Artificial Intelligence (AI) is changing the world at record speed, but it’s also consuming its energy resources at an alarming rate. The rapid rise of generative AI has triggered an explosion in demand for computational power, pushing data centres to their limits. Estimations from Statistia highlighted that AI-related operations accounted for around eight percent of global data centre electricity use in 2023. Goldman Sachs Research estimates that data centre power demand will grow 160% by 2030, with AI expected to represent about 19% of data centre power demand by 2028.
The cost isn’t just financial; it’s environmental. But what if AI didn’t have to come with such a massive energy bill? What if the way we manage data could make AI more sustainable?
Deepseek has already proven that AI can be built differently. Instead of relying on ever-growing computational resources, Deepseek optimised data processing to achieve efficiency without excess. According to DeepSeek’s own research papers, their servers consume 50% to 75% less energy than Nvidia’s latest GPU units. This shift in mindset highlights an emerging truth: it’s not just the volume of data that determines AI’s impact—it’s how that data is managed.
The hidden energy cost of batch processing
Most AI systems still rely on data fed through batch processing, a decades-old method where data is collected, stored, and processed at scheduled intervals. This approach is costly, slow, and inefficient. Since batch processing efficiency is often measured by data volume and time windows, longer processing times mean higher energy consumption. A study on web search engines has shown that longer query processing times lead to higher power usage. By the time businesses extract insights, much of the data is already outdated, and power-hungry data centres, where batch processing often occurs, are working overtime to process unnecessary information. Storage demands skyrocket, and companies lock themselves into a cycle of waste.
Batch processing is like driving in stop-and-go traffic - every restart burns more fuel than simply maintaining a steady speed. AI running on batch processing wastes energy in the same way, forcing data centres to crunch numbers in massive, inefficient bursts. The environmental impact of this approach is creeping up, especially in Asia, where the rapid rise of AI is putting immense pressure on already strained power grids.
See also: How AI-powered coding is revolutionising software development
The real-time revolution
A more sustainable approach is already transforming AI infrastructure. Real-time data streaming processes information as it arrives, eliminating unnecessary storage, reducing latency, and cutting down on the sheer computational power required to keep AI running efficiently. This shift—what Confluent identifies as ‘data in motion’—allows AI to continuously process data at all times, rather than hoarding and processing vast amounts of excess data at scheduled intervals.
Governments across Asia are already recognising the need for AI to be greener. Singapore’s Green Plan 2030 is pushing for more energy-efficient data centres, China’s Next Generation AI Development Plan emphasises balancing technological growth with sustainability, and Taiwan is exploring nuclear power to sustain AI-driven industries. Businesses looking to stay ahead of regulations—and their own sustainability commitments—will need to rethink how they handle data.
See also: Embracing the cloud-native and AI transformation journey
Sustainability and AI can coexist
Real-time data streaming isn’t just a theoretical fix—it’s already driving sustainability across industries. Australia’s Powerledger uses a data streaming platform to power peer-to-peer energy trading over blockchain, optimising the way solar energy is distributed. By processing energy transactions in real-time, Powerledger reduces reliance on traditional grids and maximises renewable energy efficiency.
Other businesses are seeing similar benefits, using real-time data streaming to cut costs, improve AI responsiveness, and slash energy use. Reworkd leverages generative AI and a data streaming platform to automate and scale real-time web scraping. The approach has transformed a traditionally manual, time-intensive process into an efficient, fault-tolerant system.
SunPower, a renewable energy company, uses a real-time monitoring system to empower customers to optimise their energy usage and reduce waste. It also makes managing solar energy systems easier, supporting smarter and more sustainable choices for both consumers and the company. These use cases showcase how companies are turning AI into a tool for both efficiency and sustainability.
Rethinking AI’s future
AI’s energy demands are becoming unsustainable, but businesses don’t have to choose between innovation and responsibility. Real-time data streaming offers a way forward, reducing AI's environmental impact while making it more powerful, responsive, and efficient. The transition from batch processing to data in motion isn’t just an upgrade - it’s the only way for AI's future.
Companies that fail to adapt will face growing pressure from regulators and consumers who demand greener solutions. Those who embrace real-time data streaming will lead the charge toward a smarter, more sustainable AI revolution. The technology is available. The moment is now. It’s time to move AI forward - without leaving the planet behind.
Suvig Sharma is the regional head of Asia at Confluent
