Alphabet and Blackstone announced a joint venture today, May 19, to create a new US-based TPU cloud company that will sell data center capacity and access to Google’s Tensor Processing Units as compute-as-a-service. Blackstone is committing $5 billion in initial equity to the venture, with the total investment expected to reach $25 billion including leverage, according to Reuters reporting of the official Blackstone press release. The company plans to bring 500 megawatts of capacity online in 2027 and scale significantly beyond that. Blackstone has named Benjamin Treynor Sloss, a Google executive with more than two decades of experience building and operating Google’s global infrastructure and operations, as chief executive officer of the new company.
Google will supply hardware including TPUs, along with software and services. The new venture will provide customers access to TPUs as an alternative to using them directly through Google Cloud, expanding the distribution of Google’s AI silicon into a standalone infrastructure vehicle for the first time.
Jon Gray, President and COO of Blackstone, described the deal as a generational opportunity to invest capital at scale building AI infrastructure, stating that the new company has enormous potential as it helps to meet the unprecedented demand for compute. Thomas Kurian, Google Cloud’s chief executive, said TPUs are optimised specifically for efficiency and performance in the AI era. The announcement positions the joint venture as a direct structural competitor to Nvidia-backed neocloud operators including CoreWeave, which has built its market position around Nvidia GPU access, and to the broader hyperscaler infrastructure model where AI compute access runs exclusively through the major cloud platforms.
What the Structure Signals
The joint venture structure mirrors the Stargate model that OpenAI, SoftBank, and Oracle established for large-scale US AI infrastructure investment, but with a critical difference: Google is supplying its own proprietary silicon rather than Nvidia GPUs. Blackstone, which manages more than $1.3 trillion in assets and operates more data center capacity than any other private investor through its BXN1 infrastructure arm, is bringing permanent capital and operational infrastructure expertise. Google is bringing a decade of TPU development, the software stack that makes TPUs accessible to enterprise customers, and a former infrastructure chief who spent twenty years building the systems that made Google’s AI capabilities possible. The combination creates a standalone AI infrastructure company that does not depend on Nvidia’s supply chain, pricing decisions, or product roadmap for its core compute capability.
The Competitive Implications for Nvidia
The Google-Blackstone venture is the most credible structural challenge to Nvidia’s AI infrastructure dominance to emerge from outside the hyperscaler ecosystem. Previous challenges to Nvidia’s position in AI infrastructure have come primarily from hyperscaler custom silicon programmes, which are closed systems that serve only the hyperscaler’s own workloads and cloud customers. The new Google-Blackstone venture is an open infrastructure company that will sell TPU compute access to any enterprise customer, competing directly with Nvidia-backed neoclouds for the same enterprise AI infrastructure market.
The venture’s success depends on whether enterprise customers will choose TPU infrastructure over GPU infrastructure for production AI workloads, a question that Google’s decade of TPU deployment in its own systems provides meaningful evidence for but that the commercial market has not yet resolved at scale. The neocloud consolidation and what it means for operators unprepared for what is coming mapped the competitive dynamics this venture is now entering. The Google-Blackstone TPU cloud is the most consequential new entrant into that market since CoreWeave’s IPO.
