# AI Compute Deals

> One row per AI compute deal since January 2023: cloud capacity, chip supply, data center leases, power and the equity that pays for them. Each row gives buyer, seller, dollar value, gigawatts, chips, term and status, and every number links to the SEC filing, company release or report it came from. Where sources disagree, the row shows both figures.

The sourced ledger of AI compute deals. 82 deals, each field linked to its public source, last verified 2026-11-09.
Open data under CC BY 4.0. When citing, credit "AI Compute Deals (aicomputedeals.com)" and link the deal page.

## Data
- [Every AI compute deal, in one sourced ledger](https://aicomputedeals.com/): the full comparison table
- [Data (JSON)](https://aicomputedeals.com/data.json): every record with sources
- [Data (CSV)](https://aicomputedeals.com/data.csv): one row per deal
- [Methodology](https://aicomputedeals.com/methodology/): what each field means and how it is verified

## Deals
- [Amazon adds $4B to Anthropic; AWS becomes primary training partner](https://aicomputedeals.com/anthropic-amazon-investment-2024-11/): Amazon invested a further $4 billion in Anthropic, bringing its total to $8 billion. Anthropic named AWS its primary training partner and committed to train future models on Trainium2.
- [Amazon invests $50B in OpenAI; AWS deal expanded by $100B](https://aicomputedeals.com/openai-aws-expansion-amazon-50b-2026-02/): Amazon agreed to invest $50 billion in OpenAI ($15 billion first, $35 billion on conditions). OpenAI expanded its $38 billion AWS deal by $100 billion over eight years and will use about 2 GW of Trainium3 and Trainium4 capacity.
- [Amazon invests up to $4B in Anthropic; AWS becomes primary cloud](https://aicomputedeals.com/anthropic-amazon-investment-2023-09/): Amazon agreed to invest up to $4 billion in Anthropic for a minority stake. Anthropic named AWS its primary cloud provider and agreed to use AWS Trainium and Inferentia chips for future models.
- [Amazon–Constellation 20-year, 690 MW Calvert Cliffs nuclear PPA](https://aicomputedeals.com/amazon-constellation-calvert-cliffs-2026-09/): Amazon signed a 20-year, 690 MW PPA plus a retail supply agreement in PJM, supporting over $3 billion of investment at Calvert Cliffs in Maryland, including a 190 MW uprate due 2030–2032.
- [Amazon–Talen 1,920 MW nuclear supply from Susquehanna](https://aicomputedeals.com/amazon-talen-susquehanna-ppa-2025-06/): Talen agreed to supply 1,920 MW from its Susquehanna nuclear plant to AWS data centers in Pennsylvania through 2042. Talen expects about $18 billion in revenue over the contract at full volume.
- [AMD–HUMAIN $10B collaboration (500 MW over five years)](https://aicomputedeals.com/humain-amd-10b-2025-05/): AMD and HUMAIN plan to invest up to $10 billion to deploy up to 500 MW of AMD-based AI compute across Saudi Arabia and the US over five years.
- [AMD, Cisco and HUMAIN joint venture (up to 1 GW by 2030)](https://aicomputedeals.com/humain-amd-cisco-jv-1gw-2025-11/): AMD, Cisco and HUMAIN plan a joint venture to deploy up to 1 GW of AI infrastructure by 2030, starting with a 100 MW phase in Saudi Arabia.
- [Anthropic $50B US data centers with Fluidstack](https://aicomputedeals.com/anthropic-fluidstack-50b-us-datacenters-2025-11/): Anthropic will invest $50 billion in US computing infrastructure, building data centers with Fluidstack in Texas and New York. First sites come online through 2026.
- [Anthropic commits $110B+ to AWS; Amazon invests up to $25B more](https://aicomputedeals.com/anthropic-amazon-aws-110b-2026-04/): Anthropic committed to spend at least $110 billion on AWS between May 2026 and April 2036, for up to 5 GW of Trainium capacity. Amazon will invest $5 billion now and up to $20 billion more tied to milestones.
- [Anthropic rents SpaceX's Colossus compute (up to $120.5B through 2029)](https://aicomputedeals.com/anthropic-spacex-colossus-2026-05/): Anthropic took all of the compute at SpaceX's Colossus 1 data center in Memphis, more than 300 MW and over 220,000 NVIDIA GPUs, and later added Colossus II capacity. Anthropic's S-1 puts the SpaceXAI agreements at up to $120.5 billion through 2029, about $3.1 billion a month once fully delivered in December 2026.
- [Anthropic–AMD: up to 2 GW of MI450 Helios racks; AMD invests up to $5B](https://aicomputedeals.com/anthropic-amd-helios-2gw-2026-07/): Anthropic agreed to deploy up to 2 GW of AMD Instinct MI450-series GPUs in AMD Helios racks, starting in the first half of 2027. AMD will invest up to $5 billion in Anthropic.
- [Anthropic–CoreWeave multi-year cloud agreement](https://aicomputedeals.com/anthropic-coreweave-multiyear-2026-04/): CoreWeave signed a multi-year agreement to run production-scale Claude workloads for Anthropic on its cloud platform, with capacity coming online in phases from later in 2026.
- [Anthropic–Google Cloud up to 1 million TPUs](https://aicomputedeals.com/anthropic-google-tpu-1m-2025-10/): Anthropic expanded its use of Google Cloud to up to one million TPUs, a deal worth tens of billions of dollars. It brings well over a gigawatt of capacity online in 2026.
- [Anthropic–Google–Broadcom multi-gigawatt TPU deal from 2027](https://aicomputedeals.com/anthropic-google-broadcom-tpu-2026-04/): Anthropic signed a deal with Google and Broadcom for multiple gigawatts of next-generation TPU capacity from 2027, mostly in the US. A Broadcom SEC filing put its part at about 3.5 GW, contingent on Anthropic's continued commercial success.
- [Anthropic–Lambda compute agreements (~$41.3B)](https://aicomputedeals.com/anthropic-lambda-41b-2026-08/): Anthropic signed agreements with Lambda in May, June and August 2026 for NVIDIA GPU compute worth about $41.3 billion over terms into 2034. The August agreement, about $34.9 billion, covers a Hut 8-built Texas campus leased by NVIDIA.
- [Anthropic–Microsoft–NVIDIA: $30B Azure commitment, up to $15B investment](https://aicomputedeals.com/anthropic-microsoft-nvidia-azure-2025-11/): Anthropic committed to buy $30 billion of Azure compute, initially up to 1 GW on NVIDIA Grace Blackwell and Vera Rubin systems. NVIDIA and Microsoft committed to invest up to $10 billion and $5 billion in Anthropic.
- [Anthropic–Nscale GPU services agreements (up to ~$44.6B)](https://aicomputedeals.com/anthropic-nscale-44b-2026-08/): Anthropic signed GPU services agreements with Nscale worth up to about $44.6 billion over terms into 2033, reported as about 460 MW at an Nscale campus in West Virginia due online by end-2027.
- [Anthropic–OnePWR $3.4B gas generators (1.7 GW)](https://aicomputedeals.com/anthropic-onepwr-gas-generators-2026-09/): Anthropic agreed to spend about $3.4 billion on 1.7 GW of gas-fired generators from Texas supplier OnePWR Solutions, taking over orders OnePWR had placed with Caterpillar, according to the Financial Times. The deal includes engineering and construction to install the equipment.
- [Anthropic's $121.3B minimum Google Cloud TPU hosting commitment](https://aicomputedeals.com/anthropic-google-cloud-hosting-121b-2026-10/): Anthropic's IPO filing disclosed a March 2026 CSP hosting agreement with Google, amended through October, committing it to spend at least about $121.3 billion between April 2026 and July 2033. It also expects $125–150 billion more for TPU compute delivered 2028–2031.
- [Applied Digital 210 MW, 15-year Delta Forge 2 lease with a hyperscaler](https://aicomputedeals.com/hyperscaler-applied-digital-delta-forge-2-2026-06/): Applied Digital signed a 15-year, 210 MW lease at Delta Forge 2 with an unnamed investment-grade hyperscaler, worth about $5.2 billion over the base term.
- [Applied Digital–CoreWeave 15-year leases at Polaris Forge 1](https://aicomputedeals.com/applied-digital-coreweave-polaris-forge-2025-06/): CoreWeave signed two roughly 15-year leases for 250 MW at Applied Digital's Ellendale, North Dakota campus, worth about $7 billion in contracted revenue. CoreWeave later took a further 150 MW, lifting the total to 400 MW and about $11 billion.
- [AWS–Cipher Mining $5.5B 15-year lease (300 MW)](https://aicomputedeals.com/aws-cipher-mining-lease-2025-11/): Cipher Mining signed a 15-year lease worth about $5.5 billion to provide AWS with 300 MW of turnkey space and power for AI workloads. Delivery runs in two phases from July 2026, with rent from August 2026.
- [Cipher Mining–Fluidstack 168 MW Barber Lake lease, Google backstop](https://aicomputedeals.com/cipher-fluidstack-google-barber-lake-2025-09/): Cipher Mining signed a 10-year AI hosting deal with Fluidstack for 168 MW of critical IT load at Barber Lake, Texas, worth about $3 billion in contracted revenue. Google backstops $1.4 billion of Fluidstack's lease and took warrants for about 5.4% of Cipher.
- [CoreWeave exercises 260 MW Helios Phase II option](https://aicomputedeals.com/coreweave-galaxy-helios-phase2-2025-05/): CoreWeave exercised its option for about 260 MW more critical IT load at Galaxy's Helios campus on 15-year terms, with deliveries from 2027.
- [CoreWeave takes the rest of Helios (526 MW critical IT in total)](https://aicomputedeals.com/coreweave-galaxy-helios-full-2025-08/): CoreWeave exercised its final option for another 133 MW at Helios, committing to all 800 MW of approved gross power (526 MW critical IT) under 15-year leases.
- [CoreWeave–Galaxy Digital 15-year, 133 MW Helios lease](https://aicomputedeals.com/coreweave-galaxy-helios-2025-03/): CoreWeave signed a 15-year lease for 133 MW of critical IT load at Galaxy's Helios campus in Dickens County, Texas. Galaxy delivered the full 133 MW in Q2 2026.
- [CoreWeave's $9B all-stock bid for Core Scientific (rejected)](https://aicomputedeals.com/coreweave-core-scientific-acquisition-2025-07/): CoreWeave agreed to buy data center operator Core Scientific in an all-stock deal valued at about $9 billion to own the power and sites behind its leases. Core Scientific shareholders voted it down on 30 October 2025 and the merger was terminated.
- [Figure–Nscale $3.5B compute deal (up to 100,000 Vera Rubin GPUs)](https://aicomputedeals.com/figure-nscale-3-5b-2026-09/): Figure signed a multi-year deal with Nscale: an initial $3.5 billion compute commitment, intended to scale past $6 billion, for up to 100,000 NVIDIA Vera Rubin GPUs in Barstow, Texas, from the second half of 2027. Nscale also invests in Figure and becomes its preferred compute provider.
- [Google rents ~110,000 NVIDIA GPUs from SpaceX at $920M a month](https://aicomputedeals.com/google-spacex-cloud-2026-06/): SpaceX signed a cloud services agreement giving Google about 110,000 NVIDIA GPUs plus CPUs and memory. Google pays $920 million a month from October 2026 through June 2029, with a reduced fee while capacity ramps.
- [Google–Constellation 20-year PPA backing 890 MW of nuclear uprates](https://aicomputedeals.com/google-constellation-uprates-ppa-2026-10/): Google signed a 20-year PPA that lets Constellation invest over $4.3 billion to add about 890 MW across 11 nuclear units in Illinois, Pennsylvania and New Jersey, with the first uprate by 2028. Google also signed a 15-year, 2,700 MW load-following supply agreement in PJM.
- [Google–Kairos Power 500 MW small modular reactor fleet](https://aicomputedeals.com/google-kairos-smr-2024-10/): Google signed a master plant development agreement to buy power from a fleet of Kairos Power advanced reactors. The first unit is targeted for 2030 and the fleet for 500 MW by 2035.
- [Google–NextEra 25-year PPA to restart Duane Arnold nuclear plant](https://aicomputedeals.com/google-nextera-duane-arnold-2025-10/): Google signed a 25-year power purchase agreement for the 615 MW Duane Arnold plant in Iowa, shut since 2020. NextEra targets a restart as soon as early 2029, subject to approvals.
- [HUMAIN expands NVIDIA deal to up to 600,000 GPUs](https://aicomputedeals.com/humain-nvidia-600k-gpus-2025-11/): At the U.S.-Saudi Investment Forum, HUMAIN said it will deploy up to 600,000 NVIDIA GPUs, including GB300, across Saudi Arabia and the US over three years.
- [HUMAIN–NVIDIA AI factories in Saudi Arabia (up to 500 MW)](https://aicomputedeals.com/humain-nvidia-500mw-2025-05/): HUMAIN, owned by Saudi Arabia's PIF, will build AI factories of up to 500 MW with several hundred thousand NVIDIA GPUs over five years. The first phase is an 18,000-GPU GB300 supercomputer.
- [IREN signs $2.8B of new multi-year AI cloud contracts](https://aicomputedeals.com/ai-developers-iren-2-8b-2026-07/): IREN signed new multi-year AI cloud contracts worth $2.8 billion with AI developers, including a new unnamed leading AI developer, and raised its year-end 2026 ARR target to over $4 billion.
- [Jane Street–CoreWeave ~$6B AI cloud commitment](https://aicomputedeals.com/jane-street-coreweave-6b-2026-04/): Trading firm Jane Street committed about $6 billion to CoreWeave's AI cloud, including NVIDIA Vera Rubin capacity across multiple sites. Jane Street also bought $1 billion of CoreWeave stock at $109 a share.
- [Meta–AMD 6 GW GPU deal with 160M-share warrant](https://aicomputedeals.com/meta-amd-6gw-2026-02/): Meta agreed to deploy up to 6 GW of AMD Instinct GPUs, starting with 1 GW of MI450 in the second half of 2026. AMD issued Meta a performance-based warrant for up to 160 million AMD shares, mirroring the OpenAI deal.
- [Meta–Blue Owl $27B Hyperion data center joint venture](https://aicomputedeals.com/meta-blue-owl-hyperion-2025-10/): Meta and Blue Owl-managed funds formed a joint venture to build and own the Hyperion AI campus in Richland Parish, Louisiana, about $27 billion of development. Blue Owl owns 80%, Meta 20%, and Meta leases the campus with a 16-year residual value guarantee.
- [Meta–Constellation 20-year PPA for Clinton nuclear plant](https://aicomputedeals.com/meta-constellation-clinton-ppa-2025-06/): Meta signed a 20-year power purchase agreement for the clean energy attributes of Constellation's Clinton Clean Energy Center in Illinois. The deal keeps the plant running after state subsidies end in 2027 and adds a 30 MW uprate.
- [Meta–CoreWeave expanded deal (~$21B through 2032)](https://aicomputedeals.com/meta-coreweave-21b-2026-04/): CoreWeave signed an expanded agreement to provide Meta about $21 billion of dedicated AI cloud capacity through December 2032, including early NVIDIA Vera Rubin deployments. CoreWeave says the commitment was signed in March 2026.
- [Meta–CoreWeave order (up to $14.2B through 2031)](https://aicomputedeals.com/meta-coreweave-14b-2025-09/): Meta committed to pay CoreWeave up to about $14.2 billion for AI cloud capacity through 14 December 2031, with an option to expand. CoreWeave disclosed the order in a Form 8-K.
- [Meta–Google Cloud six-year deal (over $10B)](https://aicomputedeals.com/meta-google-cloud-2025-08/): Meta agreed to rent compute, storage and networking from Google Cloud in a six-year deal worth more than $10 billion, first reported by The Information.
- [Meta–Nebius five-year GPU clusters (~$3B)](https://aicomputedeals.com/meta-nebius-3b-2025-11/): Nebius agreed to provide Meta two dedicated GPU clusters over five years, deployed in December 2025 and February 2026. The 6-K gives a total contract value of about $2.9 billion; Nebius's results release rounded it to $3 billion.
- [Meta–Nebius five-year Vera Rubin deal (up to ~$27B)](https://aicomputedeals.com/meta-nebius-27b-2026-03/): Nebius will give Meta $12 billion of dedicated NVIDIA Vera Rubin capacity over five years from early 2027, and Meta commits to buy up to $15 billion of unsold capacity on certain Nebius clusters.
- [Meta–NVIDIA multiyear deal for Blackwell, Rubin GPUs and Grace CPUs](https://aicomputedeals.com/meta-nvidia-multiyear-2026-02/): Meta signed a multiyear, multigenerational deal for millions of NVIDIA Blackwell and Rubin GPUs, Grace CPUs (the first large Grace-only rollout) and Spectrum-X networking. Vera CPUs could deploy at scale in 2027.
- [Meta–Oracle cloud commitment (reported ~$20B)](https://aicomputedeals.com/meta-oracle-cloud-2025-12/): Oracle's fiscal Q2 2026 results said remaining performance obligations rose $68 billion to $523 billion, highlighted by new commitments from Meta, NVIDIA and others. Reuters and Bloomberg had reported in September 2025 that Oracle and Meta were negotiating a cloud deal worth about $20 billion.
- [Microsoft extends OpenAI partnership (third investment)](https://aicomputedeals.com/microsoft-openai-investment-2023-01/): Microsoft announced a multiyear, multibillion-dollar investment in OpenAI, its third after 2019 and 2021. Azure became OpenAI's exclusive cloud provider for research, products and API services.
- [Microsoft invests $1.5B in G42](https://aicomputedeals.com/microsoft-g42-investment-2024-04/): Microsoft made a $1.5 billion strategic investment in Abu Dhabi's G42, which runs its AI applications and services on Azure. Brad Smith joined G42's board.
- [Microsoft winds down its Oracle AI cloud capacity rental](https://aicomputedeals.com/microsoft-oracle-bing-capacity-winddown-2026-11/): Bloomberg reported that Microsoft did not renew its 2023 deal to rent AI capacity on Oracle Cloud Infrastructure for Bing and has been switching the capacity off, because the older NVIDIA GPUs no longer meet its specifications. Oracle said demand for its AI infrastructure remains strong.
- [Microsoft–Brookfield 10.5 GW renewable energy framework](https://aicomputedeals.com/microsoft-brookfield-renewables-2024-05/): Brookfield and Microsoft signed a five-year framework for Brookfield to deliver over 10.5 GW of new renewable capacity in the US and Europe between 2026 and 2030.
- [Microsoft–Constellation 20-year PPA to restart Three Mile Island Unit 1](https://aicomputedeals.com/microsoft-constellation-crane-ppa-2024-09/): Microsoft signed a 20-year power purchase agreement for the output of Three Mile Island Unit 1, renamed the Crane Clean Energy Center. The 835 MW unit is due to restart around 2028 to match Microsoft's data center load in PJM.
- [Microsoft–IREN $9.7B GB300 cloud contract](https://aicomputedeals.com/microsoft-iren-childress-2025-11/): IREN signed a five-year, $9.7 billion contract to give Microsoft NVIDIA GB300 capacity at its Childress, Texas campus, with a 20% prepayment. IREN is buying the GPUs and equipment from Dell for about $5.8 billion.
- [Microsoft–Lambda multibillion-dollar GPU deal](https://aicomputedeals.com/microsoft-lambda-gb300-2025-11/): Lambda agreed a multi-year, multibillion-dollar deal to deploy tens of thousands of NVIDIA GPUs, including GB300 NVL72 systems, for Microsoft. Value and GPU count were not disclosed.
- [Microsoft–Nebius GPU capacity from Vineland, NJ](https://aicomputedeals.com/microsoft-nebius-vineland-2025-09/): Nebius signed a five-year deal to supply Microsoft dedicated GPU capacity from its Vineland, New Jersey data center. Base value is $17.4 billion, rising to $19.4 billion with extra services.
- [Microsoft–Nscale ~200,000 NVIDIA GB300 GPUs (US and Europe)](https://aicomputedeals.com/microsoft-nscale-gb300-2025-10/): Nscale contracted to deliver about 200,000 NVIDIA GB300 GPUs to Microsoft across Texas, Portugal, the UK and Norway. The largest block is about 104,000 GPUs at a 240 MW Texas campus from Q3 2026.
- [NVIDIA guarantees up to $105B of OpenAI leases at SB Energy's PORTS campus (4.25 GW)](https://aicomputedeals.com/openai-sb-energy-nvidia-ports-2026-08/): NVIDIA agreed guarantees capped at $105 billion backing an OpenAI affiliate's 20-year leases for about 4.25 GW of IT load at SB Energy's PORTS campus in Pike County, Ohio. The site will host NVIDIA AI infrastructure exclusively.
- [NVIDIA invests $2B in CoreWeave; 5 GW AI factory framework](https://aicomputedeals.com/nvidia-coreweave-2b-5gw-2026-01/): NVIDIA bought $2 billion of CoreWeave Class A stock at $87.20 a share. The companies set a framework for CoreWeave to build more than 5 GW of AI factories on NVIDIA platforms by 2030.
- [NVIDIA invests $30B in OpenAI's $110B round](https://aicomputedeals.com/nvidia-openai-30b-investment-2026-02/): NVIDIA committed $30 billion to OpenAI's funding round alongside SoftBank ($30 billion) and Amazon ($50 billion). The round, announced at $110 billion, later closed at $122 billion.
- [NVIDIA–CoreWeave $6.3B unused-capacity order](https://aicomputedeals.com/nvidia-coreweave-capacity-backstop-2025-09/): CoreWeave disclosed a $6.3 billion initial order from NVIDIA. NVIDIA must buy any of CoreWeave's capacity its own customers leave unused, through 13 April 2032.
- [OpenAI–AMD 6 GW GPU deal with 160M-share warrant](https://aicomputedeals.com/openai-amd-6gw-2025-10/): OpenAI agreed to deploy 6 GW of AMD Instinct GPUs, starting with 1 GW of MI450 in the second half of 2026. AMD issued OpenAI a warrant for up to 160 million AMD shares at $0.01, vesting with deployment and share-price milestones.
- [OpenAI–AWS $38B seven-year compute deal](https://aicomputedeals.com/openai-aws-38b-2025-11/): OpenAI signed a $38 billion, seven-year agreement to run workloads on AWS using NVIDIA GB200 and GB300 GPUs on EC2 UltraServers. All capacity is targeted for deployment before the end of 2026.
- [OpenAI–Broadcom 10 GW custom accelerators](https://aicomputedeals.com/openai-broadcom-10gw-2025-10/): OpenAI and Broadcom will deploy 10 GW of OpenAI-designed AI accelerators and Broadcom Ethernet networking. Racks start deploying in the second half of 2026, with all 10 GW by the end of 2029.
- [OpenAI–Cerebras 750 MW inference compute (over $10B)](https://aicomputedeals.com/openai-cerebras-750mw-2026-01/): OpenAI agreed to buy up to 750 MW of Cerebras-powered compute for low-latency inference, delivered in tranches from 2026 to 2028. Cerebras's IPO prospectus values the deal at more than $20 billion.
- [OpenAI–CoreWeave expansion (up to $4B)](https://aicomputedeals.com/openai-coreweave-expansion-2025-05/): OpenAI expanded its CoreWeave contract by up to $4 billion. CoreWeave disclosed the add-on alongside its first quarterly results as a public company.
- [OpenAI–CoreWeave expansion (up to $6.5B; $22.4B total)](https://aicomputedeals.com/openai-coreweave-expansion-2025-09/): CoreWeave expanded its OpenAI contract by up to $6.5 billion to train OpenAI's next-generation models. Total contract value with OpenAI rose to about $22.4 billion.
- [OpenAI–CoreWeave initial cloud contract (up to $11.9B)](https://aicomputedeals.com/openai-coreweave-initial-2025-03/): OpenAI signed its first compute contract with CoreWeave, worth up to $11.9 billion. OpenAI also received a $350 million equity stake in CoreWeave ahead of its IPO.
- [OpenAI–Microsoft recapitalization: $250B incremental Azure commitment](https://aicomputedeals.com/openai-microsoft-azure-250b-2025-10/): As part of OpenAI's recapitalization, OpenAI contracted to buy an incremental $250 billion of Azure services. Microsoft gave up its right of first refusal on OpenAI compute and holds about 27% of OpenAI Group PBC, valued at about $135 billion.
- [OpenAI–NVIDIA 10 GW letter of intent (up to $100B investment)](https://aicomputedeals.com/openai-nvidia-10gw-loi-2025-09/): OpenAI and NVIDIA signed a letter of intent for at least 10 GW of NVIDIA systems, with NVIDIA intending to invest up to $100 billion in OpenAI as each gigawatt is deployed. The first gigawatt was targeted for the second half of 2026 on Vera Rubin.
- [OpenAI–Oracle 4.5 GW Stargate capacity](https://aicomputedeals.com/openai-oracle-stargate-2025-07/): OpenAI and Oracle agreed to develop 4.5 GW of additional Stargate data center capacity in the US. The Wall Street Journal later reported the contract at $300 billion over about five years starting in 2027.
- [Oracle discloses cloud contract worth over $30B a year from FY2028](https://aicomputedeals.com/oracle-openai-30b-annual-contract-2025-06/): Oracle's Form 8-K said it signed multiple large cloud deals, including one expected to bring in more than $30 billion in annual revenue starting in fiscal 2028. The customer was not named in the filing; it was later tied to OpenAI's Stargate capacity.
- [Oracle–AMD 50,000 MI450 GPU supercluster](https://aicomputedeals.com/oracle-amd-mi450-supercluster-2025-10/): Oracle will offer a public AI supercluster of 50,000 AMD Instinct MI450 GPUs on AMD Helios racks starting in calendar Q3 2026, with more planned for 2027 and beyond.
- [Samsung and SK hynix join Stargate (DRAM supply LOIs)](https://aicomputedeals.com/openai-samsung-skhynix-stargate-memory-2025-10/): Samsung and SK hynix signed letters of intent with OpenAI to supply high-performance DRAM for Stargate, targeting up to 900,000 wafers a month. The companies also agreed to explore data centers in Korea.
- [SpaceX (xAI) ~$40B GPU finance lease with Apollo-managed lessor](https://aicomputedeals.com/spacex-apollo-gpu-finance-lease-2026-10/): SpaceX signed a binding commitment for a master finance lease under which an Apollo-managed entity buys GPU pods and related equipment and leases them to SpaceX for its AI data centers.
- [SpaceX hosting deal worth $1.11B a month (customer undisclosed)](https://aicomputedeals.com/spacex-undisclosed-compute-customer-2026-09/): SpaceX's CFO said the company closed a new AI hosting deal in early September worth about $1.11 billion a month, roughly $13 billion a year, with billing from 1 December 2026. SpaceX did not name the customer.
- [Stargate adds five US sites (nearly 7 GW planned)](https://aicomputedeals.com/stargate-five-new-sites-2025-09/): OpenAI, Oracle and SoftBank named five new Stargate sites in Texas, New Mexico, Ohio and the Midwest. With Abilene and CoreWeave projects, planned capacity reached nearly 7 GW and over $400 billion of investment over three years.
- [Stargate Norway (Nscale, Aker, OpenAI) in Narvik](https://aicomputedeals.com/stargate-norway-nscale-aker-2025-07/): Nscale, Aker and OpenAI announced Stargate Norway near Narvik: 230 MW of renewable-powered capacity, with up to 290 MW more, targeting 100,000 NVIDIA GPUs by end-2026. In 2026 Nscale took sole control of the site and added a Microsoft expansion there.
- [Stargate Project launch (OpenAI, SoftBank, Oracle, MGX)](https://aicomputedeals.com/stargate-project-launch-2025-01/): OpenAI announced Stargate, a new company to invest $500 billion over four years in US AI infrastructure for OpenAI, with $100 billion deployed immediately. SoftBank, OpenAI, Oracle and MGX are the initial equity funders, and construction was already under way in Texas.
- [Stargate UAE 1 GW cluster (G42, OpenAI, Oracle, NVIDIA, Cisco, SoftBank)](https://aicomputedeals.com/stargate-uae-1gw-2025-05/): G42 will build a 1 GW Stargate UAE compute cluster inside Abu Dhabi's 5 GW UAE–U.S. AI Campus, operated by OpenAI and Oracle on NVIDIA GB300 systems. The first 200 MW is due in 2026.
- [Stargate UK (Nscale, OpenAI, NVIDIA)](https://aicomputedeals.com/stargate-uk-nscale-openai-2025-09/): Nscale, OpenAI and NVIDIA set up Stargate UK. OpenAI will explore offtake of up to 8,000 NVIDIA GPUs in Q1 2026, potentially scaling to 31,000, across UK sites including Cobalt Park in the North East AI Growth Zone.
- [TeraWulf–Fluidstack 200 MW hosting, Google backstop](https://aicomputedeals.com/terawulf-fluidstack-google-2025-08/): TeraWulf signed two 10-year colocation deals with Fluidstack for over 200 MW of IT load at Lake Mariner, New York, worth about $3.7 billion. Google backstops $1.8 billion of Fluidstack's lease obligations and got warrants for about 8% of TeraWulf.
- [xAI $20B GPU financing via SPV for Colossus 2 (NVIDIA-backed)](https://aicomputedeals.com/xai-colossus2-gpu-spv-2025-10/): xAI arranged about $20 billion through a special-purpose vehicle that buys NVIDIA GPUs and leases them to xAI for five years for Colossus 2 in Memphis. Reports put it at about $7.5 billion equity, including up to $2 billion from NVIDIA, and up to $12.5 billion debt.
- [xAI–HUMAIN 500 MW+ data center in Saudi Arabia](https://aicomputedeals.com/xai-humain-500mw-2025-11/): HUMAIN and xAI will develop data centers in Saudi Arabia, anchored by a flagship facility of 500 MW or more with xAI as first customer. The first cluster is about 18,000 NVIDIA GB300 GPUs for training Grok.

## Companies
- [OpenAI AI compute deals](https://aicomputedeals.com/company/openai/): 21 deals
- [Anthropic AI compute deals](https://aicomputedeals.com/company/anthropic/): 14 deals
- [CoreWeave AI compute deals](https://aicomputedeals.com/company/coreweave/): 14 deals
- [Microsoft AI compute deals](https://aicomputedeals.com/company/microsoft/): 11 deals
- [Meta AI compute deals](https://aicomputedeals.com/company/meta/): 10 deals
- [Amazon AI compute deals](https://aicomputedeals.com/company/amazon/): 8 deals
- [Google AI compute deals](https://aicomputedeals.com/company/google/): 8 deals
- [NVIDIA AI compute deals](https://aicomputedeals.com/company/nvidia/): 7 deals
- [AMD AI compute deals](https://aicomputedeals.com/company/amd/): 6 deals
- [Oracle AI compute deals](https://aicomputedeals.com/company/oracle/): 6 deals
- [HUMAIN AI compute deals](https://aicomputedeals.com/company/humain/): 5 deals
- [Nscale AI compute deals](https://aicomputedeals.com/company/nscale/): 5 deals
- [Constellation Energy AI compute deals](https://aicomputedeals.com/company/constellation-energy/): 4 deals
- [SpaceX AI compute deals](https://aicomputedeals.com/company/spacex/): 4 deals
- [Fluidstack AI compute deals](https://aicomputedeals.com/company/fluidstack/): 3 deals
- [Galaxy Digital AI compute deals](https://aicomputedeals.com/company/galaxy-digital/): 3 deals
- [Nebius AI compute deals](https://aicomputedeals.com/company/nebius/): 3 deals
- [Applied Digital AI compute deals](https://aicomputedeals.com/company/applied-digital/): 2 deals
- [Broadcom AI compute deals](https://aicomputedeals.com/company/broadcom/): 2 deals
- [Cipher Mining AI compute deals](https://aicomputedeals.com/company/cipher-mining/): 2 deals
- [G42 AI compute deals](https://aicomputedeals.com/company/g42/): 2 deals
- [IREN AI compute deals](https://aicomputedeals.com/company/iren/): 2 deals
- [Lambda AI compute deals](https://aicomputedeals.com/company/lambda/): 2 deals
- [xAI AI compute deals](https://aicomputedeals.com/company/xai/): 2 deals
- [Apollo Global Management AI compute deals](https://aicomputedeals.com/company/apollo-global-management/): 1 deal
- [Blue Owl AI compute deals](https://aicomputedeals.com/company/blue-owl/): 1 deal
- [Brookfield Renewable AI compute deals](https://aicomputedeals.com/company/brookfield-renewable/): 1 deal
- [Cerebras AI compute deals](https://aicomputedeals.com/company/cerebras/): 1 deal
- [Core Scientific AI compute deals](https://aicomputedeals.com/company/core-scientific/): 1 deal
- [Figure AI compute deals](https://aicomputedeals.com/company/figure/): 1 deal
- [Jane Street AI compute deals](https://aicomputedeals.com/company/jane-street/): 1 deal
- [Kairos Power AI compute deals](https://aicomputedeals.com/company/kairos-power/): 1 deal
- [NextEra Energy AI compute deals](https://aicomputedeals.com/company/nextera-energy/): 1 deal
- [OnePWR Solutions AI compute deals](https://aicomputedeals.com/company/onepwr-solutions/): 1 deal
- [Samsung Electronics AI compute deals](https://aicomputedeals.com/company/samsung-electronics/): 1 deal
- [SB Energy AI compute deals](https://aicomputedeals.com/company/sb-energy/): 1 deal
- [SK hynix AI compute deals](https://aicomputedeals.com/company/sk-hynix/): 1 deal
- [SoftBank AI compute deals](https://aicomputedeals.com/company/softbank/): 1 deal
- [Stargate AI compute deals](https://aicomputedeals.com/company/stargate/): 1 deal
- [Talen Energy AI compute deals](https://aicomputedeals.com/company/talen-energy/): 1 deal
- [TeraWulf AI compute deals](https://aicomputedeals.com/company/terawulf/): 1 deal

## Optional
- [All records in one file](https://aicomputedeals.com/llms-full.txt)
- [Sitemap](https://aicomputedeals.com/sitemap.xml)