TECfusions has announced that TensorWave has selected two TECfusions sites for the next phase of its AI infrastructure growth: a new 10 MW deployment at the Keystone Connect campus near Pittsburgh, Pennsylvania, and an additional 10 MW expansion at TECfusions’ existing Tucson, Arizona facility. TECfusions says the Tucson deployment already hosts the largest AMD-based AI deployment in North America, and the new phase will replicate the same high-density reference design across both locations.
TECfusions reports the deployments are scheduled for delivery in the first half of 2026. It says the added capacity extends TensorWave’s AMD training cluster using direct liquid cooling and high-rack-density designs intended for large-scale AI training workloads.
At Keystone Connect, TECfusions says TensorWave is starting a multi-phase plan to scale to 1 GW at the Pennsylvania location. TECfusions also states the site is capable of hosting 3 GW primarily through on-site power generation. TECfusions describes Keystone Connect as designed for a multi-phase, behind-the-meter power model that will begin with utility power and then transition most of the Pennsylvania deployment to behind-the-meter generation, with the ability to interconnect back to the grid.
In Tucson, TECfusions says the added 10 MW expands an existing 14.4 MW phase delivered to TensorWave in under four months in 2025. It also reports that the initial 10 MW phase at Keystone Connect is expected to support more than 100 skilled trade roles and create ongoing local tax revenue.
“By adding fresh AI clusters in these regions, our customers have the flexibility across our sites to spin up GPUs as needed for urgent workloads, performing training and other testing without pausing to rework their tooling or rewrite their stack,” said Darrick Horton, CEO of TensorWave. “TECfusions has shown, in production, that they can deliver on aggressive timelines especially given the complexity of high-density deployments for the customer needs we are serving.”
Source: TECfusions
