The all engulfing tidal wave of AI adoption is completely changing the game in virtually every industry on the planet. And none more so than in the very sector which curates it. Data Centre operators are being forced into a radical rethink about how they approach their operations as they struggle to cope with rapidly spiralling energy and resilience models demanded by the AI revolution.
The data centres of today must deliver operational capability of 50-100KW per rack to support AI functionality. A considerable leap from the 5-12KW demanded by traditional computing. What’s more, as AI adoption increases, the data centres of tomorrow will need to ramp this capability up to 200KW and beyond.
To put this into an energy perspective, the way things are moving in the sector, a single hyperscale AI data centre campus will have energy needs on a par with small cities.*
And while the operational capability and energy demanded spiral ever higher, the Alternating Current (AC) power architecture expected to deliver it remains largely unchanged.
AC systems in data centres get energy from grid to rack through a meandering journey via transformers, uninterruptible power supplies (UPS), power distribution units (PDUs), heavy copper conductors and multiple conversion stages. As AI demands ever increasing power densities, this circuitous approach to moving energy compounds into significant inefficiencies which could see potential losses of multiple percentage points at hyperscale level becoming the uncomfortable norm.
The result then, in ironic contrast to other power- hungry emergent sectors like EV charging for instance, grid constraint isn’t the primary blocker to progress. It’s getting power from grid to rack without losing lots of it on the way.
And there is an answer. HVDC or High Voltage Direct Current, particularly in 800VDC architecture, can be the gamechanger data centres need to allow operators to expand operating capacity ahead of the spiralling demands of the industry, while minimising energy losses and driving CAPEX savings through the reduced hardware requirement by DC architecture.
So, let’s look at the core elements of HVDC to understand its advantage over AC systems.
A more streamlined path with less hardware
HVDC architectures can drastically streamline the power path from grid to server. Instead of four or more conversion stages typical in AC distribution, HVDC would use rectification from medium-voltage AC to a stable 800VDC microgrid followed by high-efficiency rack-level conversion to a server-ready 50VDC – just two conversion stages.
As a result, efficiency over AC systems would rise by several percentage points, becoming significant at hundreds of megawatts hyperscale levels. Less energy would be lost through heat, too, reducing cooling costs and extending equipment lifespan. DC systems would also dispense with the need for internal power supplies within the racks themselves, freeing up space to get more power density from smaller footprints lending itself to numerous other efficiencies at operational level.
Power flowing at higher voltages dramatically reduces current, which in turn lowers resistive levels and reduces the need for expensive copper -based hardware. A typical 800VDC deployment can reduce cabling requirement by up to 50% in some configurations with associated installation labour reductions of 30–50%.
Combining these efficiencies delivers substantial cost savings. Taking a relatively- in today’s AI terms – modest 10MW facility we can project approximate figures like this:
And if we expand our thinking into the scaling energy needs of the data centres of tomorrow, such as the 100MW facility Magnora ASA is planning in Norway, these figures start to ramp dramatically:
And to expand our model even further into the hyperscale campus of the future – QTS’ facility planned for Blyth, North East England for instance- adoption of an HVDC approach could see estimated annual energy savings alone in the region of 400 Gigawatt Hours***
That’s enough to power around160 of our VELOX Ultra-Rapid EV chargers to charge continuously, 24/7 for one year.
But it’s not just straight efficiency where HVDC excels over AC.
While the efficiency advantages speak for themselves, there’s more to the HVDC story. Ironically, reliability in AC architectures is most threatened by the very components intended to enhance stability – in particular UPS units and integral server power supplies.
The heat generated by multiple conversions on the AC journey from grid to rack often causes them to overheat and fail. Meanwhile more broadly, the hardware dense architecture in general provides more opportunities for failure, over cooler, smoother, simpler DC architecture.
And this simplicity delivers maintenance advantages too – hot-swappable converters that allow server repair and maintenance without downtime – for instance. Particularly valuable in the AI world where service continuity is critical.
F2R2G (Facility to Renewables to Grid) the Big Resilience Advantage
Solar PV generation and battery systems inherently operate in DC, within voltage ranges compatible with data centre 800VDC architecture. So being able to couple local clean generation and battery storage directly into the DC microgrid, removes even more cumbersome inefficient AC conversion stages, decreasing energy loss and decreasing the infrastructure costs of integration.
Being Bi-Directional, and able to both supply to as well as draw from the grid, 800VDC architecture positions data centres as dynamic participants in the local and national energy eco system. While excess onsite generation can be supplied to the grid in times of high demand, excess renewable generation grid side can be taken off it and absorbed into storage by the facility, enhancing grid stability, reducing costly curtailment**** and unlocking new resilience models for campus-scale deployments.
Standardisation and industry momentum
Emerging Industry standard are pressing the shift toward HVDC too. Server platforms built around the OCP (Open Compute Project) 48/50VDC standard are becoming the expected norm. At the same time, global efficiency and sustainability pressures, supply-chain constraints around transformers and UPS systems, along with tightening regulatory frameworks are all playing their part in driving the sector to the HVDC inflection point.
Is there a but coming?
HVDC architecture has huge capabilities. Not least in giving data centres the means to thrive through the demands of AI rather than getting burned by them. But deployment of it in data centres is breaking entirely new ground and brings with it the need for a very different approach to core operations. At the very least, far more sophisticated thinking around safety and resilience practices. And this sort of thinking doesn’t currently exist within the sector.
In simple terms
When faults occur within DC architectures they behave very differently to AC faults. The higher energy levels involved mean more heat or thermal stress is generated which significantly increases stress on components and risk of injury to people.
This calls for a deep understanding of DC fault architecture, informing validated PPE and training frameworks for personnel, and even re-thinking designs of facilities themselves to mitigate risk to people and equipment if something goes wrong.
The same applies to switching technologies. Currently geared to the less demanding needs of AC architecture, they fall way below the tolerance level demanded by DC operation. Particularly in the event of a fault. This calls for an engineering re-think with an expert DC lens.
And then we get to HR. By what means will people running HVDC environments be trained to cope with this new, unfamiliar world with its greater risk levels? Compounded too, when we understand that the immature standards landscape within the industry offers little meaningful guidance.
So, adoption of HVDC isn’t a light undertaking. In fact, it’s an understandable stumbling block for operators and investors.
Even more so when you understand that if it isn’t deployed correctly, the efficiencies we’ve talked about above will be seriously diminished and operators will be left with all the costs, none of the benefits and no way forward.
Your Partner with Deep HVDC Expertise
TPS has over 45 years’ experience designing and deploying HVDC microgrids into industries from EV charging to the energy sector itself, becoming a globally recognised partner for organisations seeking to unlock the exponential benefits this technology can deliver for them. Our proprietary technology reaches the highest standards of resilience known to the industry and with typical modular hardware rated to 150KW, our solutions are infinitely scalable.
For operators, investors and policymakers alike, the implication is clear. HVDC is not just a technical alternative—it is a foundational enabler of AI readiness. In the next generation of digital infrastructure, the question will no longer be whether to adopt HVDC, but how quickly it can be implemented at scale.
If you want to talk more about how our HVDC solution will accelerate your Data Centre expansion journey please drop us a line. We’ll be happy to arrange a call to discuss how we can help.
Footnotes
*QTS’ facility planned for Blyth, North East England is a prime example – see more above.
**To keep things as simple as possible, this cost model works on the basis of 100% utilisation and steers clear of PUE, the industry measure of realistic facility consumption when other factors such as cooling are added. Figures are based on a 5% energy saving.
***Assumes a 1 Gigawatt facility running the same model as above. A VELOX Ultra Rapid EV Charger delivering 300KW.
**** According to the UK Curtailment Monitor we have curtailed 2.8TWh of wind generated energy this year to date (18/3) ie removed it from the grid because there is no available capacity for it.
The Grid-Standard Paradox: Why MCS Needs a Brain, Not Just a Bigger Pipe
Speak to one of our experts about your specialist project requirements