AI leader Anthropic has reportedly begun leasing its own data center capacity across the United States.
This marks a significant strategic shift toward what's known as a 'hybrid infrastructure' model. Instead of relying solely on renting computing power from public clouds like Google Cloud, Anthropic is building its own capacity in leased facilities. This gives them more control and potentially better economics for their massive, ever-growing AI workloads. But why is this happening now? A few key events created the perfect conditions for this move.
First, the necessary capital has been secured. Just before this news, private equity giants Apollo and Blackstone arranged a massive $35 billion financing platform to scale Anthropic's computing power. This provides the financial muscle needed to sign long-term data center leases and purchase expensive hardware.
Second, and perhaps most importantly, Google changed its business model. In April 2026, Alphabet announced it would begin selling its powerful, custom-designed AI chips—TPUs—directly to select customers for use in their own data centers. This policy change was a game-changer. It allows Anthropic to build its own infrastructure using the same cutting-edge Google technology it relies on, all without severing its deep partnership with Google.
Third, the supply chain is locked in. Google already has a long-term agreement with Broadcom to develop and supply these TPUs through 2031, ensuring a stable source of the critical silicon needed for these new data centers.
Finally, there's a regulatory backdrop. U.S. regulators have been scrutinizing the close relationships between major cloud providers and AI labs. By diversifying its infrastructure, Anthropic can demonstrate a degree of independence, which may help ease regulatory concerns even as it seeks more funding from Google.
For Alphabet, this signals a major evolution. While it might see a slight slowdown in cloud revenue growth from Anthropic, it opens up a new, significant revenue stream from hardware sales. This shift is already reflected in Alphabet's soaring Capex, as it invests heavily to build out capacity for both its own cloud and for its partners' new data centers.
- TPU (Tensor Processing Unit): A custom-designed microchip developed by Google specifically for accelerating AI and machine learning workloads.
- Hybrid Infrastructure: An IT environment that combines public cloud services with private infrastructure, such as leased or owned data centers, to run applications.
- Capex (Capital Expenditure): Funds used by a company to acquire, upgrade, and maintain physical assets such as property, buildings, and equipment.