Is AI The Secret to Turbocharging Proven Climate Tech?
Imagine it's 2030
The world has undergone a remarkable transformation. Renewable energy sources, primarily solar and wind, now generate 60% of global electricity - from 30% in 2023. Within six years, installed renewables capacity has tripled from 3,600 GW to a staggering 11,000 GW by 2030. This incredible growth is most noticeable in emerging markets and developing economies (EMDE), where renewable capacity has quadrupled to 3,000 GW!
The widespread adoption of renewables has been made possible by significant enhancements and modernization of power grids. Transmission and distribution networks have expanded by nearly 10 million kilometers since 2020. Additionally, we are harnessing various sources of flexibility, such as energy storage—utility-scale battery storage capacity is now 36 times larger than in 2022—along with low-emission dispatchable power like nuclear plants and improved demand management strategies.
Key sectors are now largely electrified, particularly road transport and buildings. Major automotive manufacturers have phased out internal combustion engines, and electric vehicles (EVs) now make up two-thirds of all car sales - from today ~20%. This revolution is spearheaded by advanced economies, including China, with EMDEs quickly catching up as EV costs continue to drop. Charging infrastructure is widespread and rapidly expanding. In the building sector, 35% of global heating demand is met by heat pumps.
Light industries that require low- to medium-temperature heat are primarily powered by a mix of heat pumps, electric boilers, and resistance heaters. Heavy industries such as iron and steel, chemicals, and cement are adopting innovative solutions: approximately 1 gigatonne (GT) of carbon capture, utilization, and storage (CCUS) capacity has been installed, evenly split between EMDEs and advanced economies, and low-carbon hydrogen production has increased tenfold since 2022. The first regional commercial flights of electric planes have taken off, signaling a new era in aviation.
Overall, global fossil fuel demand has dropped by 25% since 2020 - we are truly moving towards a net-zero world.
To get there, we need to turbocharge the deployment of climate tech we mostly already have
The "Vision 2030" outlined above is based on the latest report from the International Energy Agency (IEA), detailing the steps needed to align with a 1.5°C climate pathway. According to the report, 80% of the necessary emission reductions by 2030 can be achieved using technologies already in the market (Figure 1).
This means that realizing "Vision 2030" largely hinges on optimizing and expanding the use of these current technologies. In a nutshell, within the next six years, we need to (Figure 2):
Triple renewable capacity and promote electrification and fuel switching: Triple renewables capacity to reach 11,000 GW by 2030. Transition from fossil fuels to low-/zero-carbon energy sources like gas, nuclear, hydropower, geothermal, and bioenergy. Drive electrification, particularly in road transport, building heating, and light industrial processes.
Double improvements in energy intensity: Enhance the efficiency of appliances, equipment, trucks, building envelopes, and planes.
Cut methane emissions from fossil fuels by 75%: Deploy reduction measures and technologies (e.g., stop non-emergency flaring and venting, enhance leak detection and repair programs).
Achieve near-zero emissions in new heavy industry capacity. Continue to mature novel technologies. By 2030, CCUS and hydrogen deliver a 1 Gt reduction in emissions, particularly by making new heavy industry capacity carbon-neutral.
AI is a crucial ally in overcoming bottlenecks to climate tech deployment
We know we need to complete this within the next six years. However, it's not easy. We're facing significant bottlenecks in the deployment of the required, mostly already proven, technologies.
I want to make the point that AI must be leveraged as an ally to turbocharge the deployment of "old school" proven climate tech, given that AI-based solutions can address many critical bottlenecks.
To support this point, let's take a closer look at the need to triple renewable capacity by 2030, the most critical target we need to reach within this decade - and how AI could support this effort by addressing bottlenecks. Note: There are many more bottlenecks, but I try to focus on the most critical ones for now.
1) Support renewable energy site selection
Bottleneck: Identifying optimal sites for renewable energy projects is often challenging. Selecting the best sites involves optimizing multiple dimensions, such as regional availability of high-quality renewable resources, potential environmental and social impacts (like job creation and social acceptance), and technical feasibility (like grid connections). This process is often slow and complex.
AI can help support the identification, assessment, and selection of optimal sites for renewable energy projects, e.g.,
AI algorithms can be trained on labeled geospatial data, including topography, land use, weather patterns, proximity to critical infrastructure (e.g., power grid, “uncongested” grids), and other environmental, economic, regulatory, or social criteria to help determine sites suitable for a renewable energy project.
2) Simplify and accelerate permitting
Bottleneck: Permitting is painfully slow today. Delays in permitting significantly hinder project deployment. For example, in Europe, 60 GW of onshore wind capacity—four times what was commissioned in Europe in 2022—is currently held up by permitting procedures. On average, permitting timelines range from around nine years for offshore wind, three to nine years for onshore wind, and one to five years for ground solar PV.
AI can help simplify and accelerate permitting of renewables projects, e.g.,
Natural Language Processing and Machine Learning (ML) algorithms can be leveraged to quickly review applications, identify missing or incorrect information, and suggest corrections (e.g., for technical drawings or long text-based submissions).
Large Language Models (LLMs) can help process public comments, streamline the application process, and train less experienced workers in the siting and permitting ecosystem.
LLMs can also be used to extract and organize unstructured data like past permits, approvals, and environmental reviews, pulling that information into structured data that can then be deployed to fine-tune models for specific contexts, building tools for e.g., developers or government reviewers.
3) Remove grid connection challenges
Bottleneck: Connecting new capacity to the grid is often challenging. Delays in securing grid connections and congested grids are major hurdles. Currently, 3,000 GW of solar and wind capacity are waiting for grid connection. Securing connections can take several years (up to three years in the US). Additionally, often, the existing physical grid infrastructure is “congested” - meaning that it cannot support new large-scale power capacity.
AI can help with removing challenges around grid connection, e.g.,
AI algorithms can forecast future renewable energy production with high accuracy, considering, for example, weather patterns, seasonal variations, and maintenance schedules. Grid operators can leverage this in planning capacity and ensuring that the grid can handle new energy projects.
Particularly, ML algorithms can support complex technical studies (e.g., process grid data, historical interconnection applications, and simulate grid scenarios) to suggest optimal points of interconnection for new generation projects and also recommend upgrades to existing grid infrastructure.
LLMs could also ease the processing of grid connection applications by helping screen and validate unstructured data in those applications, such as documents pertaining to land ownership
4) Support the creation of resilient grids that can accommodate renewables
Bottleneck: Integrating renewables at a large scale requires smart, flexible, and expanded grids. Today's grids are built around predictable thermal power plants (e.g., coal, nuclear, and gas) and cannot handle large loads of intermittent renewable energy (e.g., solar, wind).
AI can support the creation of smart, resilient, and expanded grids to enhance the integration of renewables, e.g.,
AI models can help optimize grid designs by analyzing geographical data, population density, and load requirements. This involves, e.g., identifying the optimal locations for substations, transformers, and other components and optimizing the routing and sizing of power lines to minimize transmission losses and maximize grid efficiency.
AI could also support Transmission Expansion Planning—a complex process that determines the optimal location and capacity of new transmission lines and the best timing for new construction.
AI enables the development of Digital Twins of power grids. These are virtual models of real-world power grids that simulate their behavior and performance using data from sensors, satellites, and other sources, enhanced with machine learning and simulations. They are crucial for integrating renewable energy into power grids. They help manage the complexities of microgeneration from sources like solar and wind and the demand from electric charging points, ensuring stable and efficient grid operations.
5) Support the ramp-up of renewables in EMDE
Bottleneck: Financing costs for renewable projects in emerging and developing markets are too high: The cost of capital in these markets is at least double that in advanced economies. A 1% increase in the weighted average cost of capital increases wind and solar PV generation costs by at least 7%. This is partly due to a regional lack of experience in clean technology project development, less favorable policy and regulatory environments, and perceived and real macroeconomic risks by investors.
AI can support project planning, risk assessment, and financing for renewable projects in EMDE, e.g.,
LLMs can be leveraged to support project development in markets with insufficient renewables project experience (e.g., help with generating scenarios, workflows, project plans, financial assessments, and first drafts of project proposals based on global best practices tailored to the unique project context).
ML algorithms can also help governments design efficient renewable energy auctions by creating scenarios that simulate different auction formats and rules to see which ones lead to the most desirable outcomes.
AI models can also generate accurate predictions about future energy prices, material costs, interest rates, and other factors that impact the financial viability of renewable energy projects. These predictions can be integrated into financial models to determine the expected return on investment for projects.
AI models can also assist in designing optimal financing strategies for projects. This includes determining how to structure the capital stack and identifying suitable investors to include, ensuring that the financial plan aligns with the project's goals and risk profile.
To unleash the scale of change needed, we must dive deeper into which critical deployment bottlenecks AI can solve today
Of course, AI is not a silver bullet here. Many of the discussed solutions also require supportive policy and regulatory reforms (e.g., permitting processes, conducive policy environments in EMDEs), changes in physical infrastructure (e.g., expansion of power grids), or supply chains that can provide the materials needed.
Furthermore, we must ensure AI can be truly net-climate positive and to the benefit of people (see risks discussed in my previous article).
However, there is real potential for AI to be a powerful ally in turbocharging the deployment of proven climate tech this decade. For each of the major activities we need to drive (from tripling renewables to electrifying the economy, driving energy efficiency improvements, to cutting methane emissions), we need to delve much deeper into understanding what could prevent us from reaching our "Vision 2030." In every relevant context—whether tech-specific, industry-specific, or region-specific—we must identify the most critical bottlenecks and explore how AI can be leveraged to solve them.
If you have any further thoughts, ideas, or questions on how we can advance this topic - please reach out and/or comment!