By: Archana Joshi, Head - Strategy, Enterprise AI & Banking and Financial Services, LTIMindtree

Today, generative AI or Gen AI is solving complex challenges that require data-driven insights, and its impact is being felt across industries and consumer segments. Gen AI is helping global leaders manage climate change, improve agricultural yield, minimize pollution, and improve drug efficacy. But all this data crunching comes at a cost.

Large Language Models (LLMs) such as Generative Pre-trained Transformers (GPTs) consume a lot of energy, raising sustainability concerns when the world is increasingly focused on reducing carbon footprints. The International Energy Agency expects the computational power needed to train AI models to increase 10 times by 2026. Goldman Sachs also predicts that data center power demand will increase by 160% by 2030.

This is particularly concerning for financial institutions as they are increasingly adopting sustainable finance practices across the board. They need to make investment decisions after careful consideration of Environmental, Social, and Governance (ESG) regulations. The EU’s Sustainable Finance Disclosure Regulation (SFDR) requires Financial Market Participants (FMPs) to disclose information about sustainability and ESG.

Financial institutions need to be able to balance these aspects when leveraging AI to ensure ESG compliance. This blog explores how you can help your organization minimize AI’s impact on the environment.

The environmental impact of generative AI

The environmental cost of Gen AI is significant. A single LLM, such as GPT-3, can consume as much as 1,287 megawatt hours of electricity—equivalent to the energy consumption of 120 U.S. households annually.

McKinsey reports that the carbon footprint of AI in financial services and other data-intensive sectors is particularly concerning due to the rapid technology demand created by compliance and regulatory needs.

The increase in energy demand is attributed to the sheer volume of data required to train models, the billions of parameters in each AI engine, and the vast computing infrastructure involved. As Gen AI models evolve, millions of parameters have quickly grown to billions and even trillions, as seen in advanced LLMs deployed by companies such as OpenAI and Google. This heavy reliance on complex algorithms and real-time data processing for even simple tasks will strain the IT infrastructure, impacting sustainability and accelerating energy consumption.

Optimizing IT infrastructure with AI

The existing global IT infrastructure is not fully optimized for Gen AI’s computational needs. According to CBRE[vi], one of the primary challenges is the inefficient utilization of existing data centers and cloud resources. In many cases, infrastructure is designed for peak loads but remains underutilized during off-peak hours, leading to wasted energy and unnecessary costs.

This is where Gen AI can help improve itself. Using AI-driven tools to optimize load balancing, financial services can dynamically allocate resources based on demand patterns. AI-enabled infrastructure monitoring can reduce energy consumption by 20% to 30% by predicting usage patterns and automating resource adjustments in real time. AI can analyze historical data to predict future workloads, allowing infrastructure to operate more efficiently.

Furthermore, generative AI can optimize data storage, retrieval, and processing. By identifying patterns in data usage and resource allocation, AI can recommend ways to consolidate workloads, reduce redundancy, and minimize idle capacity—thereby cutting energy costs and improving sustainability.

Sustainable approaches to AI deployment

Gartner predicts that 30% of Gen AI projects will be abandoned after the Proof-of-Concept (PoC) stage by the end of 2025 due to inadequate risk controls, escalating costs, poor data quality, or unclear business value. Yet, we are seeing businesses race to create new PoCs. Given the environmental cost of AI, financial institutions must adopt sustainable practices in their AI development and deployment strategies. The “Reduce, Recycle, Reuse” framework offers a practical approach to minimizing AI’s environmental impact.

• Reduce: Instead of relying on massive models that require extensive computational power, AI in financial services can use smaller, more efficient models for specific tasks. Companies like Apple, Facebook, and Google are already experimenting with tiny AI or lightweight AI models, which leverage smaller models without compromising accuracy. This approach reduces the energy required for training and inference, aligning AI deployments with environmental goals.
• Reuse: Leveraging reusable components like prompt libraries, pre-trained models, and caching mechanisms can significantly reduce computational overhead. Instead of developing AI systems from the ground up each time, companies can build on pre-existing frameworks, reducing energy consumption and development time.
• Recycle: Financial services firms can repurpose existing AI models and data, thereby avoiding the need to retrain models from scratch. This not only conserves energy but also reduces costs associated with computational resources. Moreover, designing AI to maximize existing data through intelligent data recycling techniques will help repurpose historical models to adapt to new challenges.

The circular economy and AI in financial services

The concept of a circular economy—where what you consume is balanced by what you give back—offers valuable lessons for AI deployment in financial services. This model encourages companies to consume only what they can replenish, reducing waste and fostering sustainable growth.

For AI, the circular economy principle can be applied in several ways, including:

• Energy efficiency: Financial services firms must prioritize AI systems that minimize energy use. This includes adopting energy-efficient processors and optimizing data centers for green energy sources. Leading data center providers, such as Google and Microsoft, expect to source and match 100% zero-carbon electricity by 2030.
• Responsible consumption: AI architects must make mindful decisions when developing models. For example, not every task requires a full-scale LLM. Financial institutions should evaluate whether a simpler model could achieve the same results with lower energy consumption. In fact, most AI models used in financial services could be fine-tuned or pruned to reduce computational load without affecting performance.

Financial firms must also consider the environmental cost of external cloud services and reduce their environmental impact by adopting a cloud-native approach emphasizing energy efficiency.

Regulatory gaps and opportunities

While much of the regulatory focus in Gen AI is on ethical use, privacy, and security, there’s a growing recognition of the need for environmental accountability. The current AI regulations are largely silent on energy consumption and carbon emissions. Governments and regulatory bodies have yet to mandate that companies disclose the environmental impact of their AI deployments.

In financial services, where regulatory oversight is already stringent, including sustainability metrics in AI strategies could offer a competitive advantage. By voluntarily reporting AI energy consumption and carbon emissions, financial firms could position themselves as leaders in the sustainability space. This could also mitigate the reputational risk associated with high-energy AI systems.

At LTIMindtree, we implement the reduce, reuse, recycle framework to ensure our AI deployments are as energy efficient as possible. We also apply the Software Carbon Intensity (SCI) metric to measure the carbon footprint of our software solutions.

I recommend that financial services firms integrate environmental sustainability into their AI governance frameworks. This would involve tracking energy consumption, calculating AI-related carbon footprints, and implementing measures to reduce the environmental impact of generative AI.

Conclusion

In conclusion, as leaders navigate the evolving landscape of generative AI, it’s crucial to focus not just on innovation but also on sustainability. The increasing energy demands of AI systems highlight an urgent need for a comprehensive sustainability index that measures the environmental impact of AI alongside its business value. This index should serve as a guiding framework, encouraging organizations to prioritize eco-friendly practices in their AI initiatives. By embracing this holistic approach, businesses can unlock the true potential of AI—not only as a driver of efficiency but as a catalyst for sustainable growth.

The responsibility lies with leaders like you to champion this change. Investing in sustainable AI practices not only benefits the planet but can also enhance your organization’s reputation, improve consumer trust, and create long-term resilience. Today, let us commit to an AI future that marries technological advancement with environmental responsibilities. Together, we can create a more sustainable and prosperous tomorrow, where we realize AI’s benefits without damaging our planet. The time for action is now—let’s innovate sustainably.

To know more about LTIMindtree’s sustainable banking initiatives, please click here.

References

• Electricity 2024: Forecast to 2026, International Energy Agency,
• GS SUSTAIN: Generational Growth — AI/data centers’ global power surge and the sustainability impact, Goldman Sachs, April 30, 2024,
• Sustainability-related disclosure in the financial services sector, European Commission,
• Carbon Emissions and Large Neural Network Training, Cornell University,
• The state of AI in 2023: Generative AI’s breakout year, Quantum Black,
• The economic potential of generative AI: The next productivity frontier, McKinsey, June 14, 2023,
• Global Data Center Trends 2024, CBRE, June 24, 2024,
• Universal workflow of artificial intelligence for energy saving, Science Direct,
• Gartner Predicts 30% of Generative AI Projects Will Be Abandoned After Proof of Concept By End of 2025, Gartner,
• Data Centres and Data Transmission Networks, International Energy Agency