• 25 per cent will exceed 100kW per rack, dedicated to large-scale training clusters.
To keep up with AI, operators must continue to look to the next and most advanced GPUs, for example, the NVIDIA Rubin CPX scheduled to release in late 2026. Coupled with NVIDIA Vera CPUs and Rubin GPUs inside the new NVIDIA Vera Rubin NVL144 CPX platform, this integrated NVIDIA MGX system packs 8 exaflops of AI compute to provide 7.5x more AI performance than NVIDIA GB300 NVL72 system.
Digital twins will permeate
design of almost everything In 2026, we will see the rise of digital twins as processing power continues to evolve in AI data centres and advanced platforms are developed, like NVIDIA’s Omniverse and Cosmos. Data centre operators will use digital twins to achieve greater efficiency and accelerate development by designing and simulating highly complex physical objects, systems and processes. Take for example a data centre’s power system itself. ETAP sophisticated modeling technology can create a virtual replica of a data centre’s electrical infrastructure through integration with NVIDIA Omniverse.
Liquid cooling goes mainstream Traditional cooling solutions are no match for the processing power required for AI. In 2026, rack densities will reach 240 kW per rack. 2028 will see densities reach 1 MW per rack, and research is underway to determine whether 1.5 MW per rack is possible. With ultra-high densities on the horizon, advanced cooling techniques are required to keep these systems operating. This means liquid cooling will transition from an outlier to the norm for data centre cooling solutions.
Retrofitting data centres for AI As AI becomes central to business strategy, organisations of all sizes need access to next- generation data centres. While hyperscalers will pursue greenfield builds as they have the need and the resources for large-scale investments, smaller companies will eye brownfield retrofits as a viable strategy for creating AI-ready facilities. Infrastructure providers have developed solutions to upgrade existing facilities, including larger IT racks, higher power PDUs and liquid cooling ‘bolt on’ rear door heat exchangers specifically designed for accelerated compute AI server racks and clusters. These components help ensure that AI is feasible for a broad spectrum of companies, not just the biggest ones in the world.
• • • 2026 PREDICTIONS • • • Continued focus on sustainability
concerning power demands Power sourcing will persist as a central theme in 2026, with data centre operators continuing to adopt a diverse array of power solutions, including natural gas turbines with carbon capture, HVO-fuelled back-up generators, wind, solar, geothermal and battery storage. Worldwide, renewables currently supply 27 per cent of the electricity consumed by data centres, mostly through wind, solar and hydropower sources. Total power generation for renewables is projected to grow 22 per cent each year until 2030, meeting nearly half of the anticipated growth of data centre electricity demand.
A new era in intelligent
infrastructure In 2026 AI will shift from a disruptive force to a foundational element of business and technology. Data centres will not simply support technology they will enable intelligence itself from liquid-cooled AI factories to highly advanced digital twins. To keep pace, operators will need global partnerships that can match the speed of technological and geopolitical changes.
https://www.se.com/ww/en/
electricalengineeringmagazine.co.uk
ELECTRICAL ENGINEERING • DECEMBER/JANUARY 2026 37
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52
