Photonics seen as key to greener AI as energy use surges in Europe
A new report has identified photonics as the only viable solution to meet the rapidly growing energy demands driven by artificial intelligence (AI) without increasing carbon emissions. As AI models and data centres expand across Europe, concerns are mounting about the environmental and economic impact of soaring electricity consumption.
Energy strain
The report, released by Photonics21 and authored by TEMATYS, warns that electricity use by AI is outpacing the capacity of existing power grids. According to the International Energy Agency, global data centres consumed 415 TWh of electricity in 2024. TEMATYS projects this figure could more than double by 2030 if current growth rates continue, largely because of increased AI workloads.
Such levels of energy consumption risk undermining Europe's efforts to meet climate commitments. The study finds that current silicon-based semiconductor technologies are unable to keep up with the increasing need for computational power without heavy energy and environmental costs.
Photonics potential
Photonics, the use of light to transmit and process information, is presented in the study as a critical complement to existing chip technology. Integrating photonic elements into data centres and chip architectures-using methods such as co-packaged optics-could significantly cut energy use and emissions associated with AI computing.
The report highlights that while photonics will not replace CPUs or GPUs, it can alleviate the pressure on them, reducing the carbon and cost implications of AI expansion. It also states that the right policy and investment choices in the coming years will determine whether Europe can maintain control over its AI hardware supply chain and uphold its environmental objectives.
"Photonics can provide the infrastructure that will determine whether AI becomes cleaner and more competitive or simply costlier and dirtier. Europe has the research base to lead; what it lacks is coordinated investment and industrial scale," said Sébastien Bigo, Nokia Bell Labs Fellow and Photonics21 Work Group Leader for Digital Infrastructure.
Current role
The study points out that photonics is already integral to the digital economy. Fibre optic technology forms the backbone of internet and data-centre networks, and the next phase-integrating optical components within chip packages-is currently underway in the industry.
Despite industry claims of significant energy efficiency improvements, the report suggests that efficiency gains from today's technologies alone will not be enough to counteract the surge in power usage as AI's footprint grows. The study also describes progress in optical computing, noting lab-based advances and examples such as MIT's 2024 demonstration of fully photonic neural-network chips.
European landscape
Europe hosts world-class research groups and a growing cluster of start-ups in photonics. However, the report expresses concern that a lack of coordinated investment, large-scale manufacturing, and skilled workforce development threaten Europe's potential to play a leading role in the global market. Without stronger support, Europe could become reliant on overseas suppliers for key components of AI infrastructure.
In its recommendations, the study calls for European governments and investors to increase funding for pilot manufacturing, support the scaling up of photonics start-ups, prioritise photonics in major technology funding schemes, and invest in the skills required for production at scale.
"If Europe wants sovereignty in future AI hardware and to meet its climate commitments, Europe's policy and investment decisions in the coming years will be decisive," said Bigo.
