For decades, the uninterruptible power supply had a simple job: sit quietly between the grid and the load, and catch the power when it fell. That job hasn’t gone away. But the way UPS systems are designed is changing faster right now than at any point in recent memory, and AI is the biggest reason why.
At Strategic Power Innovations, we work with hyperscale and enterprise data centers, colocation providers and enterprise facilities, and each of them is feeling these changes differently. Here are five shifts we think everyone in critical power should be watching.
1. The move toward 800 VDC
The most significant change is happening at the architecture level. Rack power demands in AI data centers are scaling beyond 300 kW, which is pushing the industry toward centralized 800 VDC architectures to reduce thermal losses, copper use and current levels.
This isn’t one vendor’s idea. NVIDIA, Google and Microsoft have been developing the 800 VDC architecture together through the Open Compute Project, publishing a joint white paper in March 2026 and a solid-state transformer specification in July 2026. Major UPS manufacturers are responding: Vertiv’s planned 800 VDC system includes centralized rectifiers, high-efficiency DC busways, rack-level DC-DC converters and DC-compatible backup systems.
It will likely arrive in stages. Schneider Electric’s roadmap starts around 2027 with sidecar converters that bring 800 VDC to high-compute rows without redesigning the whole facility, then moves to a centralized DC UPS and building-wide 800 VDC busway, and longer term toward solid-state transformers converting medium-voltage AC directly to 800 VDC.
That staged path matters a lot for colocation operators. Most colos will be serving mixed tenants for years: some running dense AI racks, others running conventional IT. Being able to add DC capacity row by row, without stranding the AC infrastructure already in place, is going to be a major design consideration.
2. Silicon carbide is raising the efficiency ceiling
Inside the UPS, the power semiconductors are changing. Manufacturers are replacing traditional silicon IGBTs with silicon carbide (SiC). SiC has roughly three times the thermal conductivity of silicon, so heat is easier to manage, and its faster switching makes the UPS more responsive while its durability extends component life and reduces maintenance.
The result: SiC-based designs are reaching around 98% efficiency in full online double-conversion mode. For a large data center, every tenth of a percent shows up in the utility bill and the cooling load. For enterprise customers with older UPS fleets, this is often the most practical upgrade on the table today, since it delivers efficiency gains without changing anything downstream.
3. Lithium-ion and modularity are now the default conversation
Battery chemistry and form factor are moving together. Analysts expect sustained demand for UPS systems with high-cycle lithium-ion batteries in data centers, and modular designs are increasingly how capacity gets added, giving customers “pay as you grow” scalability with high power density in a compact footprint.
That said, VRLA isn’t disappearing overnight; it still holds the largest share of the data center UPS market in 2026. Most sites we see are in a transition period, not a clean break, which makes lifecycle planning and battery strategy just as important as the equipment choice itself.
4. The UPS is becoming a grid asset
A UPS full of batteries is also an energy storage system. Smart grid capabilities can turn a UPS from a reactive piece of equipment into a dynamic “virtual power plant,” with uses like peak shaving, where battery power limits how much a site draws from the mains, and frequency stabilization services.
With grid capacity now one of the hardest constraints on new data center builds, we expect this to shift from “nice feature” to a real part of the business case, especially for operators looking to get more out of the utility capacity they already have.
5. Total cost of ownership is replacing upfront price
Put it all together and the buying conversation is changing. Data center customers are focusing more on total cost of ownership than initial capacity, and hyperscale operators have begun specifying double-conversion UPS systems above 97% efficiency. Efficiency, footprint, battery life, grid participation and serviceability now all belong in the same evaluation.
What this means going forward
The UPS isn’t becoming less important in the AI era; it’s becoming more central and more complex. The right answer looks different for a hyperscale AI build, a multi-tenant colo and an enterprise server room, and the teams that do well will be the ones that understand the whole power path, from grid to chip, rather than just the box in the electrical room.
That’s the approach we take at Strategic Power Innovations: helping data center, colocation and enterprise customers make power decisions that work today and leave room for what’s coming next.
I’m curious what others are seeing: are your customers asking about 800 VDC yet, or is the focus still on getting more out of the AC systems they already own?
#UPS #CriticalPower #DataCenters #Colocation #AIInfrastructure #800VDC #SiliconCarbide #EnergyEfficiency #StrategicPowerInnovations
Planning a UPS upgrade? See our critical power solutions in Atlanta and UPS system installation services.