By Lieven Levrau, Senior Director of Product Strategy at Nokia and Vice Chair of the Technology Steering Committee at The IOWN Global Forum

Lieven Levrau

The rise of generative AI is reshaping industries and redefining the pace of technological advancement. Global adoption has surged – ChatGPT had 400 million weekly active users globally as of February 2025. However, such rapid growth comes with significant environmental implications.

The energy required to train large-scale AI models is substantial. A single AI-focused hyperscale data center can have a capacity of 100 MW – equivalent to the annual electricity consumption of 100,000 households. As investments in these facilities increases, so does AI’s environmental footprint. So, how can we sustain the momentum of AI innovation while ensuring environmental responsibility?

The infrastructure gap

Today’s AI systems depend on an infrastructure that is increasingly outdated. Many existing data centers lack the capacity to support the high-performance GPUs need for advanced AI workloads. Simply upgrading these facilities with new chips is often insufficient, as legacy systems cannot meet the computational demands of modern models.

Moreover, the expansion of AI-ready infrastructure faces regulatory and logistical challenges. In several markets, the construction of new data centers is constrained by energy, land use, or sustainability considerations. According to projections by McKinsey, meeting the anticipated global demand for compute capacity by 2030 could require up to $6.7 trillion in investment.

This poses a significant challenge for businesses. While large enterprises may have the capital to invest in advanced infrastructure, smaller organizations often face trade-offs – pursuing technological competitiveness at the risk of environmental and financial sustainability.

A shift toward photonic solutions

All-Photonics Networks (APNs) present a promising path forward. These systems replace traditional electronic data transmission with photonic technology, offering vastly improved energy efficiency and transmission speed. When deployed in conjunction with green data centers powered by renewable energy in more remote locations, APNs have the potential to support large-scale AI training, while significantly reducing environmental impact.

Remote GPU services allow organizations to offload intensive computational tasks to specialized, energy-efficient facilities. This enables compute data centers to be located further away while maintaining the same or similar latency budget, allowing them to be situated in regions with greener energy production and more efficient cooling. This model not only reduces local infrastructure demands but also cuts electricity usage, operating costs, and emissions. It represents a scalable solution to managing AI’s increasing resource requirements.

Enhancing access and efficiency

These innovations are not only environmentally beneficial – they also democratize access to AI capabilities. Currently, only well-resourced tech giants can afford the infrastructure necessary to train state-of-the-art models. Remote services and high-speed photonic networks will enable smaller enterprises and research institutions to compete, accelerating innovation across the board.

Additionally, photonic infrastructure significantly reduces training time. While traditional AI model development can span several months, APNs enable the same work to be completed within days, using a fraction of the energy. This enhanced efficiency aligns both business goals and sustainability imperatives.

Prioritizing security in a distributed ecosystem

As AI systems grow more sophisticated, so do the risks associated with data privacy and security. AI training often involves large volumes of sensitive information, necessitating advanced protections. Photonic infrastructure must be accompanied by robust safeguards such as end-to-end encryption, remote attestation, and zero-trust architecture to ensure data integrity and regulatory compliance.

Sustainable AI infrastructure must be secure by design. Environmental efficiency and operational scalability are vital, but they must not come at the expense of information security.

A strategic path toward sustainable AI

The evolution of APNs and remote AI services demonstrates that technological progress and environmental responsibility need not be mutually exclusive. By improving energy efficiency, reducing emissions, and broadening access to computational resources, these innovations offer a strategic pathway toward sustainable, inclusive AI.

The challenge is no longer whether sustainable AI is possible – but whether industry and government can move quickly enough to implement these changes at scale. The pace at which APNs and renewable-powered infrastructure are adopted will determine whether AI continues to thrive without overwhelming the planet’s resources.