Tech to Solve the Data Center Energy and Resource Dilemma

“The data center of the future is modular, liquid-cooled, photon-powered, and flexible — an energy ecosystem, not a consumer.”

The Challenge: Explosive Growth Meets Energy Constraints

Emerald’s latest webinar tackled one of the most pressing questions in the energy and technology world: how can data centers keep up with skyrocketing AI-driven demand without breaking the energy system?

Computing requirements have multiplied 500,000× over the past decade, and global data center energy use—already around 415 TWh in 2024—is projected to more than double by 2030, rivaling Japan’s total consumption.

Watch the full webinar (registration required):

Emerald Data Center Webinar Recording

Six Pain Points Shaping Innovation

Christoph Frei identified six interlocking challenges driving both stress and opportunity across the data center landscape:

Firm Energy Supply – How can we reliably power this surge? Options range from geothermal and 24/7 renewables with storage to super-capacitors for grid resilience.
Centralized vs. Modular Design – Gigawatt-scale hyperscalers meet smaller, modular centers closer to the edge. Each use case—training LLMs vs. low-latency inference—demands a different infrastructure model.
Cooling & Waste Heat – Transitioning from air to liquid cooling is inevitable as rack power density leaps from 10 kW to 100 kW+. Higher-temperature waste heat opens up reuse in district heating, agriculture, and industrial processes.
Flexibility – Can data centers act as flexible loads? With smart orchestration, non-urgent AI training or batch jobs could shift in time or geography, turning data centers into grid-balancing assets.
Tech Shifts – From electrons to photons, optical interconnects and chips could massively cut energy loss while boosting performance.
Underlying Systemic Shifts – AI’s impact on energy is two-way: AI for energy and energy for AI are both redefining how we build digital and physical infrastructure.

Key Technical Nuggets

AI Symphonies, Not Soloists – As Graham Carey explained, the industry is moving from one giant model to “fit-for-purpose orchestration,” where smaller, specialized models handle different tasks. This could cut energy per query dramatically.
Photonics on the Rise – Every transition from photon to electron wastes power. Optical communication—and eventually optical computing—could revolutionize bandwidth efficiency within and between data centers.
Liquid Cooling Inflection Point – Demand for immersion and near-die cooling is real this time. Top-end chips now push 1 kW per processor, making air cooling obsolete.
DC Power Systems & Efficiency Gains – Startups are pioneering high-voltage DC architectures, cutting transformer losses and even copper usage, while improving energy conversion efficiency.
Flexibility Case Study – BitDeer and Google experiments show data centers can shed up to 25% of load for hours without affecting user experience—a glimpse into the future of demand-responsive computing.
Next-Gen Power Sources – Geothermal and small modular reactors (SMRs) are emerging as credible firm-power solutions. Deep drilling startups and fast-track SMR projects (e.g., Hitachi-GE’s 300 MW unit in Canada) could supply clean, reliable power within the decade.
Fusion Watch – Once “always 20 years away,” Frank Balas pointed out that fusion now has startups reaching ignition temperatures, with Commonwealth Fusion Systems targeting a working reactor by 2027.

The Takeaway

Emerald’s experts see the future data center as a symphony of solutions—liquid-cooled, photon-powered, flex-enabled, and co-located with firm renewable or nuclear sources.