# Xeal Plans Parking-Lot Compute Pods for Up to 100,000 Nvidia GPUs > Xeal says its Laitent Pods could use spare electrical capacity at EV charging sites for AI inference. The first installation is planned by the end of 2026. > > OpenCheese AI · 2026.10.03 · Web page: https://opencheese.dev/blog/xeal-parking-lot-compute-pods-nvidia-gpus Parking lots are built for cars, but much of the electrical capacity at some charging sites sits unused. Xeal wants to put that spare capacity to work running AI inference, with a plan to install enough compact compute modules to house as many as 100,000 Nvidia GPUs. ## Power already connected Xeal operates EV charging infrastructure in 500 US cities. The company says its 1,600 sites have more than 200 megawatts of approved, grid-connected capacity, while charging stations typically draw less than 10% of their permitted maximum load. That gap is the basis for Laitent, Xeal’s platform for placing small compute modules directly in parking areas. The pitch is straightforward: use power connections that already exist, rather than wait for a new data center and its separate energy infrastructure. Existing connections may shorten the path to deployment, though the available capacity still varies with charging demand. ## A data center in one parking space Xeal’s Laitent Pods are designed to fit within the footprint of a parking space. Each can hold up to 48 Nvidia Hopper or Blackwell GPUs. The modules do not require a water connection, which could make them easier to place at sites where building a conventional data center would be difficult. Laitent software is meant to adjust computing activity to the power available. It can use part of a GPU or combine capacity across multiple modules in the same city. Xeal is targeting inference, the process of running models that have already been trained, rather than model training. That distinction gives the proposal a defined job. Xeal is not pitching a parking lot as a replacement for every kind of data center. It is aiming to distribute capacity for workloads that can run across smaller installations, with computing resources matched to changing power availability. ## Plans, not a deployed network Xeal expects to install its first module by the end of 2026. The company says the approach could bring new computing capacity online in months rather than years because sites would not need to obtain new power-connection approvals or build separate energy infrastructure. There is also a property-owner pitch. Xeal estimates that adding compute modules could raise a parking site’s value by about $1 million, with little or no investment from the owner. That is the company’s estimate, not a reported result from an operating installation. The first deployment will have to show how the economics work in practice. The broader idea is already attracting other plans. In March, Auddia announced an intention to place solar-powered GPUs at parking areas serving medical-property sites near Dallas. Belgian startup Tonomia is also developing distributed AI infrastructure in solar parking canopies. These projects differ in their described setups, but share the premise that parking infrastructure might serve another purpose when it is not busy doing its first one. For Xeal, the main asset is not the parking space itself. It is the power connection already waiting there. The first pod will show whether that is enough to make a parking lot a useful home for AI inference. ## Related posts - 2026.10.29 · [Anthropic files S-1 and plans Nasdaq listing as ANT after an $80.7 billion half-year loss](https://opencheese.dev/blog/anthropic-files-s-1-plans-nasdaq-listing-ant-80-7-billion-half-year-loss.md): Anthropic has made its S-1 public, with a planned Nasdaq listing under the ticker ANT. It shows $16.3 billion in first-half revenue, an $80.7 billion net loss and hundreds of billions in compute commitments. - 2026.11.09 · [Apple Plans to Turn Siri Into an AI Agent for Third-Party Apps](https://opencheese.dev/blog/apple-plans-to-turn-siri-into-an-ai-agent-for-third-party-apps.md): Apple is preparing a more autonomous Siri that could book tables, order food and carry out multistep tasks in its own apps. Some features could arrive in December 2026, with third-party integrations expected in spring 2027. - 2026.11.06 · [Airbnb plans an AI travel agent for the first half of 2027](https://opencheese.dev/blog/airbnb-plans-an-ai-travel-agent-for-the-first-half-of-2027.md): Airbnb says its agent will assemble maps, photos and other travel tools around guest requests, then help with reservation changes, payments and refunds. The company is betting that travel needs more than a chat window.