For AI to drive a green revolution, innovations in computing efficiency and clean energy integration are crucial. This article explores next-gen AI hardware, energy-aware algorithms, and policy strategies that can align AI’s growth with sustainability.
Pathways to sustainable AI: Tech innovations and policy responses
For AI to truly spark a green revolution, innovation must focus on making computing more efficient and integrating AI growth with clean energy systems. This involves advances in hardware and software, as well as smart policies, to nudge the industry in the right direction.
Technological levers for efficient AI
- Next-gen AI chips (ASICs and photonics)
Traditional CPU/GPU architectures are not very energy-efficient for AI workloads. Enter specialised AI accelerators. Companies like Lightmatter are developing photonic (light-based) chips that perform AI computations using photons instead of electrons, massively reducing energy loss as heat.
Lightmatter’s chip reportedly achieves 9 petaflops per watt of performance — orders of magnitude beyond conventional silicon. If such optical computing scales up, future AI models could run on a fraction of the energy today’s do. Similarly, Google’s TPUs and various startups’ AI ASICs are tuned for maximum throughput per watt, offering 2 to 5× improvements over general GPUs.
Inspired by the human brain, neuromorphic chips (like Intel’s Loihi 3) use networks of “spiking” neurons that are extremely low-power. They excel at tasks like pattern recognition with minimal energy. Intel reports up to 76 per cent lower energy for LLM inference with neuromorphic approaches on some workloads. While still experimental, these could allow AI systems that learn and operate continuously on tiny power budgets — think AI co-processors that sip power like a LED lightbulb.
- Algorithmic efficiency (better software)
On the software side, there’s a push for efficient AI algorithms — for example, techniques like model pruning, quantisation, and knowledge distillation, which create smaller models that run faster. A pruned or distilled model can often achieve 90 per cent of the accuracy of a large model with, say, 50 per cent less computation required.
OpenAI and others are actively researching ways to maintain capability while cutting out the “waste” in neural networks. In training, new optimisation methods and architectures (like sparsely activated models) promise to reduce the compute needed to reach the same accuracy. These advances directly translate to energy saved.
Software is also helping schedule computing tasks at times and places where energy is clean. For instance, Microsoft Azure’s carbon-aware workload scheduling now shifts nearly 40 per cent of AI jobs to regions or times where renewable energy is abundant. If the wind is blowing in one data centre region, Azure will queue more AI jobs there, and pause or move jobs from another region that’s on fossil power at that moment. This kind of intelligent orchestration can significantly cut the effective carbon footprint of AI computations.
- Energy-proportional computing and PUE improvements
Data centre engineers continue to drive down overhead so that almost every watt goes to computing, not waste. Average Power Usage Effectiveness (PUE) has improved (some hyper-scale centres are at a PUE of 1.1 or lower, meaning 90+ per cent of energy powers IT equipment).
Techniques like better airflow management, AI-controlled cooling (as discussed), and even waste heat reuse (heating nearby buildings with server heat) all contribute. The closer we get to a PUE of 1.0 and fully utilised servers, the more work (AI tasks) we can get done per unit of energy input.
Also Read: 5 reasons why energy management is key to individual and organisational success
Policy interventions
Governments can guide the AI-energy trajectory with targeted policies and standards:
- Energy Efficiency Standards for AI Models
Just as there are fuel economy standards for cars, we may see efficiency standards for AI. The EU’s contemplated rule requiring 15 per cent energy efficiency improvement in new AI models is a first step. If major markets adopt similar rules or require reporting of AI energy use, it creates a competitive incentive to design greener AI. Transparency is key — imagine an “Energy Star” rating for AI services, where customers could choose a provider that is more energy-efficient.
- Carbon-adjusted pricing and credits
Some regions are introducing tariffs or credits to encourage clean energy usage. For example, California and Bavaria (Germany) have floated the idea of carbon-adjusted power purchase agreements that penalise data centres drawing power from grids below a certain renewable percentage.
Under such schemes, if an AI facility isn’t using (or contracting for) at least 80 per cent clean power, it would pay a surcharge or face limits. This kind of policy pushes companies to invest in renewables or locate where clean power is available, to avoid financial penalties.
- Dynamic electricity pricing
Grid operators like PJM in the US are implementing real-time pricing to manage peaks. PJM’s dynamic tariffs have encouraged data centres in its region (e.g., northern Virginia) to reduce peak load by 19 per cent — they respond to price spikes (often corresponding to dirty peaker plants coming online) by dialling down non-urgent workloads. Wider use of dynamic pricing will reward AI operations that can flex around grid conditions, effectively incentivising them to be more grid-friendly and efficient.
- Accelerating clean energy permitting
One practical bottleneck for sustainable AI power is the slow permitting of new renewable projects and transmission lines. Policymakers can streamline this — for instance, the US Nuclear Regulatory Commission is fast-tracking approvals of advanced reactor designs aiming to have a set of SMRs approved by 2026, specifically with data centre use cases in mind.
Governments can also designate “energy corridors” for easier building of high-voltage lines to data centre regions, or provide grants for battery storage at data hubs. All these reduce the risk that AI’s growth outstrips green energy availability.
Supporting the development of the above-mentioned technologies (optical computing, neuromorphic, etc.) through grants and public-private partnerships can speed their arrival. Given AI’s strategic importance, one can envision national programs to develop next-gen low-power AI hardware (the way there were initiatives for supercomputing in past decades). This not only helps the climate but also ensures a country’s AI industry remains globally competitive as efficiency becomes a differentiator.
The big picture is that a combination of technology innovation and forward-thinking policy can bend the trajectory of AI’s energy impact. It’s analogous to the auto industry — without better tech (EVs, hybrids) and policies (fuel standards, incentives), car emissions would have kept rising unabated. With them, it’s possible to have the benefits of mobility (or in our case, AI capabilities) while mitigating the harms.
For corporate leaders, staying ahead on these fronts means:
- Monitoring and adopting emerging efficient AI tech — perhaps experimenting with new accelerators or AI model optimisations that cut costs and footprint.
- Engaging with policymakers or industry groups to help shape sensible standards (it’s better to help craft the rules than be caught off-guard by them).
- Committing to transparency in AI energy use and emissions. Some leading companies already publish the PUE and carbon data of their data centres; extending this culture to AI operations builds trust and prepares the company for a future where stakeholders demand to know the climate impact of AI initiatives.
Next, we turn these insights into a concrete action plan for executives — what steps to take to ride the AI wave without capsising under energy costs or sustainability risks.
A tactical AI-energy strategy for corporate leaders
How can corporate decision-makers apply these insights in practice? Here we distill a practical guide — key questions to ask, and steps to take — to balance AI’s opportunities with energy and sustainability considerations.
Also Read: Why the future of space and energy storage might be growing in a Thai hemp farm
Five key questions every CEO should ask about AI and energy
- How much energy do our AI operations consume? — Get a handle on the current state. Measure the power usage of your AI workloads (on-premise and in cloud). Understand the scale: is it five per cent of your IT energy use? 50 per cent? Quantify it in kWh and dollars, so you have a baseline. Also project how this might grow with planned projects (if you adopt a new AI tool, will it double your compute hours? Triple?). You can’t manage what you don’t measure.
- Are we using the most energy-efficient AI models and infrastructure available? — Audit your AI stack. Are there opportunities to use smaller models, or algorithm optimisations like batching and quantisation to cut compute? Are you running on last-gen hardware out of habit, when new AI chips could do the job with 1/2 the energy? Push your tech teams and vendors to justify choices in terms of efficiency, not just accuracy or speed.
- Are we leveraging AI to optimise our own energy use? — This flips the script: use AI as part of the solution. Could AI tools help reduce energy waste in your operations (factories, offices, supply chain)? For example, using AI for route optimisation in logistics to save fuel, or for energy management in buildings (as some have done to cut HVAC costs by 15–30 per cent). Ensure your sustainability and facilities teams are exploring AI solutions — the ROI can be significant, and it creates a positive offset for the energy your AI projects consume.
- Are we investing in clean-energy-powered cloud services (or data centres)? — When choosing where to run AI workloads, factor in the energy source. Major cloud providers now offer regions or options powered by 100 per cent renewable energy — utilising those can drastically cut the carbon footprint. If you run your own data centre, consider power purchase agreements for renewables or even on-site solar. Essentially, align your digital infrastructure with your renewable energy procurement.
- Are we prepared for potential AI energy regulations? — Scan the horizon for laws that might affect your AI deployments. For instance, if efficiency standards for AI or reporting requirements come in a year or two, do you have the data to comply? If carbon pricing rises, do you know which AI projects would become more expensive to run? Engaging with industry groups and regulators proactively can give you a voice and early insight. Internally, scenario-plan for a future where “green AI” might be mandated either by law or by customers/investors.
Asking these questions at the C-suite level ensures that AI initiatives are not happening in a silo, but are integrated with energy management and corporate strategy.
Practical steps for sustainable AI adoption
Conduct an AI energy audit
Much like financial auditing, do an energy audit for AI. Map out all AI-related compute (data centres, cloud usage, edge devices) and tally the power usage. Identify hotspots — e.g., a particular analytics cluster or training workflow that draws a lot of power. This audit gives you a clear picture of where to target efficiency efforts.
It might reveal, for example, that 20 per cent of your AI jobs account for 80 per cent of the energy — maybe heavy model training that could be scheduled during off-peak hours or moved to a more efficient cloud zone.
Optimise and right-size AI workloads
Use the findings to implement quick wins:
- Model right-sizing: Where possible, replace giant models with smaller ones or use transfer learning to avoid training from scratch. If a 500-million parameter model can solve the problem, don’t use a 50-billion one. This can cut computation dramatically.
- Lifecycle management: Not all AI tasks need to run at highest frequency. Determine which jobs are mission-critical vs. which can be throttled or delayed in high load times. Leverage cloud auto-scaling to shut down idle resources (many companies find servers running when not needed — a pure waste).
- Use AI to tune AI: It’s meta, but you can apply AI to improve scheduling and resource allocation for your AI jobs (similar to how DeepMind’s system works for Google). This can maximise utilisation and reduce idle energy burn.
Leverage AI for broader energy management
As noted, deploy AI solutions in your operations to save energy and costs. For example:
- Implement an AI-based energy management system in corporate offices or factories (many vendors offer these).
- Use machine learning to analyse production line data for energy inefficiencies (maybe a certain machine uses more power than it should — predictive maintenance can fix that).
- Optimise logistics and travel with AI to reduce fuel use. Every kilowatt-hour or gallon saved here helps offset the extra energy your data centres might consume. And they directly save money, improving the business case for AI investments.
Adopt hybrid computing strategies
Not all workloads must run in power-hungry central clouds. Consider a hybrid AI approach: run smaller, latency-sensitive tasks on energy-efficient edge devices (or on end-user devices), and reserve big cloud compute for the truly heavy tasks. By using edge AI (which has no network transit and can be highly optimised), you reduce total energy per inference.
Also explore techniques like model distillation to create lighter versions of cloud models that can run on-premises or on cheaper hardware when appropriate. This hybrid mindset ensures you’re not always using a sledgehammer (huge cloud instance) for a nail (simple task).
Also Read: On the precipice of energy transition
Prioritise green cloud providers and contracts
When negotiating with cloud or data centre vendors, make sustainability a key criterion. Ask providers about their PUE, their renewable energy percentage, and their roadmap for low-carbon operations. Some cloud providers now offer dashboards showing the carbon emissions of your cloud usage — use those insights.
If you operate your own facilities, sign renewable energy contracts (PPAs) to cover your AI electricity use with clean energy. Also, work with utilities on programs (many utilities have “green tariffs” or will help with renewable projects if you’re a large load). Align your procurement so that as your AI energy use grows, your renewable supply grows in step.
Collaborate with industry and policymakers
Given the broader grid challenges, it’s wise for companies running big AI workloads to have a seat at the table. Join industry consortia focused on sustainable data centres or AI ethics that include energy impact. Engage local governments if you’re building data facilities — perhaps partner on community solar/storage so the investment benefits both you and the grid.
Being proactive can also help shape favorable policies (for instance, incentives for using local clean power or faster permitting for your backup generators etc.). Don’t wait to be caught by surprise regulations; help shape the narrative that AI can be part of the climate solution.
Scenario planning and risk mitigation
Finally, include energy security in your risk assessments for AI. Ask “what if” questions: What if power is constrained in Region X — do we have failover in a different region? What if electricity prices spike 3× — does our AI project still make economic sense, and can we hedge that risk?
Have backup plans for critical AI services if rolling blackouts or energy rationing ever hit (not unthinkable in some grids). By planning for these contingencies, you ensure AI deployments are resilient and won’t be derailed by external energy shocks.
By taking these steps, executives can balance efficiency, cost, and sustainability in their AI adoption. The companies that follow this playbook will likely have a smoother ride scaling AI — with lower bills and stronger ESG credentials — than those who treat energy as an afterthought.
Conclusion: A contested energy future
AI’s rise presents both a monumental challenge and an opportunity for the energy landscape.
On one hand, AI’s energy demands are forcing a reckoning: power grids are under strain, carbon goals are at risk, and companies may face tough trade-offs or regulatory hurdles if they ignore the issue. On the other hand, AI offers unprecedented tools to drive efficiency, optimise energy systems, and accelerate the transition to cleaner power.
For corporate leaders, the takeaway is clear: the future will belong to those who integrate AI and energy strategy. The organisations that treat energy as a core element of their AI plans — investing in efficiency, securing sustainable power, innovating with AI in their operations — will lead the pack.
They’ll enjoy more reliable growth (because they won’t hit energy ceilings), better public trust, and likely cost advantages as well. Those that ignore the linkage may find themselves facing energy supply crises, skyrocketing costs, or regulatory roadblocks that stall their AI ambitions.
The choice isn’t whether to adopt AI — that wave is here and necessary to remain competitive. The choice is how to do so responsibly and strategically. Companies that can harness AI and champion sustainability will shape the narrative of the coming decades. They’ll prove that innovation and green objectives can reinforce each other, not collide.
In the end, will your company spark the AI energy revolution, or be caught flat-footed by it?
By asking the hard questions now and taking decisive action, you can ensure that AI becomes a driver of efficiency and positive change — a win-win for your business and the planet, rather than a zero-sum trade-off. The green energy revolution and the AI revolution can be two sides of the same coin, but it will take foresight and leadership to make that vision a reality.
Will your company shape the AI-energy future — or be shaped by it? The decisions made today will determine the answer. The opportunity is to lead boldly, invest wisely, and create an AI-powered future that is sustainable, secure, and full of possibility for generations to come.
Thanks for reading!
What do you think?
This is part three of a three-part series exploring AI’s energy impact. Read part one here and part two here.
This article was originally published here and co-authored by Xavier Greco, Founder and CEO of ENSSO.
—
Editor’s note: e27 aims to foster thought leadership by publishing views from the community. Share your opinion by submitting an article, video, podcast, or infographic.
Join us on Instagram, Facebook, X, and LinkedIn to stay connected.
Image courtesy: DALL-E
The post The AI-energy paradox: Will AI spark a green energy revolution or deepen the global energy crisis? — Part 3 appeared first on e27.
Begin with a pilot → Test with one team
Integrate smartly → Connect AI with your existing systems (ERP, CRM, HRMS)
Refine continuously → Use KPI-driven feedback to improve performance
Recognise AI Champions in your teams and incentivise adoption!
A strong team with a less-than-perfect model can still create significant investment value.
On the other hand, even the most well-crafted business model will fail if the team lacks the ability to execute.
CTOs to handle a larger scope of coding tasks with fewer engineers.
CEOs to manage marketing, pitch decks, product design, and proof-of-concept (PoC) validation independently.
AI-enhanced businesses → Companies that add AI to existing products but haven’t fundamentally changed their approach. These businesses often struggle to prove AI’s core value to their model.
A willingness to learn and adapt is more crucial than ever.