Google and Blackstone form £4bn AI cloud venture for TPU compute

TL;DR:

• Alphabet’s Google and Blackstone will form a joint AI cloud venture aimed at meeting demand for AI compute infrastructure, with Blackstone committing an initial $5bn (£4bn) of equity.
• The venture aims to bring 500MW of new data centre capacity online in 2027, with total investment potentially reaching $25bn (£20bn) including leverage, according to Bloomberg News.
• The US-based business will offer Google’s custom Tensor Processing Units (TPUs) on a compute-as-a-service basis, with Google’s long-time executive Benjamin Sloss as CEO.

The Google-Blackstone deal is one of the largest single AI infrastructure commitments yet announced. It also signals the maturing of the Big-Tech-plus-private-capital model that has become the dominant funding pattern for the data centre build-out underpinning frontier AI.

Blackstone — the world’s largest alternative asset manager — will provide the equity and capital-structure backbone; Google contributes the chips, software stack and operational expertise. The 500MW first wave in 2027 puts the venture in the same scale bracket as the largest single hyperscale campuses now under construction in the US, with capacity expansion expected over time.

A play on TPU demand and capital intensity

Google’s TPUs have become a credible alternative to Nvidia GPUs for AI training and inference, with Anthropic among the high-profile customers anchoring TPU demand. By offering TPUs through a compute-as-a-service model — rather than only via Google Cloud’s own platform — the venture broadens the addressable market and gives organisations alternative procurement routes.

“This isn’t the biggest headline number we’ve seen. But it’s a high-quality bet on sustainable growth in AI infrastructure,” said Brittain Ladd, an AI and supply chain consultant at Florida-based Chang Robotics. Blackstone President Jon Gray framed the partnership as a response to “rising demand for AI infrastructure and the need for large-scale capital deployment.”

Total Big Tech AI infrastructure spending in 2026 is projected to exceed $700bn (£560bn) — a figure that explains why structures like the Google-Blackstone venture, blending operational expertise with deep-pocketed private capital, are becoming the default rather than the exception.

Looking forward

For the UK, the venture has two indirect but important implications. First, on-grid energy: AI data centre energy demand is now the dominant new-build pressure on European grids, and a US 500MW TPU campus puts further demand on energy supply chains and power-purchase markets that UK operators also compete in. Second, capacity geography: a US-headquartered TPU-as-a-service venture concentrates frontier AI compute even further on the US side of the Atlantic, putting more pressure on UK government commitments to sovereign AI compute (most recently via the AI Growth Zones programme). Whether UK financial-services and public-sector AI workloads can be served by US-based TPU capacity — or whether data-residency requirements force a parallel UK build-out — will be the policy question Whitehall has to confront over the next 18 months.