BL!XT selected for NVIDIA Inception — Why power is becoming the bottleneck in AI infrastructure

BL!XT has been selected for NVIDIA Inception — NVIDIA’s program for companies building the next generation of AI infrastructure.

Inception supports a focused set of companies with access to NVIDIA’s full stack: compute platforms, developer tools, and a global ecosystem shaping how AI infrastructure is designed and deployed. For us, the significance is not only the program itself — it’s what it signals about where constraints and value are moving in the stack.

AI is no longer constrained by compute alone — it is increasingly constrained by power.

Global data center electricity use is projected to more than double — from roughly 415 TWh in 2024 to around 945 TWh by 2030 — with AI as the primary driver. In the US, installed data center power capacity is expected to grow from just over 30 GW in 2025 to more than 90 GW by 2030.

For the past decade, the industry has focused on scaling compute: faster chips, larger clusters, more parallelization. That strategy has worked. But beneath that exponential growth sits an infrastructure layer that has barely changed.

A growing mismatch

Modern AI data centers are entering a new regime: extreme power densities, rapid load fluctuations, and increased sensitivity to power quality. Yet the electrical systems supporting them remain largely mechanical.

Protection is still handled by circuit breakers designed for a different era. Control is limited. Visibility is fragmented. Response times are slow relative to the value and sensitivity of the compute they protect.

We are effectively running dynamic, software-defined compute on top of static, mechanical power systems.

As AI workloads scale, this mismatch becomes economic. Power density limits cap how tightly compute can be deployed, directly impacting revenue per rack and per megawatt. Inefficiencies in legacy AC architectures compound at scale, while limited real-time control reduces utilization and flexibility. Protection systems not designed for microsecond-level dynamics introduce operational risk in GPU-dense environments.

The bottleneck is shifting — not away from compute, but toward the ability to deliver, control, and economically optimize power at the same pace.

From electrical hardware to programmable power

To support the next generation of AI infrastructure, the power layer must evolve in the same way compute and networking already have. It needs to become dynamic, observable, controllable, and software-defined.

This is the transition BL!XT is building toward.

At the core is BL!XT Zero®, a solid-state switchgear platform that replaces mechanical components with semiconductor-based control. Response times move from milliseconds to microseconds, and electrical flows can be monitored and controlled in real time at the granularity AI workloads demand.

Zero is designed for both AC and DC environments — enabling a transition toward more efficient DC architectures now being explored in hyperscale data centers.

BL!XT is extending this programmability into conversion and storage with XVerter®, a software-defined approach that combines solid-state transformation and storage in a single system (think SST + BESS all-in-one) — enabling more flexible, efficient, and integrated on-site power architectures.

Together, this creates a programmable power fabric — from protection and distribution through to conversion and storage — turning the power layer from a static cost center into a performance- and value-driving layer of the stack.

Where NVIDIA fits — and the shift ahead

If NVIDIA is defining the compute layer of AI infrastructure, the layer beneath it becomes increasingly critical.

Through NVIDIA Inception, BL!XT collaborates with partners who are redefining AI data center architectures — ensuring the power layer is designed alongside compute rather than bolted on after the fact.

The scaling limits of AI are not disappearing — they are moving. For a long time, the constraint was compute. Increasingly, it is the ability to deliver, control, and optimize power at the same rate as investment in GPUs, cooling, and real estate.

AI data centers will not scale on legacy electrical infrastructure.

They will require a new layer — one that brings programmability, visibility, and control to power.

BL!XT is building that layer — making power a controllable input to performance, efficiency, and ultimately infrastructure returns.