Tag: energy crisis

We are at a critical point in time. Across the globe, data demand is growing while AI workloads are accelerating.

This trend has a remarkable impact. Data centers in 2022, used between 240 and 340 terawatt-hours (TWh) of electricity—which is about 1–1.3 percent of power demand from across the globe. Experts project that by 2026, this will grow exponentially to 945 TWh, about the same as the entire annual electricity consumption of Japan. Companies like Equinix are doubling their global footprint in just five years, matching the scale of their first 27 years of growth, to keep pace with demand.

The amount of power required to support this growth is not sustainable. Continuing to use only electron-based systems will create a global energy crisis—which will stretch data center infrastructure and hurt the planet. 

The industry is now facing a critical question: How can the world continue to scale computing power while avoiding an energy and environmental crisis?

I believe the solution to the data center energy crisis is photonics. We can create systems – by harnessing light – that need less power, demand less heat, and provide communication speeds that we could never imagine before. Programs such as NTT’s IOWN, Europe’s F5G program, Microsoft’s hollow-core fiber trials, and technology innovation from Ayar Labs, Lightmatter, and Celestial AI will bring us to a future when photonics are foundational, not experimental. 

How Electronics Create Heat

Relying on electrons is the fundamental problem. Networks and devices run based on electrons moving through circuits. This movement generates heat, which is impossible to avoid. In an effort to avoid overheating in data centers, operators establish a substantial cooling infrastructure – which creates heat itself.

As a result, a tricky cycle is developed – electrons make heat, then cooling systems try to reduce the impact of heat. These together increase the demand on power. And recently the trend toward increasing AI workloads makes it unsustainable. As an industry, we need a new approach to tackle the issue.

The Benefits of Photonics

Looking for an answer, let’s consider photonics. These are particles of light, which can be the basis for communication and computation. What differs photons from electrons is that they do not create heat as they move through optical systems. As a result, photonic communications can decrease latency, reduce energy consumption, and ultimately cut cooling needs. 

It’s important to consider optical fiber, used broadly across modern networks. Currently it is used primarily to replace long-haul electrical cables. However, there remains a bottleneck at the endpoints, where processors and chips rely on electronics. To make a more dramatic transformation, it is critical to extend photons end-to-end. This means from the network backbone and then to the chip and device level. By doing so, we can reach an ultra-low level of power consumption, generate minimal heat, and enable almost instant communications – which is mandatory for the next era of computing. 

Moving Toward a Photon Future as an Industry

Looking across the industry, an impressive effort to reimagine networks based on photons is the IOWN (Innovative Optical and Wireless Network). This program from NTT was established in 2019. IOWN is creating a photonic network with the ability to provide real-time communication – with much less power and little heat. In 2023, a worldwide consortium was created by NTT to move forward. Industry leaders – more than 130 – joined including Sony, Intel, Ericsson, Nokia, Google, Microsoft, Red Hat, and Toyota.

If successful, IOWN will reshape global communication infrastructure. Objectives include reducing delays which currently hold back industries such as telemedicine and autonomous vehicles. The results include less consumption of energy which reduces environmental strain, ultimately lowering costs for corporations. It’s amazing to consider the potential environmental and economic impact. 

Initiatives Across Europe and America

NTT is not alone. In Europe, the ETSI F5G and F5G-Advanced programs are working toward all-optical fixed networks that push fiber “everywhere.” Their targets include tenfold increases in bandwidth and energy efficiency, as well as sub-millisecond latency—goals closely aligned with IOWN’s All-Photonics Network.

At the same time, Microsoft – based in the United States – is developing hollow-core fiber (HCF) technology after acquiring Lumenisity. HCF is meant to reduce signal delay vs. standard fiber. It is designed to provide enhanced security, which makes it appealing for backbone links and cloud networks.

Data Center Photonics

We also see chip and rack level innovation. Ayar Labs created TeraPHY, an in-package optical I/O solution meant to replace copper for chip-to-chip communication. This increases bandwidth while using less power. Lightmatter developed Passage, a photonic interconnect fabric capable of moving data across accelerators and memory very quickly, reaching hundreds of terabits per second. Another example is Celestial AI, which created Photonic Fabric, meant to speed up links between memory and AI chips while requiring less energy. 

Together, innovations like these are designed to get rid of key bottlenecks in AI computing – the great amount of power needed to move data inside and between servers.

Ecosystem Standards

Work groups such as the Optical Internetworking Forum (OIF) are establishing a framework of photonics-at-scale. OIF is creating specifications for 800G and 1600G coherent optical interfaces, energy-efficient optical modules, and co-packaged optics for network switches and ASICs. Standards like these are critical to align industry efforts and guarantee interoperability.

Conclusion: Light at the End of the Tunnel

We’ve seen global data creation almost triple in the last five years, achieving 180 zettabytes in 2020 vs. only 64 zettabytes in 2015. Some consider this rapid growth the beginning of the “Zettabyte Era.” This is a time period recognized increased demand of cloud computing, digital services, and artificial intelligence (AI).

AI data centers are projected to consume 90 TWh by 2026, about one-seventh of the total global data center load. The GPU- and CPU-intensive nature of AI training is pushing data center power demand toward 96 gigawatts (GW), with cooling systems alone responsible for up to 40 percent of that energy bill.

This is a major paradigm shift, not just a technology upgrade. For us to meet digital age demands and remain sustainable, policy makers, industry leaders, and investors should move forward with light-based networks. If widely adopted, photonics could usher in a new era of connectivity: One that is faster, cleaner, and sustainable for generations to come.

The early-rate deadline for the 2026 Inc. Regionals Awards is Friday, November 14, at 11:59 p.m. PT. Apply now.