Blackstone, the colossal private owner of data centers, is injecting $5 billion in equity into a new artificial intelligence infrastructure venture, forging a strategic alliance with Google. This significant investment, announced Monday, positions the asset management titan at the forefront of the burgeoning AI hardware race.
Under the terms of the agreement, Google will furnish the newly formed, U.S.-based company with its proprietary Tensor Processing Units (TPUs). These specialized chips are engineered to accelerate the demanding computations inherent in artificial intelligence workloads. The initial deployment aims to bring 500 megawatts of compute capacity online by 2027, with ambitious plans for substantial expansion thereafter, according to a statement released by Blackstone.
“This new company possesses immense potential as it addresses the unprecedented demand for compute power,” stated Jon Gray, President and COO of Blackstone, in the release. The unnamed entity is slated to be led by Benjamin Treynor Sloss, who previously held the role of Google’s chief programs officer. A Google spokesperson declined to comment on the tech giant’s specific leadership involvement.
The Wall Street Journal had previously reported on the joint venture, citing sources familiar with the matter, indicating that Blackstone would hold a majority stake. Blackstone did not publicly disclose the venture’s ownership structure, nor did it immediately respond to requests for further comment. Reports also suggest that the joint venture has already identified potential data center locations, with some already under construction.
This move follows Blackstone’s aggressive investment strategy across the AI landscape. Just earlier this month, the firm established a similar venture with Anthropic, further solidifying its commitment to the AI infrastructure sector.
The deal highlights the intensifying competition between Google and industry leader Nvidia in the realm of AI hardware. Google’s TPUs have long been positioned as its direct counterpoint to Nvidia’s dominant Graphics Processing Units (GPUs). While Google continues to leverage Nvidia’s GPUs within its cloud infrastructure, the company has actively pursued a strategy to diminish its reliance on Nvidia by developing its own bespoke TPUs.
This trend is not unique to Google. Other major cloud providers, such as Amazon Web Services, have also embarked on developing their own custom semiconductor chips. Google was an early adopter of in-house hardware development, having manufactured its first TPU in 2015.
Unlike the general-purpose nature of GPUs, which are well-suited for a broad spectrum of computing tasks, Google emphasizes that its TPUs are purpose-built for highly efficient processing of a more specialized set of AI applications, particularly those involving agentic AI. Google’s own Gemini AI model, for instance, operates on its TPUs. The company’s client roster for TPUs also includes prominent organizations like Anthropic and Citadel Securities.
Nvidia’s GPUs, initially designed for graphics rendering and gaming, operate by parallelizing complex computational problems. The surge in demand for GPUs was significantly amplified following the launch of OpenAI’s ChatGPT in 2022, propelling Nvidia to become the world’s most valuable company in 2024.
In a notable development earlier this month, Google briefly surpassed Nvidia in market capitalization, underscoring its strong positioning in the AI ecosystem. Analysts suggest that Google’s parent company, Alphabet, remains strategically well-placed in the AI sector, benefiting from its internal AI development capabilities, extensive distribution network, and its robust cloud division. This latest venture with Blackstone signals a clear intent to further solidify its hardware prowess and capitalize on the explosive growth in AI compute demand.
Original article, Author: Tobias. If you wish to reprint this article, please indicate the source:https://aicnbc.com/21841.html
