Alphabet Inc.'s Google has entered into an agreement with the world’s largest alternative asset manager, Blackstone Group LP, to establish a new artificial intelligence cloud computing company. This joint venture aims to compete directly with emerging cloud service providers like CoreWeave Inc. in the rapidly expanding market. According to a statement released on Monday, Blackstone will provide an initial $5 billion in equity capital and will serve as the controlling shareholder of the new entity. Sources familiar with the matter indicate that, including leveraged financing, the total investment could reach approximately $25 billion.
The company plans to achieve 500 megawatts of computing capacity by 2027. Its data centers will be powered by Google's proprietary seventh-generation Tensor Processing Unit (TPU), codenamed "Ironwood." Officially launched in April 2025, this chip is Google's first TPU specifically designed for AI inference tasks, marking a strategic shift from a primary focus on training to the large-scale deployment of inference. The Ironwood chip delivers a peak FP8 performance of 4,614 TFLOPs, is equipped with 192GB of HBM3e memory, and offers a memory bandwidth of 7.4 TB/s, achieving approximately double the energy efficiency of its predecessor, Trillium. Its single-chip performance is comparable to NVIDIA's GB200, while offering a significant total cost of ownership advantage: internal estimates show a TCO roughly 44% lower per chip than the GB200, and 30% to 41% lower in external sales scenarios.
Google veteran Benjamin Treynor Sloss has been appointed Chief Executive Officer of the new company. This leadership decision underscores Google's commitment, placing a key executive involved in its core technology architecture at the helm to ensure seamless integration of capabilities from chip design to data center operations and cloud services.
This $25 billion partnership exemplifies the global investment surge in AI infrastructure. A recent Morgan Stanley report forecasts that capital expenditures from the top five hyperscale cloud providers will reach approximately $800 billion in 2026, rising to about $1.16 trillion by 2027, signaling the dawn of a "trillion-dollar era" for AI infrastructure investment. Analyst research indicates that in Q1 of this year, the combined capital expenditures of the top four North American cloud providers (Amazon.com, Google, Microsoft, and Meta Platforms, Inc.) totaled $131.63 billion, a year-on-year increase of 70.3%. Overall capital spending for the world's top nine cloud service providers is projected to reach $830 billion, a substantial 79% increase year-over-year.
In this expansion cycle, Blackstone's role is particularly pivotal. The firm, with over $1.3 trillion in assets under management, has long made significant bets on the data center sector. As early as 2021, Blackstone acquired data center operator QTS, and in 2024, it spent approximately 16 billion Australian dollars to acquire Australian computing services provider AirTrunk, further solidifying its position as a leading global data center provider. Recently, Blackstone completed an IPO for its data center acquisition platform, Blackstone Digital Infrastructure Trust Inc., designed to continuously acquire built and leased data center assets benefiting from the AI boom.
By partnering with Google, Blackstone is not only providing substantial capital but will also be deeply involved in the design, construction, and operation of the data centers—a path that differs from the traditional "build-and-operate" model dominated by cloud providers.
The Google-Blackstone venture further intensifies competition in the AI infrastructure space. In recent years, Google has consistently expanded its proprietary AI chip portfolio, with each generation from the first TPU to the latest seventh-generation Ironwood achieving leaps in computing power, energy efficiency, and memory bandwidth. Concurrently, Google is accelerating efforts to secure data center capacity globally, including retrofitting existing facilities, participating in new construction projects, and using joint venture models to secure long-term power and land resources. These moves aim to meet the computational demands of internal large models like Gemini, as well as the explosive external demand from enterprise customers for AI training and inference services.
The newly formed company will compete directly with so-called "neoclouds" or next-generation cloud service providers like CoreWeave and Nebius Group NV. These companies specialize in offering highly customized compute rental services for AI developers and enterprises, rapidly capturing market share outside traditional cloud providers with flexible pricing and rapid deployment capabilities. Notably, many neoclouds have capital ties to AI chip leader NVIDIA, which not only provides a stable supply of GPUs but also embeds itself deeply in their growth through investments and priority supply arrangements.
Google has emerged as one of the major beneficiaries of the current AI investment wave. Its cloud computing business revenue continues to soar, while its consumer-facing AI services are gradually gaining market acceptance.
Comments