LT350 Plans Solar-Powered Data Centers in Parking Lots

Parking lots as AI infrastructure

A startup called LT350 is proposing to transform underutilized parking lots into distributed data center sites, combining solar canopies with battery storage and GPU compute capacity to support growing AI inference workloads.

CEO Jeff Thramann outlined the company's vision in an interview, describing a model that would deploy smaller-scale data center infrastructure across multiple parking lot locations rather than concentrating capacity in massive single facilities. The solar canopies would serve dual purposes: generating power for the GPU systems while providing shade for vehicles below.

Why it matters

As AI adoption accelerates, data center capacity has become a critical bottleneck. Traditional hyperscale facilities face mounting opposition from communities concerned about power consumption, water usage, and land development. LT350's distributed approach could offer a path to scale AI infrastructure while reducing the concentrated environmental and social impact that has made large data centers increasingly controversial in residential areas.

The distributed model

Thramann argues that spreading data center capacity across smaller, distributed locations could mitigate the community disruption associated with large-scale facilities. By leveraging existing parking infrastructure that often sits underused, the approach would avoid the need for new land development while adding functional value to spaces that currently serve only vehicle storage.

The integration of solar generation and battery storage addresses one of the primary concerns around AI infrastructure: energy demand. By generating power on-site, the facilities could reduce grid strain while potentially operating with greater energy independence than traditional data centers.

Technical considerations

The model relies on deploying GPUs—the specialized processors that power AI workloads—in modular units beneath solar canopy structures. Battery systems would store excess solar generation for use during peak demand periods or nighttime operations.

While Thramann provided the conceptual framework, technical details about cooling systems, network connectivity requirements, and the scale of GPU deployment per site were not disclosed in the interview.

Market context

The proposal comes as AI companies and cloud providers face mounting pressure to expand inference capacity—the computational resources needed to run trained AI models at scale. Inference workloads differ from training in that they can be distributed more easily across geographic locations, making LT350's parking lot model potentially viable for certain use cases.

Details about LT350's funding, deployment timeline, and customer pipeline were first reported by Fox Business during Thramann's appearance on Varney & Company.