Alphabet Inc. and Blackstone Inc.'s artificial intelligence cloud venture has encountered delays at key data-center locations intended to house Google's custom Tensor Processing Units, jeopardizing the partnership's target of delivering 500 megawatts of compute capacity by 2027.
Blackstone committed $5 billion in equity capital to the unnamed U.S.-based venture, which will offer data center capacity paired with Google's TPUs through a compute-as-a-service model. Total investment could reach $25 billion including leverage. The initial plan called for the first 500 megawatts operational by 2027, with aggressive scaling beyond that date.
Benjamin Treynor Sloss, Google's former chief programs officer, will lead the venture as CEO. Google has not specified whether it will retain direct governance.
Blackstone, which manages more than $1.3 trillion in assets, is pursuing multiple plays in AI infrastructure. The firm established a similar venture with Anthropic earlier this month.
Google developed its first TPU in 2015 as part of a broader strategy to reduce dependence on external hardware providers like Nvidia. Unlike general-purpose graphics processing units, TPUs are optimized for specific workloads—particularly agentic AI models. Google runs its Gemini model on proprietary TPUs; Anthropic and Citadel Securities also use Google's TPU services.
The venture reflects intensifying competition between Google and Nvidia in AI hardware. Nvidia's GPUs, originally developed in 1999, parallelize complex computing tasks. Demand for Nvidia's processors surged after OpenAI's ChatGPT launch in 2022, and the company became the world's most valuable in 2024. Amazon Web Services and other technology giants have similarly pursued custom semiconductor development.
Site-level execution delays underscore the gap between capital commitment and infrastructure buildout. Blackstone President and COO Jon Gray stated the venture holds "enormous potential" to meet "rare demand for compute"—a characterization now tested by practical construction obstacles.
