Google and Blackstone Launch AI Cloud Venture

Google and Blackstone have formed a joint venture to build an AI-focused cloud business, aiming to deliver 500 megawatts of data center capacity by 2027 [1]. The new, as-yet-unnamed company will be led by longtime Google executive Benjamin Sloss and will utilize Google’s proprietary Tensor Processing Units (TPUs) to provide computing power to outside organizations [2].

Blackstone is providing an initial $5 billion in equity to fund the project, with the firm expected to support roughly $25 billion in total compute investments over time [2]. While Google has previously provided its chips to select partners like Anthropic and Meta, this venture marks the company’s most significant effort to commercialize its hardware for broader use [2]. By offering TPU-based infrastructure, the partnership aims to challenge the dominance of cloud providers like CoreWeave, which currently rely heavily on Nvidia hardware [2].

The move reflects a broader shift in the AI infrastructure market, where hyperscalers are increasingly commercializing their proprietary chips while private capital takes a larger role in funding massive data center projects [3]. Industry observers note that this strategy creates a viable alternative for companies seeking to avoid the Nvidia-centric ecosystem that has defined the "neocloud" model [3]. For Google, the venture serves as a way to monetize its specialized hardware beyond its own internal cloud platform [2].

The partnership leverages Blackstone’s extensive experience in the data center sector, including its acquisitions of QTS Realty Trust and AirTrunk [2]. As demand for the computing power required to train and run advanced AI models continues to surge, the venture is positioning itself to capture a significant share of the market by integrating chips, software, and energy infrastructure into a single, vertically integrated stack [3].

Whether this TPU-based model can successfully disrupt the established market for AI compute remains the primary question for the industry. The success of the venture will depend on how effectively it can scale its capacity to meet the massive energy and hardware requirements of the current AI boom [2].