More Barclays On GE Vernova And The Data Center Migration To 800 Volt DC (GEV)

Following on June 16's quick hit: "GE Vernova, nVent best positioned for 800VDC technology, says Barclays" (GEV; NVT).

From AI Invest, June 13:

GE Vernova and the 800VDC Bet: Why Barclays Thinks It Can Win AI's Next Infrastructure Layer 

- Barclays argues AI's next bottleneck is power infrastructure, not servers, as data centers surge in electrical equipment861011-0.55% demand.

- GE Vernova's 800VDC technology, partnered with NVIDIA, could reshape AI factories by improving efficiency and reducing copper use.

- The company's electrification orders doubled in 2026 Q1, but timing risks remain if 800VDC adoption lags expected 2028 commercialization.

- GridOS software and grid expertise position GE Vernova to capture value, though valuation concerns and insider selling raise caution.

- Investors must watch 800VDC deployment progress, revenue conversion, and hardware-software integration to validate the infrastructure thesis.

Barclays' core thesis: AI's next bottleneck is power, not servers

The key question is not whether AI demand is strong. It is whether the market is still valuing GE Vernova mostly as an industrial earnings story while missing the power infrastructure opportunity underneath AI compute. That debate matters because the demand signal has intensified: data centers bought more electrical equipment in the first three months of 2026 than in all of 2025, and GE Vernova said orders across its business rose almost 80%, with electrification orders essentially doubling. If power infrastructure becomes the binding constraint, companies tied to that layer could be revalued before the earnings impact is fully visible.

Why 800VDC matters in the AI power stack
800 VDC power distribution improves efficiency, reduces copper usage, and simplifies system architecture as AI facilities move to higher power density. That shift is not just theoretical: it is linked to NVIDIA's upcoming Kyber rack architecture, and GE Vernova has partnered with NVIDIA on reference designs for gigascale AI factories centered on 800 VDC. Barclays has highlighted the technology's commercial rollout as it is expected to begin from 2028, with GEV seen as a potential beneficiary.

The broader buildout also helps explain why this is an infrastructure story rather than a short-cycle server trade. Global data-center capacity is projected to move from 135 gigawatts to 300 gigawatts by 2030. The main risk to the thesis is timing: if adoption stalls after design wins, or if deployment slips well beyond the expected commercial ramp, the opportunity may take longer to show up in results than bulls expect.

GE Vernova's position in the grid-to-rack chain
That 800VDC shift matters because power delivery is becoming the real constraint. AI facilities are moving toward 1 to 5 gigawatt AI factories and rack loads that can reach one megawatt per rack. In that environment, power is no longer a support function; it shapes how quickly a site can be built and how much compute can actually be installed. The recent procurement signal also points to acceleration, with data-center buyers purchasing more electrical equipment in the first three months of 2026 than in all of 2025. 

Hardware sits at the critical grid interface
The bottleneck is not just megawatts. It is the ability to move those megawatts through transformers, switchgear, substations, and grid controls without creating reliability problems. GE Vernova's electrification business makes that equipment, and the recent doubling in those orders matters because those products sit at the interface between the utility grid and the AI factory. They are the hardware nodes where voltage is stepped down, distributed, and protected before power reaches the compute floor....

....MUCH MORE 

The winners in the whole data center buildout will be those companies that can deliver the most compute for the fewest watts.