Could Data Center Demand Spark A Climate Tech Rebound?

Global demand for data centres has surged dramatically in recent years, with energy consumption rising in tandem as artificial intelligence (AI) and digital services become more widespread. According to the International Energy Agency (IEA), data centres worldwide consumed approximately 415 terawatt-hours (TWh) of electricity in 2024, which represented about 1.5% of the total global electricity use in that year. This demand has grown at a rate of 12% per year over the last five years. Hyperscale data centres—those operated by major cloud and AI providers—have doubled their energy consumption in recent years, even as traditional data centres have become more efficient. The IEA estimates that global data centre electricity consumption will more than double by 2030, reaching around 945 TWh (Chart 1).

Chart 1: Global data centre electricity consumption by equipment, Base Case, 2020‒2030

Data Centre Build-Outs: Scale of the Energy Demand Challenge

By way of comparison, one large hyperscale data centre can consume as much electricity as 100,000 average North American households, and some of the new AI “hyperscale” centres under construction will require 20 times more. Between January and August 2024 alone, Microsoft, Meta, Google, and Amazon committed roughly $125 billion to building and operating AI-focused data centres (JPMorgan 2024) (Table 1).

Table 1: AI Capital Expenditures and Total Data Centre Operating Costs between January and August 2024

The growing appetite for energy presents a challenge and an opportunity. According to Canary Media, the looming increase in electricity demand has led utilities in the U.S. to plan the construction of new fossil fuel power plants and keep coal plants online for longer. These actions, experts fear, will put even more strain on the decarbonisation goals, aside from U.S. utilities and the ratepayers who will ultimately pay for the power plants and grid infrastructure needed to support data centre growth. Globally, some countries are responding with regulations to the rising demand for electricity from data centres. Amsterdam has paused new builds to prioritise urban sustainability, while Singapore now mandates higher server-room set points (26°c or higher) to reduce cooling loads, albeit at the cost of shorter chip longevity.

Startup Are Developing New Solutions To Decarbonise Data Centres

As AI drives up data-centre demand, climate startups are seizing a golden opportunity to decarbonise digital infrastructure. In 2024, 280 Earth, a startup that spun out of Google’s “moonshot factory” X, signed a $40 million agreement to capture the carbon dioxide emissions of some major companies through an initiative called Frontier, which includes Stripe, Alphabet, Meta, Shopify, and McKinsey. More recently, new startups, such as Scalvy, have sought to develop modular power electronics for data centre racks to convert voltage and current with lower losses.

However, despite new solutions coming on board to address the energy problem, climate tech startups face a challenging fundraising environment combined with an uncertain policy landscape. According to BloombergNEF data, global climate tech equity financing plunged 40% in 2024, dropping from $84 billion in 2023 to about $51 billion. This marked the third consecutive year of decline in climate tech funding, despite the overall increase in venture capital investment. Meanwhile, investors funnelled nearly $100 billion into AI companies in 2024, a massive uptick in AI funding that “siphoned” capital from decarbonisation ventures.

Yet, this divergence belies a deeper synergy. Many experts note that AI and climate solutions need not be competitors; AI can, in fact, accelerate climate technology. Notably, sectors tied to the AI-driven demand for data infrastructure defied the downturn. Energy investment surged to $9.4 billion—a 12% increase—while buildings saw a 10% rise, totalling $2.7 billion. According to Sightline Climate’s 2024 Climate Tech Investment Trends report, of the ten largest climate tech deals last year, four (Scala, Crusoe, X-energy, and Form Energy) were for businesses involved in the development of sustainable data centres (Chart 2).

Chart 2: Top 10 climate deals, 2024 ($bn)

In 2023, the top deals were dominated by financing for batteries and the supply chain, which averaged $780 million. However, in 2024, the average deal size fell to $500 million, roughly one-third lower than the previous year’s average. For the climate tech sector to rebound and leverage the parallel surge in sustainable data-centre and AI investment, governments should first promote broader R&D funding and tax incentives aimed at further improving the efficiency of AI-related computer chips (Chart 3).

Chart 3: Efficiency improvement of AI-related computer chips, 2008‒2023

At hyperscale, chip-level innovations can significantly reduce power consumption per inference, resulting in substantial electricity savings at the grid level. Table 2 highlights leading examples from China’s DeepSeek, the U.S.’s Groq and South Korea’s DEEPX, each employing novel optimisations to slash compute requirements and boost efficiency. Furthermore, on-device AI paradigms, which move inference onto local edge devices rather than centralised cloud data centres, can cut energy use by two to three orders of magnitude per task.

Table 2: Emerging AI Chip Efficiency Innovations by Company and HQ

Energy Credits Could Spur Low-Carbon Data Centres And Broader Investment into Climate Tech

To accelerate the low-power transition, policymakers in all major data centre markets should also implement an energy credit trading system that rewards companies for executing workloads on energy-efficient hardware or shifting computing tasks to periods of surplus renewable energy (Table 3).

Table 3: Potential Benefits of Energy Credit Trading

By contrast, carbon-credit trading remains uneven. Although the EU’s Emissions Trading System mandates carbon credit trading, most major economies lack comparable nationwide frameworks. In the U.S., compliance markets are currently limited to California’s Cap-and-Trade Program, the Northeast’s Regional Greenhouse Gas Initiative, and Washington’s Climate Commitment Act, with no federal scheme in place. This geographic patchwork prevents data centre operators in underserved regions from monetising low-carbon computing strategies, leading to underinvestment in energy-efficient chips and a continued reliance on carbon-heavy grid power.

Scaling both the energy credit trading system and national carbon-credit markets beyond today’s patchwork would ensure that the data centres powering tomorrow’s AI also drive the next wave of climate tech investment, while creating market-friendly price signals and liquidity that catalyse broader climate tech deployment, rather than locking in another decade of fossil fuel incumbency.