The AI infrastructure company created by Blackstone and Google launched publicly Wednesday under the name Crux AI. It plans to bring 500 megawatts of Google Tensor Processing Unit capacity online in 2027, backed by an initial $5 billion equity commitment from Blackstone-managed funds.
The capital and capacity target are not new financing announcements. Blackstone disclosed both when it announced the joint venture in May. Wednesday’s launch supplies the name, website and operating plan for the company, which is aimed at AI labs, technology companies, enterprises and governments seeking dedicated compute.
Google will provide TPUs, software and services. Crux will combine them with power, data center capacity, networking and operations, giving customers one provider responsible for the physical site through production deployment. The service offers another route to Google’s custom accelerators outside the standard Google Cloud service.
Google has used TPUs in production for more than a decade. The purpose-built chips support training and inference and power Gemini as well as other Google products, according to the companies’ May announcement.
Crux says it will own and operate its data center sites over the long term. Its website describes a capacity ramp of two gigawatts a year beginning in the second half of 2027. The launch announcement does not identify facility locations, customers or pricing.
Chief executive Benjamin Treynor Sloss spent more than two decades at Google, where he originated the discipline of site reliability engineering and led infrastructure spanning networks, data centers, production software and operations. Blackstone named him to lead the venture when it announced the partnership in May.
The company enters a neocloud market in which specialist operators lease large blocks of accelerated computing without running a general-purpose hyperscale cloud. Many such services use Nvidia GPUs. Crux’s hardware relationship makes Google TPUs its defining supply choice, while Blackstone supplies capital and data center experience.
Google has been expanding the supplier base around its TPU platform. A long-term custom-silicon agreement with Marvell covers inference accelerators, networking and memory systems, alongside Google’s established work with Broadcom. Crux adds a customer-facing operator to that hardware ecosystem.
Blackstone is already financing AI infrastructure through other structures. It joined Apollo in a $35 billion private-credit transaction for Broadcom-based Anthropic capacity, and the wider market has begun using accelerators as loan collateral to fund construction. The Crux commitment is equity in an operating company rather than debt secured against chips.
Crux says its first 500 megawatts will be the opening phase of a multi-gigawatt plan, followed by a capacity ramp of two gigawatts a year beginning in the second half of 2027.
Sources: Crux AI, Crux AI company site, Blackstone
–
By the Control Plane Editorial Team