Google and Blackstone Forge $25 Billion AI Cloud Joint Venture to Rival CoreWeave

Blackstone is pouring $5 billion in equity into a new AI cloud company that Google will operate, a deal that hands the private equity giant majority ownership of a business built on Google's proprietary tensor processing units. Including leverage, the total investment reaches $25 billion, according to a person familiar with the matter cited by Bloomberg and the Straits Times. By 2027, the venture targets 500 megawatts of computing capacity, a scale that would place it in direct competition with neocloud providers like CoreWeave and Nebius Group.

The structure is unusual. Google, one of the world's largest cloud providers, is effectively spinning out infrastructure capacity into a vehicle controlled by its financial partner rather than building it on Alphabet's balance sheet. Veteran Google executive Benjamin Treynor Sloss will lead the new entity as chief executive. The arrangement suggests Google sees strategic value in separating this capital-intensive buildout from its core reporting, even as it retains operational control through chip design and technical leadership.

CoreWeave built its business renting Nvidia GPUs to AI startups desperate for compute. Google's joint venture with Blackstone enters that same market with a different formula: custom silicon rather than commodity chips, and a balance sheet backed by the world's largest alternative asset manager. The competitive dynamic shifts from a simple GPU arms race to a contest over total cost of ownership and vertical integration.

The timing matters. Neoclouds have raised billions on the premise that hyperscalers move too slowly for AI-native workloads. Google is now signaling it can match their speed and capital access without sacrificing its own chip roadmap. For CoreWeve, Nebius, and Lambda, this represents a direct challenge from a player with both manufacturing scale and now dedicated institutional capital. The Straits Times noted the project adds to a broader boom in computing infrastructure spending that underpins AI models and services across the industry.

The $5 billion equity commitment from Blackstone is the headline number, but the $25 billion total including leverage tells the real story. This is a leveraged infrastructure play dressed in AI clothing, structurally closer to a data center REIT or fiber rollout than a typical venture investment. Blackstone's Infrastructure and Private Equity divisions have deployed similar templates across energy, logistics, and telecom; here they are applying it to compute capacity.

For Google, the deal preserves capital. Alphabet has spent heavily on data centers and faces ongoing pressure to show discipline. By bringing in Blackstone as majority owner, Google funds expansion without expanding its capital expenditure line in the same way. The risk sits with Blackstone's investors, who gain exposure to AI demand growth through a yield-oriented structure. If utilization lags or chip economics shift unfavorably, the leverage amplifies downside.

The new business will run Google's tensor processing units, or TPUs, the home-grown chips that power its internal AI models and cloud offerings. This is not a generic cloud resale arrangement. Google has been expanding its AI chip lineup and hunting for additional data center capacity to house them, as reported by the Straits Times. The Blackstone joint venture gives those chips a dedicated, scaled deployment vehicle outside Alphabet's wholly owned footprint.

TPUs have historically served Google's internal needs and select cloud customers. A standalone, Blackstone-backed operator changes the go-to-market. It suggests Google believes its silicon can compete as a standalone product in the open market, not merely as a bundled component of Google Cloud. Whether outside AI labs and enterprises will commit to TPUs over Nvidia's CUDA ecosystem remains an open question, but the joint venture structure provides the capital base to find out.

Blackstone's involvement fits a larger pattern. Privateequitywire reported in early May that Blackstone and KKR were in discussions with Google to provide AI capabilities across portfolio companies through broad, portfolio-wide agreements. The Google-Blackstone cloud venture appears to be the infrastructure backbone for that strategy, not merely a financial investment.

The structure would allow firms to deploy AI capabilities across dozens of portfolio companies simultaneously, rather than negotiating separate deals for each business. Other buyout groups including Thoma Bravo and Vista Equity Partners have pursued similar AI integration strategies, but the Google arrangement stands out for its direct access to models and now dedicated compute capacity. For the private equity industry, this represents a shift from advisory AI consulting to owning the underlying infrastructure, with all the recurring revenue and capital intensity that implies.

The 2027 capacity target gives the venture roughly 18 months to scale, an aggressive timeline for data center construction and power procurement. Regulatory scrutiny is likely, particularly in Europe and the US, where concentrated AI infrastructure ownership draws increasing attention. Google will need to navigate its own cloud business competing with its partly owned offspring, a tension that could shape pricing and customer access.

For the industry, the deal validates the neocloud model while potentially undercutting it. If Google can offer TPU-based compute at scale with Blackstone's cost of capital, standalone GPU renters face margin pressure. The question becomes whether AI workloads actually shift to TPUs in volume, or whether Nvidia's ecosystem stickiness preserves the status quo. Either way, the capital commitment signals that infrastructure investors view AI compute as a generational opportunity worth betting billions on, even at the risk of overbuild.