Latest / Minds, Bodies, and Terawatts / Google and Blackstone team up on $5B AI cloud and chip infrastructure company

Transcript

Google and Blackstone team up on $5B AI cloud and chip infrastructure company · Minds, Bodies, and Terawatts

• 0:00
  [Host] This is Minds, Bodies, and Terawatts — where the cloud gets a new landlord. Business Insider reported today that Google and Blackstone are forming a five billion dollar joint venture focused on AI cloud infrastructure and chips. That's five billion in equity from the world's largest alternative asset manager, paired with Google's entire cloud stack.
• 0:24
  [Guest] This is private equity planting its flag in the compute layer. Blackstone already owns QTS Realty, one of the biggest data center operators in North America. This deal gives them a direct pipeline into Google's AI customer base while Google offloads capital risk to a partner sitting on over a trillion dollars in assets.
• 0:49
  [Host] Google guided 175 to 185 billion in capital spending this year. Why do they still need a partner?
• 0:57
  [Guest] Because they're not the only ones writing checks. Morgan Stanley estimated in March that Amazon, Google, Meta, and Microsoft are collectively spending roughly 630 billion on AI infrastructure in 2026. At that scale, even Alphabet wants to share the load.
• 1:15
  [Host] So who gets hurt first?
• 1:17
  [Guest] Independent cloud providers. CoreWeave just went public. Lambda Labs has been raising aggressively. These companies built their business renting Nvidia GPU clusters to startups that couldn't get allocations during the chip shortage.
• 1:31
  [Host] And now Google walks in with Blackstone's checkbook.
• 1:35
  [Guest] When you can build dedicated AI compute at five-billion-dollar scale, pricing power shifts overnight. A startup choosing between a Google-backed facility with guaranteed chip supply and a smaller provider scrambling for its next Nvidia shipment, that decision makes itself.
• 1:53
  [Host] The chip angle here matters too.
• 1:55
  [Guest] Google already designs its own TPUs, now in their sixth generation with Trillium. A venture that explicitly includes a chip push means deeper vertical integration. The silicon and the cloud become one bundled offering that locks customers in tighter than renting commodity GPUs ever could.
• 2:16
  [Host] What does Blackstone get beyond financial returns?
• 2:19
  [Guest] A toll road. They've spent years buying data centers and energy assets because infrastructure is sticky. Once a company migrates AI workloads onto your hardware, switching costs are brutal.
• 2:32
  [Host] Google also gets to match Microsoft, which has Stargate with OpenAI and SoftBank.
• 2:38
  [Guest] Each hyperscaler is assembling its own financial coalition because the buildout exceeds any single balance sheet. The IEA's Electricity 2025 report projected that global data center electricity demand will more than double by 2030, driven almost entirely by AI. That kind of energy footprint requires capital from Wall Street, not just tech earnings.
• 3:02
  [Host] So the compute layer is consolidating into a few joint ventures backed by the deepest capital pools on earth. Once those facilities are built, can anyone catch up?
• 3:14
  [Guest] The power contracts run decades. Chip supply agreements lock in for years. Customer migration costs compound the longer workloads sit on your hardware.
• 3:25
  [Host] Five years out, what's the defining question?
• 3:28
  [Guest] It won't be who trained the best model. It'll be who owns the physical layer those models depend on. That's what Patrick Deglon calls the Brain layer in Unscarcity, where whoever controls the compute substrate holds the leverage.
• 3:45
  [Host] For more on how infrastructure concentration reshapes AI, head to unscarcity dot ai. The book is on Amazon, the audiobook is on Spotify. This has been Minds, Bodies, and Terawatts.