8 billion. Set that against the $242 billion that went into AI startups in a single quarter — the world's entire annual investment in AI for climate is roughly what AI startups raised every three days at the start of this year.
The three breakthroughs I mentioned at the beginning of this article are just the beginning of what AI can do for the climate — in many cases they’re the easy breakthroughs. They’re prediction, search, and optimization problems where the AI is essentially a faster pair of eyes.
The larger prize is what my colleagues at Obvious Ventures and I have come to call “generative science.” These are models trained in chemistry, physics, and biology that can propose genuinely novel arrangements of atoms rather than merely sorting through existing ones. This is where we unlock nuclear fusion, carbon-free cement and steel, and grid systems that balance themselves. We can make cancer vaccines and drugs optimized with a single patient’s DNA. If we ever make it to another planet, it will be because of AI. The same is true if we ever learn to sustainably feed 10 billion people.
This is not a speculative category anymore. A series of startups are making meaningful breakthroughs in these kinds of technologies. In Cambridge, England, a materials science company called CuspAI is building foundation models for chemistry to find materials for direct air capture of carbon dioxide. In California, Periodic Labs, founded by the researcher who led materials and chemistry at Google DeepMind, raised a $300 million seed round at a $1 billion valuation to run autonomous synthesis labs hunting for superconductors that work at higher temperatures (its valuation has since risen dramatically). And Zanskar, a company Obvious Ventures has backed, trained its models on subsurface data and a century of drilling and satellite records to find geothermal resources the industry had already written off. Last year, it identified a blind site in western Nevada with no geysers or surface expression that has the potential to generate over 100 megawatts.
Zanskar, however, is an exception. None of these other technologies were backed by climate funders. CuspAI and Periodic Labs have received financing from sovereign wealth funds, chipmakers, generalist growth firms, and individual investors who made their fortunes in software.
I’ll be the first to acknowledge that building a climate tech company isn’t easy. Investors often have to make two bets at once: that the science will work and that, if it does, there will be a viable business on the other side. Unlike chatbots from the frontier AI labs that have grown to $1 trillion valuations in less than five years, meaningful climate tech breakthroughs take longer to deploy, and even longer for their impact to put a meaningful dent in climate change.
But companies like CuspAI, Periodic, and Zanskar are proof of what’s possible when we point AI and climate tech in the same direction.
There are three main things we can do differently to continue that progress, and we can start each of them today.
- Change how we underwrite climate technology. The main strategy of an early-stage deep climate tech fund is to kick the tires on a project’s financing and hardware. But a firm that cannot read a model architecture will not fund one. Firms should have at least one investor who has been involved with AI in addition to chemistry and infrastructure specialists. They should also have experts who can assess scientific risk as early as possible.
- Focus on systemic challenges. At Obvious, we refer to this as “world positive,” meaning we favor technology that enlarges the pie for the maximum number of people, rather than benefiting an exclusive few. The Meta–Georgia Tech carbon capture dataset was released as open-source infrastructure that any company can build on. Data sets, benchmarks, and simulation environments for climate-relevant science are chronically underfunded, but they’re exactly what philanthropic and public research capital is good at. This isn’t altruism. It’s the highest-leverage dollar in our arsenal that can create more technologies, more customers, and more impact.
- Stop treating the frontier AI labs as adversaries and start treating them as infrastructure. The top frontier labs are producing climate-relevant science as a byproduct. That work is currently underrecognized, underfunded, and uncommercialized. Someone should be standing at the end of that pipeline with a checkbook, or at least using these platforms with a focus on commercializable outcomes. This is what’s happening in techbio. We should be doing the same for climate.
As with any economic shift, aligning the incentives gets us much further than any fleeting policy commitment. When alignment happens, it creates a flywheel where AI powers research in climate tech, whose breakthroughs get fed back into AI to run models more cleanly and efficiently.
Jakob Uszkoreit, the former Google engineer who co-authored the transformer architecture that powers today’s leading large language models, described to me the paradox this way: AI needs carbon to get off the ground, but once airborne, it becomes the mechanism that solves the carbon problem. The question is whether we achieve liftoff before the end of the runway.
