Major AI compute deals

Value:

$35B (reported)

Type:

Capacity commitment

Term:

~6 years (reported)

Hardware:

NVIDIA chips at a Nueces County, Texas facility leased by NVIDIA and developed by Hut 8

Anthropic signed a cloud computing agreement worth approximately $35 billion over about six years with NVIDIA-backed Lambda, first reported by the WSJ. NVIDIA holds the lease on the Texas facility and supplies the chips Lambda resells to Anthropic.

Sources: Bloomberg · Reuters via Investing.com

Value:

$30B (reported)

Type:

Compute rental

Term:

October 2026 through June 2029 ($920M/month)

Hardware:

~110,000 NVIDIA GPUs plus CPUs, memory and related hardware, housed in xAI data centers

Implied rate:

~$11.46/chip-hr

Disclosed in a SpaceX regulatory filing ahead of its IPO: Google agreed to pay SpaceX $920 million per month from October 2026 through June 2029 — roughly $30 billion total — for access to about 110,000 NVIDIA GPUs housed in xAI's data centers.

Sources: CNBC · TechCrunch

Value:

over $40B

Type:

Compute rental

Term:

Through May 2029 ($1.25B/month, 90-day termination right)

Hardware:

The entire Colossus 1 cluster near Memphis — over 220,000 NVIDIA GPUs, ~300 MW

Implied rate:

~$7.78/chip-hr

Disclosed in SpaceX's S-1 filing: Anthropic agreed to pay xAI roughly $1.25 billion per month through May 2029 — potentially over $40 billion total — to lease about 300 MW of compute spanning the entire Colossus 1 data center. A rare case of one frontier lab renting a rival's cluster outright.

Sources: TechCrunch · Network World

Value:

$6B

Type:

Capacity commitment

Hardware:

NVIDIA Vera Rubin systems

Trading firm Jane Street signed a ~$6 billion AI cloud agreement with CoreWeave for capacity on NVIDIA Vera Rubin systems, alongside a $1 billion equity purchase — a sign that large-scale compute buying now extends beyond AI labs and hyperscalers.

Sources: CoreWeave

Value:

$21B (reported)

Type:

Capacity commitment

Term:

Through December 2032

Meta expanded its CoreWeave relationship with an additional commitment of approximately $21 billion for AI cloud capacity through December 2032, bringing its total CoreWeave commitments to roughly $35 billion on top of the September 2025 deal.

Sources: CNBC · AI Business

Value:

over $20B

Type:

Investment + compute

Term:

3 years, deployed in tranches 2026-2028 (option for +1.25 GW by 2030)

Hardware:

Cerebras wafer-scale inference systems, 750 MW

OpenAI committed to buy 750 MW of wafer-scale AI inference capacity from Cerebras over three years, receiving warrants for a minority equity stake. Early reports pegged the deal at over $10 billion; Cerebras later confirmed more than $20 billion.

Sources: The Information · Yahoo Finance

Value:

$30B

Type:

Investment + compute

Hardware:

NVIDIA Grace Blackwell and Vera Rubin systems; option for up to 1 GW of additional capacity

Anthropic committed to purchase $30 billion of Azure compute capacity, with an option to contract up to 1 GW more on NVIDIA Grace Blackwell and Vera Rubin systems. In the same three-way alliance, NVIDIA agreed to invest up to $10 billion in Anthropic and Microsoft up to $5 billion.

Sources: CNBC · GeekWire

Value:

$9.7B

Type:

Compute rental

Term:

5 years, deployments phased through 2026

Hardware:

NVIDIA GB300 GPUs at IREN's Childress, Texas campus (200 MW critical IT load)

IREN signed a five-year, ~$9.7 billion GPU cloud services contract — including a 20% prepayment — giving Microsoft access to NVIDIA GB300 GPUs deployed in phases through 2026 at IREN's Childress, Texas campus.

Sources: IREN · Data Center Dynamics

Value:

Undisclosed

Type:

Compute rental

Hardware:

Tens of thousands of NVIDIA GPUs, including GB300 NVL72 systems

Lambda announced a multibillion-dollar, multi-year agreement to deploy AI infrastructure for Microsoft powered by tens of thousands of NVIDIA GPUs including GB300 NVL72 systems. The exact dollar value and GPU count were not disclosed.

Sources: Lambda · CNBC

Value:

$38B

Type:

Capacity commitment

Term:

7 years

Hardware:

Hundreds of thousands of NVIDIA GPUs (GB200/GB300 via EC2 UltraServers)

OpenAI signed a $38 billion, seven-year agreement giving it access to AWS infrastructure comprising hundreds of thousands of NVIDIA GPUs for training and inference — its first major cloud deal with Amazon, enabled by the end of Microsoft's right of first refusal.

Sources: Amazon · CNBC

Value:

$250B

Type:

Capacity commitment

As part of its for-profit recapitalization — which left Microsoft with a ~27% stake — OpenAI committed to purchase an incremental $250 billion of Azure services, while Microsoft gave up its right of first refusal on new OpenAI cloud workloads.

Sources: Microsoft · CNBC

Value:

Undisclosed

Type:

Capacity commitment

Hardware:

Up to ~1 million Google TPUs; well over 1 GW coming online in 2026

Anthropic agreed to expand its use of Google Cloud TPUs in a deal both companies described as worth 'tens of billions of dollars' (no exact figure disclosed), giving it access to up to one million TPUs and bringing well over a gigawatt of capacity online in 2026.

Sources: Google Cloud · CNBC

Value:

$20B (reported)

Type:

Capacity commitment

Oracle confirmed a multiyear cloud deal to supply Meta with computing power for training and deploying AI models. Bloomberg and Reuters reported the value at about $20 billion, a figure the companies did not officially confirm.

Sources: CNBC · Reuters via Investing.com

Value:

$14.2B

Type:

Capacity commitment

Term:

Through December 2031

Hardware:

NVIDIA GB300 (Blackwell) systems

Meta signed a $14.2 billion agreement with CoreWeave running through December 2031 for AI cloud capacity on NVIDIA GB300 systems, with an option to materially expand through 2032.

Sources: CNBC · Data Center Dynamics

Value:

$300B (reported)

Type:

Capacity commitment

Term:

5 years, beginning 2027

Hardware:

NVIDIA GPUs across Stargate data center sites (4.5 GW of capacity)

OpenAI agreed to purchase roughly $300 billion of Oracle cloud compute over about five years starting in 2027, tied to 4.5 GW of new Stargate data center capacity in the US. The figure was first reported by the WSJ; the companies never published it themselves.

Sources: SiliconANGLE · Data Center Dynamics

Value:

$17.4B

Type:

Compute rental

Term:

5 years, through 2031 (options can raise total to $19.4B)

Hardware:

Dedicated NVIDIA GPU capacity from a new Vineland, New Jersey data center

Microsoft signed a five-year, $17.4 billion agreement — expandable to $19.4 billion — with Nebius for dedicated GPU infrastructure capacity from a new data center in Vineland, NJ, Nebius's first long-term contract with a major tech company.

Sources: Nebius · CNBC

Value:

over $10B

Type:

Capacity commitment

Term:

6 years

Meta signed a six-year cloud deal with Google Cloud worth more than $10 billion, covering servers, storage, networking and AI infrastructure services. The figure was reported by CNBC and Bloomberg via sources rather than announced by the companies.

Sources: CNBC · Bloomberg

Value:

up to $22.4B

Type:

Capacity commitment

Hardware:

NVIDIA Blackwell-generation GPU infrastructure

CoreWeave's initial March 2025 agreement with OpenAI was worth up to $11.9 billion and included OpenAI taking equity; expansions of up to $4 billion (May 2025) and up to $6.5 billion (September 2025) brought the cumulative contract value to about $22.4 billion.

Sources: CoreWeave · Bloomberg