admin The US-based venture will offer data centre capacity, operations, networking and Google Cloud’s Tensor Processing Units through a compute-as-a-service model. Blackstone is making the initial equity commitment, with the first 500 megawatts of capacity expected to come online in 2027. Google will supply the TPUs, software and services, giving customers another route to access its custom AI chips beyond direct procurement through Google Cloud.
The plan marks one of Google’s most significant moves to commercialise its in-house AI chip technology at scale. TPUs, developed over more than a decade, are designed for the training and inference of advanced AI models. They already support Google’s Gemini products and serve workloads for AI laboratories, capital markets firms and high-performance computing users. The new company is expected to target organisations facing shortages, high costs or capacity constraints in the broader accelerated computing market.
Blackstone will hold the financial muscle behind the platform at a time when AI data centres are becoming one of the largest investment themes in global infrastructure. The alternative asset manager, which oversees more than $1.3 trillion in assets, has expanded its exposure to digital infrastructure, data centres, energy generation and transmission networks. AI workloads require large, reliable and power-intensive facilities, making the availability of land, electricity, cooling systems and capital as important as access to chips.
Benjamin Treynor Sloss, a long-serving Google executive with extensive experience in global infrastructure and operations, will lead the venture as chief executive. His appointment signals that the company is intended to function as an operational cloud infrastructure platform rather than a passive financing vehicle. Jon Gray, Blackstone’s president and chief operating officer, described AI infrastructure as a “generational opportunity” for large-scale capital deployment, while Google Cloud chief executive Thomas Kurian has framed the venture as a way to meet rising demand for TPU access.
The partnership also reflects a deeper shift in the AI compute market. Nvidia’s GPUs remain the preferred hardware for many frontier model developers, cloud operators and enterprise AI teams, supported by the company’s mature software ecosystem and supply chain reach. Yet major cloud providers are increasing investment in proprietary chips to lower costs, improve performance for specific workloads and reduce dependence on external suppliers. Amazon has Trainium, Microsoft has Maia, and Google has built the most established custom AI accelerator platform among the hyperscalers.
For Google, the Blackstone venture creates a capital-efficient channel to expand TPU capacity while preserving Google Cloud as the anchor platform for its own enterprise AI strategy. It also gives the company a more visible role in the fast-growing market for dedicated AI compute, where specialised cloud providers and GPU-backed infrastructure firms have attracted strong demand from start-ups and model developers.
The move comes as demand for AI infrastructure strains global supply chains. Large technology companies are committing hundreds of billions of dollars to data centres, chips, power systems and networking. Alphabet has raised its 2026 capital spending plans sharply, with AI infrastructure and cloud capacity among the major drivers. Anthropic, one of Google’s key AI partners, has already expanded its use of Google Cloud TPUs, including plans for very large-scale capacity to train and serve future Claude models.
Blackstone’s role is equally strategic. Data centres have become a core infrastructure asset class, but AI facilities require larger power commitments and more complex construction planning than conventional cloud sites. The venture allows Blackstone to combine its capital base and physical infrastructure portfolio with Google’s chip and software stack, potentially creating a platform that can scale as enterprise AI adoption accelerates.
Competition will be intense. GPU capacity remains central to the market, and Nvidia-backed cloud providers are moving quickly to secure customers that need flexible access to high-end accelerators. Microsoft, Amazon and Oracle are also expanding AI infrastructure offerings, while specialist operators are trying to win clients by promising faster deployment and more tailored compute clusters.
