Oct 31, 2025

ZeeOctober 31, 2025 AI may be one of the main forces behind modern progress for the tech world, but it is also becoming one of the hungriest users of electricity and water. NTT DATA’s new white paper, Sustainable AI for a Greener Tomorrow, says AI workloads could make up over 50% of all data centre power use by 2028. Cooling systems also demand huge amounts of water, while producing hardware depends on rare earth minerals that leave behind pollution and waste. David Costa, who leads Sustainability Innovation at NTT DATA, said that AI’s fast expansion has serious environmental consequences. But he added that the same technology could also manage power grids, model climate risks and improve water use if it is built with efficiency at its core. The report says sustainability needs to be part of AI design from the very start, not added later. That means tracking electricity, carbon and water use with shared tools such as the AI Energy Score and the Software Carbon Intensity system. The paper also calls for “lifecycle thinking”, which is designing hardware to last longer, using recyclable parts and reducing e-waste through refurbishment. NTT DATA says the responsibility cannot sit with one group. Hardware makers, data centre operators, cloud providers and users all need to rethink how AI systems are built and maintained. Running AI tasks when renewable energy is available, refurbishing old equipment and using modular components are all small steps that could make AI cleaner and less wasteful. Costa said that sustainability should no longer compete with performance targets. “Efficiency must be treated as a design principle,” he said, “because technology cannot thrive on a damaged planet.” More experts have shared how startups in the uK in particular can ensure sustainable AI… “UK startups have a major problem with operational sustainability. Too many brilliant ideas are buried under procedural clutter. Most teams adopt AI reactively, chasing quick wins without thinking about integration or long-term business growth. The result is fragmented systems, duplicated costs, and no clear understanding of where data flows or what’s actually driving value. Sustainable AI adoption isn’t about using less tech, it’s about using it wisely, with clarity, accountability, and intention. Here’s how founders can set themselves up for success. “Map out all of the software tools in use across your organisation – then regularly review your AI stack to remove duplication and complexity. A small number of well-integrated tools will almost always outperform a pile of disconnected experiments. “Know your purpose. Every AI tool should have a clear reason to exist: team members should know exactly what problem it solves and what outcome you expect. If you can’t define that in a sentence, it’s probably not worth adopting. “Give each AI system a clear owner responsible for maintaining it, measuring impact, and ensuring it integrates properly with the rest of your workflow. “Build for the long term. Treat AI as part of your operational backbone, not a side project. Sustainable growth comes from continuous iteration, not from chasing the newest tool.” “Sustainable AI use for UK start-ups is not about your carbon footprint; it’s about discipline and integration. True sustainability is achieved when building tools and tech solutions that are transparent, data-efficient and that actually solve a legitimate problem. We see too many start-ups bolt-on AI as a shiny new feature to ensure they aren’t left behind but fail to make it part of their core infrastructure; this is unsustainable. Bolt-on’s are easily recognisable as unsustainable AI practice. “To benefit from AI properly, it should be crucial to the success of their business, not an optional extra. Sustainable AI use will be seen in companies using it to truly improve their business, by increasing automation, improving accuracy and accelerating operations, while also reducing waste. We see it in more complex markets, like global trade or B2B procurement, where AI becomes essential precisely because of the value it creates, improving efficiency, trust and long-term resilience.” “At JoinDig, we see first-hand how AI can transform consumption. By connecting shoppers to both new and pre-loved products, we extend product lifecycles and reduce the carbon footprint of manufacturing. “Sustainable choices should be convenient. By using machine learning to aggregate and analyse millions of listings, AI makes it easier for consumers to find the best-value, eco-friendly options instantly. “UK startups like ours are proving that AI can do more than optimise ads. It can underpin smarter, circular economies where conscious shopping becomes the default. “AI is helping to remove the barriers eco-conscious shoppers face: lack of transparency, time-consuming searches, and uncertainty about quality or authenticity. “By embedding sustainability into technology, UK startups can show that innovation and environmental responsibility are not mutually exclusive; they’re complimentary. “The next era of commerce will be defined by ‘conscious convenience,’ where sustainable choices are as easy, affordable, and attractive as unsustainable ones, and AI is central to that shift . “For businesses, insights from AI-driven resale platforms can inform circular supply chain strategies, helping brands design products that retain value and can be efficiently reintroduced into the market. “With AI as an ally, the future of shopping can finally align with the future of our planet.” “GenAI is here to stay, but UK start-ups can cut their footprints without sacrificing speed. They should begin by auditing their AI use and identifying where models are embedded, who is using them, and how often. Start-ups should also ask tough questions of their AI providers: where the models are hosted, how energy is sourced, and what transparency exists around emissions. “Next, it’s about choosing the right model for the task. Large language models (LLMs) are not always the answer. Smaller, domain-specific models are faster, more affordable, and far less energy-intensive, especially if you are training the models on your own data and intelligence rather than asking an agent to scan the whole internet for knowledge for each query For more advanced applications, improving prompt efficiency and reducing unnecessary processing can significantly cut emissions. “Finally, sustainability should be treated as a performance metric alongside speed and cost. When teams see CO₂ per request fall without compromising quality, behaviour changes naturally. Using AI more sustainably shouldn’t limit innovation – it should make progress measurable, proving that smarter, more efficient systems can drive both productivity and sustainability.” “It’s complicated, because “sustainable AI” almost feels like an oxymoron. The thing itself eats energy. But for start-ups, I think it starts with being deliberate, not just doing what’s fastest or cheapest. If you never ask where the power’s coming from, you’re part of the problem. There are greener options now (e.g. Carbon3.ai) but you have to care enough to look. “What we need is restraint. Everyone wants to build the biggest, flashiest model, but most of the time you don’t need it. A smaller model that does one thing well is better than a bloated one that needs a warehouse of GPUs to keep it alive. I think we’ve started to confuse scale with brilliance, and that’s a mistake. “Measurement is also important. Hardly anyone tracks the energy their systems use, they just assume it’s fine. But once you start actually measuring it, you can’t un-see it. It changes how you build. “Maybe what’s more hopeful is using AI to fix the things it breaks. There are lots of start-ups using it to cut energy waste, predict renewable output, automate sustainability reporting. That feels like a fair trade-off, using the tool to offset its own footprint.” “At Photoroom, we’ve always believed efficient AI is better AI. Smaller architectures, faster inference, less wasted compute – that focus pushes you to build smarter. It’s how we scale powerful visual tools without the environmental or financial overhead of heavier models. “For AI start-ups, building sustainably isn’t a constraint: it’s a win for performance, cost, and the planet. Every technical choice – your model size, data source, and cloud provider – compounds over time, shaping not only your carbon footprint and cost base but also the user experience. Leaner systems run faster, feel more responsive, and are easier to maintain as you grow. Transparency also matters. Sharing how your models work, what data trains them, and what their real impact is builds trust – something every company in AI will need for the long term.” “As AI becomes central to business operations, sustainability should be part of its foundation, not an afterthought. Companies can start by selecting energy-efficient infrastructure. Hosting AI workloads on servers powered by renewable energy or in data centres that repurpose excess heat can significantly reduce carbon impact. “Next, they should prioritise model efficiency over scale. Instead of training massive models from scratch, organisations can fine-tune existing open-source models for their specific needs. This approach reduces computational and environmental costs while improving transparency. “Data governance also plays a key role. Using curated, bias-monitored datasets prevents wasteful retraining cycles and supports fairness, which is an important aspect of social sustainability. “Companies should measure and disclose their AI footprint, including energy usage and emissions, just as they would for any ESG metric. Transparency drives accountability and encourages innovation toward greener practices. “Finally, sustainability is cultural as much as technical. Embedding environmental and ethical considerations into AI governance frameworks ensures long-term alignment with both corporate and societal goals. “Responsible, sustainable AI is not only good for the planet but also a strategic advantage for organisations that want to innovate with integrity.” “Through our La Fosse Academy, which helps organisations upskill their people for the digital age, we’ve seen that sustainable AI use starts with people, not platforms. The real differentiator isn’t which tool a start-up adopts – it’s whether teams have the capability to use AI responsibly. That means understanding how to question outputs, spotting bias, and identifying potential risks before they escalate. “AI is developing faster than most organisations’ skill frameworks, so continuous learning and adaptability need to be built into the culture. Start-ups that focus on building AI literacy and confidence across their teams, rather than chasing efficiency or short-term wins, tend to achieve far more lasting and responsible outcomes. “Ultimately, embedding AI responsibly is as much a cultural challenge as it is a technological one. Creating an environment where people feel supported, equipped, and empowered to use AI thoughtfully encourages innovation while minimising risk. Sustainable AI isn’t about the tools you use, it’s about the skills, awareness, and confidence you build across your organisation to ensure it has a positive impact for the long term.” “UK start-ups are getting smarter with how they build and train models, focusing on approaches that cut energy use without compromising accuracy. Techniques like model pruning and quantisation help shrink models so they use less power and memory, which directly lowers computational costs. “Another prominent change we’re seeing is in data handling. By streamlining data pipelines and removing redundancies early, start-ups avoid unnecessary processing. And that’s something that often drains the most energy. Cloud platforms with renewable energy commitments are also becoming a standard part of tech stacks, letting smaller firms scale responsibly. “The practical side of sustainability comes down to habits. For instance, clean coding, regular performance audits, and tracking metrics like carbon intensity per model run. Those details add up. Investors, regulators, and clients increasingly expect visibility on these factors, so building efficiency into your AI systems from day one isn’t just environmentally responsible anymore, it actually makes good business sense.” “Sustainable AI use starts with a clear sense of purpose. Most start-ups rush into using large models because they can, not because they should. The most responsible approach is to design AI systems to address a real business need while using data efficiently and avoiding unnecessary computational load. The goal is to come up with scalable and reliable systems that can be easily maintained. “From a technical perspective, focusing on lightweight architectures, optimized cloud deployment, and selective fine-tuning can reduce the use of resources without losing capability. Sustainability also involves transparently sourcing data, testing models responsibly and managing bias. Far from being a constraint, sustainable AI should be seen as an advantage for UK start-ups. It controls costs and nurtures long-term innovation while cementing the trust that customers need before wholeheartedly embracing AI-driven products.” “Regardless of personal opinions, Sam Altman was right about one thing: there’s never been a better time to start a company. AI has torn down what used to be barriers to entry for entrepreneurs and innovators; what once required an entire workforce to operate can now be done by a single individual focused on precision and speed of output. “This democratisation is the climate opportunity. If all emerging AI companies were focused not only on valuation but on carbon removal, the combined global effect would be transformational. Yes, AI has a significant energy bill, but its potential is far more substantial. If every watt spent training a language model is put towards finding solutions to offset climate change , AI’s footprint becomes a down payment on a cleaner future. “AI is arguably the most intelligent thing humanity has ever developed; it only makes sense to put that intelligence towards solving humanity’s biggest problem.” “A challenge many fast-growing startups face is balancing rapid innovation with the dual demands of being secure and compliant. As new regulations emerge across regions and sectors, startups often struggle to scale while maintaining consistent data protection and governance standards. “Security and compliance shouldn’t be seen as obstacles to growth but as foundations for it. Embedding automation and continuous monitoring into cloud infrastructure allows young companies to build secure environments from day one, reducing human error and ensuring regulatory readiness as they expand. “For startups, the goal isn’t simply to tick compliance boxes, but to create resilient systems that can adapt as threats and rules evolve. By integrating security into development and operational workflows, teams can innovate confidently, knowing they’re protecting customers, data, and reputation in equal measure. Ultimately, sustainable growth in the digital economy depends on treating compliance and security as inseparable parts of the same mission.” Related Articles