The Ultimate Guide to AI in Cleantech

The cleantech world is experiencing a quiet revolution. Artificial intelligence is no longer knocking at the door, it’s quietly remodeling the entire house. At Cleantech Group, we’ve been diving deep into how advanced AI is reshaping the industries we analyze over the past year or so, and three persistent themes have emerged:

• AI as an enabling technology that goes beyond pure-play software and is layered across the hardware value chain, and is already impacting more spaces than most people realize
  _
• The opportunities for integration of AI into the innovation process are likely still under-leveraged, and we expect to see advantages quickly gained by those adopting AI in their innovation process
  _
• Even so, the stunning pace of LLM (large language model) development is creating competitive pressure on the ecosystem, and we’re fast moving to a place where AI in the innovation process is tables stakes, and demonstrable results will be paramount

Our analysis reveals several ways that AI is transforming the clean technology landscape:

• AI as a Business: A software or cloud solution, typically aimed at efficiency gains or resource use – this is what comes to mind for most people when thinking of AI
  _
• AI as a Tool: Perhaps more important, AI is becoming an enabler in developing next-generation hardware and molecular innovations that were previously thought to be years or decades away

The Rapid Growth of AI Adoption

The adoption curve of generative AI looks less like a gentle slope and more like a rocket launch, outpacing even the internet and personal computers in their heyday (see an analysis from the IEA below). Couple this with GPU costs plummeting over the past two decades, and we’re witnessing AI models evolve from clever tools to genuinely transformative technologies.

This evolution creates a fascinating paradox. On one hand, data centers are gulping electricity at unprecedented rates, raising legitimate sustainability concerns. On the other, AI applications hold the potential to “turn the clock backwards” on climate risks by catalyzing innovations that were previously stuck in the realm of science fiction.

Current State: AI in Cleantech Remains Underleveraged

Despite its potential, AI is still surprisingly underutilized across the cleantech spectrum. Even with our intentionally broad definition of AI as a differentiator, AI-enabled companies represent just a fraction of investments in the space, a missed opportunity that’s likely to be more widely recognized in short order.

Geographically, there’s an interesting story unfolding. While cleantech investments have gone increasingly global, AI-focused cleantech remains concentrated in North America, claiming approximately 70% of dollars invested this year—significantly higher than the 50% share for cleantech investments overall. This geographic imbalance hints at both untapped markets and the potential for more diverse innovation sources.

But while the growth of AI-enabled cleantech out of APAC has been slow, the potential is enormous. The statistic below shows where despite enormous gains in dollar per hour worked in APAC, the fastest growing countries are nowhere near the peak. In the current environment of economic uncertainty, we expect a firm embrace of process and resource efficiency – expect more local solutions to local efficiency challenges to emerge out of APAC, and fast.

As an enabling technology, AI can be layered across dozens of technologies within Cleantech Group’s taxonomy. However, we see the most evidence of AI’s impact, when analyzing thematically, emerging in three broad categories today:

1. Accelerating Deep Tech Innovation
2. Enhancing Manufacturing Efficiency and Sustainability
3. Enabling Climate Change Adaptation

AI in Deep Tech: Enabling Faster Innovation

For those who have followed Cleantech Group’s research over the past few years, you will have noticed that we have tracked the trend of deep tech innovation by identifying a few proxy technology categories. Encouragingly, we’ve observed that deep tech innovation has become one of the most prominent themes in cleantech over the past few years.

First reaction for many is to see AI and deep tech as two exclusive categories, or even simpler, view things only through the lens of software and hardware. We’ve thought carefully about how to evaluate the impact of AI and have worked to identify just over 2,500 companies in our taxonomy that we believe are deriving some sort of benefit from AI – i.e., are AI-enabled and not AI-only.

Within that data set, we’ve identified just over 162 companies who both fit into the deep tech proxy categories and are AI-enabled. What one will notice first is that, on a dollars-invested basis, AI-enabled deep tech companies account for an average of 20% of dollars invested in deep tech for cleantech. That said, AI-enabled fusion is the most prominent category, although others are beginning to emerge.

It is our position that AI is still significantly under-leveraged in deep tech innovation, as indicated in the chart below, where deep tech innovators that use AI somewhere in their process are raising funds at earlier development stages than are deep tech companies on average.

Avalanche Energy: Making Desktop Fusion a Reality

Avalanche Energy is taking a distinct approach to fusion development, building desktop-sized fusion reactors that can be stacked and scaled or used in industrial settings. Avalanche Energy employs particle-in-cell plasma simulation that models fusion reactor configurations with unprecedented precision.

Fusion research can take months to set up experiments, so Avalanche’s AI-powered approach enables:

• Detailed modeling of reactor configurations down to the subatomic level
  _
• The ability to go from simulation to physical lab experiments within hours, a process that traditionally could take weeks or months
  _
• Lightning-fast iteration cycles for testing fusion reaction variables – accelerating the path to commercial uses

Zanskar: Cutting-edge Subsurface Models Flatten Geothermal Project Cost

Geothermal energy has always promised abundant clean power, but the financial risks of exploration have kept many projects on the drawing board. Enter Zanksar, whose AI-driven subsurface sensing technology is like giving geothermal developers X-ray vision.

Their system measures temperature gradients, gravity variations, geological formations, and tectonic movement before a single drill bit touches soil. The impact is transformative:

• Exploration costs – often 30-40% of total project budgets – are slashed dramatically
  _
• The key promise here is that commercial developers who previously could not grapple with geothermal’s risks are now taking a second look
  _
• Sites that were previously considered too uncertain are being reevaluated as viable energy sources

AI in Manufacturing: Improving Efficiency and Sustainability

We have made an effort at Cleantech Group to understand where cleantech innovation is most impactfully playing a role in the “real economy”, e.g., manufacturing processes and the manufacturing of new energy products. Despite being a consistent theme in clean tech investments, AI’s penetration into heavy industry remains surprisingly minimal, and a blue ocean opportunity waiting to be capitalized on.

Perhaps most obviously, materials discovery is where we have observed the majority of industry- and manufacturing-related AI innovation.

The trend of innovators using AI as a differentiating component of their R&D and product formulation processes, is, however, beginning to take shape now.

Mitra Chem: AI Enabling Batteries with Abundant Materials

Mitra Chem is using AI to simulate and synthesize thousands of cathode materials and has created a battery innovation engine that:

• Aims to speed development timelines by 90%, which would bring new battery formulations from lab to market in months rather than years
  _
• Enables the rapid development of high-performance batteries using abundant, ethically sourced materials, e.g., reducing problematic materials like cobalt in the supply chain
  _
• Dramatically reduces R&D costs that make battery innovation a challenge from which to finance and profit

Cosmos Innovation: Compressing Timelines to High Efficiency Solar

Singapore-based Cosmos Innovation is aiming to get more juice out of the solar squeeze on two fronts: producing high-efficiency solar cells but in a faster and cheaper way. Their Mobius platform acts as both materials’ scientist and process engineer, supporting every step from molecular design to manufacturing execution.

This AI-powered formulation and manufacturing process:

• Eliminates the costly trial-and-error approach that has hampered advanced solar development
  _
• Enables manufacturers to rapidly respond to material availability challenges and supply chain disruptions
  _
• Improves the economics of perovskite-silicon tandem cells, which promise cell efficiencies well beyond the 22-24% common today, potentially reaching 30-35%
  _
• If successful, use of these cells can reduce the physical footprint of solar installations by a third or more—critical in land-constrained environments

Fero Labs: Putting AI Tools Onto the Steel Manufacturing Floor

Steel production is one of humanity’s oldest and most carbon-intensive industries, but Fero Labs is offering an opportunity for major efficiency increases, without waiting for next-gen facilities. What makes their approach particularly revolutionary is how they’ve designed their AI to work with operators, not data scientists.

Their system:

• Makes direct, real-time interventions in manufacturing processes that reduce energy consumption and improve quality
  _
• Uses “white box” AI that allows operators to understand exactly what parameters the system is monitoring and why it’s making specific recommendations
  _
• Creates cross-functional understanding between floor operators, engineers, and management about process optimization
  _
• Bridges the gap between data scientists and steel workers, building trust in AI systems and avoiding challenges of entrenched interests
  _
• Could potentially reduce emissions from steel production by 8% while improving quality and reducing costs

Atacama Biomaterials: Teaching AI to Reinvent Plastics

The journey to ubiquity of alternative plastics presents a consistent cost and materials availability challenge. Atacama Biomaterials’ Marie Curie AI platform represents a potential leap in bio-based alternatives. Rather than incremental improvements to existing bioplastics, their system fundamentally reimagines material formulation.

Their AI platform:

• Analyzes thousands of natural fiber combinations against a proprietary biomass database to develop plastics alternatives
  _
• Identifies specific combinations of plant fibers and natural binders that can meet or exceed the performance requirements of conventional plastics
  _
• Addresses the performance limitations that have traditionally kept bioplastics from mainstream adoption
  _
• Creates a pathway to truly sustainable packaging and products that doesn’t require consumers to compromise on quality or performance

Critical Materials Innovation: Urgency Creating Demand for New Approaches

The clean energy transition depends on a reliable supply of critical minerals, and AI is transforming how we discover and extract them. In Q1 2025, critical materials innovation comprised the highest percentage of cleantech deals it ever has, with mining innovation playing the leading role.

Kobold Metals: Mining Exploration Gets the AI Treatment

Backed by investors including Bill Gates and Jeff Bezos, Kobold Metals is applying machine learning to the age-old challenge of finding mineral deposits. Their proprietary TerraShed database combined with their Machine Prospector tool is changing the economics of exploration.

The Kobold system:

• Analyzes geological data from diverse sources to identify mineral deposits with unprecedented accuracy
  _
• Dramatically reduces the need for expensive and environmentally disruptive physical exploration
  _
• Continuously improves its data models with each new data point, creating a virtuous circle of increasing accuracy
  _
• Makes site identification and drilling decisions significantly faster and cheaper, potentially unlocking previously uneconomic deposits
  _
• Could help close the growing supply gap for critical battery materials like nickel, cobalt, and lithium

Earth AI: Finding Mineral Needles in Geological Haystacks

Earth AI has developed a mineral targeting platform so precise it can identify promising deposits in areas as small as two square kilometers. By leveraging 50 years of geological data and continuously refining their model, they’ve created a system that gets smarter with every drill hole.

Their breakthrough approach:

• Identifies previously overlooked or unknown deposits with 25 times greater accuracy than traditional methods
  _
• Significantly reduces the environmental footprint of exploration by enabling precise, targeted drilling
  _
• Creates a continuous feedback loop that improves prediction accuracy with each new sample
  _
• Verifies the extent and quality of deposits up to four times faster than conventional techniques
  _
• Could dramatically increase the success rate of mineral exploration; their estimates claim up to 66% success rate eventually

Novamera: Precision Extraction for the 21st Century

Novamera has developed a closed-loop system for mineral extraction that combines AI-powered imaging with precision drilling techniques. Their proprietary near-borehole imaging tool represents a fundamental rethinking of how we access critical materials.

The Novamera system:

• Collects high-resolution subsurface data
  _
• Creates detailed 3D maps of ore body geometry
  _
• Calculates optimal drill trajectories that maximize resource recovery while minimizing energy use
  _
• Increases success rates dramatically while reducing the financial and environmental risks of extraction
  _
• Delivers stunning results: 95% waste reduction, 50% cost reduction, and 44% GHG emissions reduction compared to conventional mining

AI for Climate Adaptation & Resilience: Bending the Curve

Climate adaptation often gets sidelined in environmental discussions, often due to a reluctance that preparing for climate impacts somehow means surrendering in the fight against their causes. But here’s the stark reality: climate change isn’t just coming; it’s already reshaping our world. The increasing frequency and intensity of extreme weather events demand urgent adaptation strategies, and AI is emerging as a powerful ally in this crucial effort.

What’s particularly interesting is that even during investment downturns in adaptation and resilience technologies, AI-enabled solutions consistently maintain a significant foothold—representing at least 15% of investments in this category. This persistent AI presence signals something important: intelligent systems are uniquely suited to help us navigate an increasingly unpredictable climate future.

Google FireSat: Spotting Wildfires Before They Rage

Wildfire resilience technologies saw growing interest in 2024, with major AI players now entering the arena. Google’s FireSat represents a quantum leap in early detection capabilities. This planned constellation of 50 low-flying satellites will revolutionize how we monitor forest landscapes:

• Updates forest imagery every 20 minutes—an unprecedented frequency for comprehensive coverage
  _
• Employs sophisticated AI models to analyze images at 5×5-meter resolution
  _
• Dramatically outperforms current systems that typically detect fires only after they’ve grown to 2-3 acres
  _
• Industry estimates are that if response times were reduced by 15 minutes, large fire frequency could be reduced between 3-7%, placing high value on Google Firesat’s improvement to the response time
  _
• Potentially saves billions in property damage and ecosystem losses annually

The difference between spotting a fire when it’s the size of a campsite versus the size of a football field can mean the difference between a minor incident and a catastrophic blaze. Google’s system aims to shrink this critical detection window from hours to minutes.

ThinkLabs AI: Creating Digital Twins for Power Grid Resilience

Today’s generational growth in electricity demand is requiring more capacity on the grid, but also placing a higher importance on resilience against weather events and costly power outages. ThinkLabs AI, a GE spin-off, is taking grid intelligence to a new level with its physics-informed digital twin technology.  It:

• Creates comprehensive virtual replicas of entire grid systems
  _
• Feeds critical training data to grid systems not just for current conditions but for events the grid has never encountered
  _
• Applies “physics guardrails” to future scenarios, keeping predictions grounded in real-world possibilities
  _
• Pairs with real-time monitoring to identify emerging threats before they materialize
  _
• Prescribes preventative actions to maintain resilience against increasingly severe weather events

This approach represents a fundamental shift from reactive to proactive grid management—essential as climate impacts intensify. ThinkLabs isn’t alone in this space; Google’s X moonshot laboratory has been operating its “Tapestry” grid modeling system with impressive real-world results:

IONATE: Reimagining the Humble Transformer

While transformers have been the backbone of electrical grids for over a century, IONATE is teaching this old dog impressive new tricks. Their reimagined transformers feature an autonomous control module that’s constantly monitoring and adjusting to grid conditions.

In a world facing both surging data center power demands and increasingly unpredictable weather events, IONATE’s innovation:

• Dynamically manages electromagnetic functions in real-time, responding to changing grid conditions and reducing downtime risks
  _
• Intelligently shifts power flow while maintaining power quality – think data centers that have a need for uninterruptible power supply
  _
• Creates additional capacity in existing grid infrastructure—potentially avoiding billions in unnecessary upgrades

Enko: Protecting Food Systems Through AI-Powered Crop Science

Perhaps the most overlooked climate vulnerability lies in food systems. Changing agricultural conditions—from emerging pest pressures to disease patterns to drought intensity—create unprecedented risks to global food security. Enko is tackling this challenge head-on with AI-powered development of crop treatments:

• Developed the ENKOMPASS platform that leverages extensive DNA libraries to identify novel crop protection formulations
  _
• Creates precisely targeted solutions for insect, disease, and weed management tailored to specific crops
  _
• Makes plants significantly more resource-efficient—critical as growing conditions become more challenging
  _
• Has potential to reduce the staggering $10B in annual pest-related crop losses in the U.S. alone
  _
• Accelerates the development of climate-resilient agricultural practices that can adapt to changing conditions

Matter Intelligence: Hyperspectral Sensing for Precision Interventions

The effectiveness of AI models depends heavily on the quality of input data—which is why companies developing advanced sensing technologies are so crucial to adaptation efforts. Matter Intelligence represents the cutting edge of this field:

• Combines hyperspectral and thermal sensing to measure – not just image – ground conditions
  _
• Captures shape, composition, and temperature data with meter-level precision
  _
• Assesses building and infrastructure risks with high enough detail to identify safety vulnerabilities
  _
• Can reportedly detect disease in individual plants, enabling ultra-precise agricultural interventions
  _
• Reduces the risk of fertilizer over-application, creating significant resource efficiency benefits

By providing this level of detailed environmental data, Matter Intelligence enables the kind of precise, targeted interventions that will be essential as climate conditions grow more volatile and resources more constrained.

The Evolution of AI in CleanTech: What’s Next?

As the AI-Cleantech ecosystem evolves, we’re seeing distinct trends emerge. Today’s market features numerous newcomers riding the adoption wave, often using efficiency gains to compensate for technical limitations. Many deploy industry-tailored small language models trained on publicly available data—satellite imagery or industry databases—or simply wrap existing large language models in sector-specific interfaces.

However, the competitive landscape is shifting rapidly. Over the next few years:

• Companies relying solely on public data or general-purpose LLMs will face increasing pressure to differentiate
  _
• The imperative will shift toward demonstrating rapid, tangible proof points of AI’s benefits
  _
• For hardware companies, this means showing how AI translates to lower end-product prices
  _
• In software, traditional SaaS models may give way to success fee structures where companies compete on actual versus promised savings

Several key elements will define tomorrow’s winners:

• Proprietary data will become increasingly valuable, with novel data acquisition techniques serving as foundations for unique AI solutions
  _
• Technologies that improve physical process efficiency—whether in manufacturing or R&D—will gain greater appreciation
  _
• Solutions that generate measurable physical proof points will outcompete purely digital offerings

For those launching or backing AI-Cleantech ventures, the message is clear: while today’s market offers substantial opportunity, differentiation will become increasingly critical. Many solutions that appear valuable today will struggle to maintain their distinctiveness as the market matures. The most successful players will be those who can quickly pivot toward innovations that deliver demonstrable value in tomorrow’s market.

AI is already making its mark in cleantech, but we’re only at the beginning of this transformation. As single-point solutions evolve into comprehensive systems—like the grid modeling examples we’ve explored—we’ll see increasing integration across previously separate domains.

The energy consumption of AI systems remains a significant challenge, but there’s a compelling symmetry in how AI must help solve its own problems by enabling more efficient grids and accelerating clean baseload power development.

Perhaps most importantly, these powerful tools must become accessible to non-specialists. As we saw with Fero Labs’ operator-friendly AI for steel production, democratizing access to these technologies will be crucial for widespread adoption.

And finally, adaptation deserves more than an afterthought in our climate strategy. With climate effects already reshaping our world, AI offers a powerful way to diversify our approach—helping us not just fight climate change but navigate the changes already underway. By spreading our bets across both mitigation and adaptation, we create a more robust response to one of humanity’s greatest challenges.