Google and Blackstone AI Data Center Joint Venture Faces Delays Across Multiple Locations
nashnova research
Google and Blackstone's $5 billion AI joint venture Project Braid has stalled at several data-center sites, with on-time delivery odds dropping from 90% to roughly 50% — exposing the hard supply-chain and power bottlenecks behind AI's buildout boom.
What is Project Braid, and why does it matter?
Blackstone committed $5 billion to a joint venture with Google called Project Braid, targeting 500 megawatts of capacity by 2027 to lease Google's AI processors to outside customers.
This means → Project Braid is not just another data-center deal. It is Google's key channel for pushing its TPUs — tensor processing units, its custom AI chips — into a broader market to compete directly with Nvidia's ecosystem.
In plain terms = Google wants more people running its chips instead of Nvidia's, but first it needs enough physical facilities to house them.
Where exactly is the project stuck?
Cheyenne, Wyoming: Google originally hired developer Crusoe to build the site, then took over the project and reapplied for permits after losing confidence in Crusoe's ability to deliver. Local officials told residents the project would be scaled back.
A separate site was abandoned entirely because its transformer configuration did not meet requirements. A McKinsey report notes that lead times for transformers and similar equipment now stretch to nearly one year.
Texas Governor Greg Abbott ordered a pause on new data-center approvals to gather information on how developers and tech companies pay for power — disrupting some of Project Braid's plans in the state.
On-time delivery odds fell from 90% to 50% — what does that signal?
Project Braid executives have accepted a new reality: the probability of on-time delivery has dropped from 90% three years ago to roughly 50% today.
A JPMorgan report from May found that over 60% of data-center capacity originally slated for 2027 completion has not yet broken ground, with supply-chain shortages and labor bottlenecks as the main constraints.
This reflects an industry-wide problem: the bottleneck is not money or demand — it is transformers, construction labor, and permitting that cannot keep pace with the capital being deployed.
How is the venture responding?
CEO Benjamin Treynor Sloss said in a statement that sites moving at different speeds is "very common." The venture has a "diversified site strategy" and still expects to hit the 500-megawatt target next year.
Project Braid has identified 29 potential sites and recently sought new-site support from partners. Blackstone has also engaged advisors and allies to help find additional locations.
In plain terms = the playbook is "cast a wide net" — since any single site can stall, run enough candidates in parallel to offset delay risk with volume.
What does this mean for Google and Blackstone respectively?
Google expects capital expenditure of up to $205 billion this year. The joint-venture structure shifts part of the spending and risk off its balance sheet. But the core goal is expanding TPU reach — right now, emerging cloud providers like CoreWeave run only Nvidia chips, and Google urgently needs more facilities to host its own.
Blackstone treats Project Braid as the flagship bet for BXN1, its new AI investment unit, and a critical step toward becoming the world's largest digital-infrastructure financier.
This means → whether the delays can be absorbed without jeopardizing the 2027 delivery commitment is a direct test of both companies' strategic credibility.
市场有风险，内容仅供研究参考，不构成投资建议。