Anthropic

“Anthropic's compute commitment will initially be up to one gigawatt of compute capacity with NVIDIA Grace Blackwell and Vera Rubin systems.”

Anthropic: “Anthropic’s compute commitment will initially be up to one gigawatt of compute capacity with NVIDIA Grace Blackwell and Vera Rubin systems.” The same announcement calls it “additional compute capacity up to one gigawatt”, beside the $30 billion Azure commitment, and names no counterparty for it.

“A 5 GW agreement with Google and Broadcom, which will begin coming online in 2027”

Anthropic sizes the agreement at 5 GW, with Google and Broadcom together, coming online from 2027. Broadcom’s 3.5 GW is part of it: a part and the whole, printed side by side and never added.

“The expansion will provide Anthropic with multiple gigawatts of TPU capacity, expected to come online starting in 2027. This capacity expansion will be delivered through Google Cloud services, as well as access to Google-built TPUs supplied through Broadcom.”

Google Cloud’s announcement of the same expansion: multiple gigawatts of TPU capacity from 2027, delivered through Google Cloud services and Google-built TPUs supplied through Broadcom.

“We have signed a new agreement with Google and Broadcom for multiple gigawatts of next-generation TPU capacity that we expect to come online starting in 2027.”

Anthropic’s announcement of a new agreement with Google and Broadcom for “multiple gigawatts of next-generation TPU capacity” from 2027. Broadcom’s 8-K of the same day puts about 3.5 GW of it through Broadcom, and Anthropic sized the whole at 5 GW on 6 May.

“The expansion is worth tens of billions of dollars and is expected to bring well over a gigawatt of capacity online in 2026.”

Anthropic’s own announcement of the TPU expansion Google announced the same day: worth tens of billions of dollars, with well over a gigawatt expected online in 2026. “Well over a gigawatt” is a floor, so no capacity value is recorded.

“secure up to 5 gigawatts (GW) of capacity for training and deploying Claude, including new Trainium2 capacity coming online in the first half of this year and nearly 1GW total of Trainium2 and Trainium3 capacity coming online by the end of 2026.”

Anthropic’s own announcement of the agreement Amazon announced the same day: up to 5 GW, with new Trainium2 capacity in the first half of 2026 and nearly 1 GW of Trainium2 and Trainium3 by the end of 2026.

“Anthropic is actively using Project Rainier to build and deploy its industry-leading AI model, Claude.”

AWS describes roughly 500,000 Trainium2 chips at Project Rainier, with Anthropic expected to use more than one million by the end of 2025. The announcement gives no power capacity. Amazon’s separate November 2025 announcement of 2.4 GW in Northern Indiana does not name Anthropic.

Amazon’s page metadata records publication on 24 June 2025 and an update on 29 October 2025.

“Anthropic has committed to purchase $30 billion of Azure compute capacity”

Anthropic commits to purchase $30 billion of Azure compute capacity; Microsoft published the same announcement that day. The sentence continues “and to contract additional compute capacity up to one gigawatt” without naming from whom, so that gigawatt is a separate record whose counterparty is not named.

“Anthropic has committed to purchasing $30 billion in Azure compute capacity”

Microsoft’s publication of the same announcement. Its own summary reads “committed to purchasing $30 billion in Azure compute capacity and may contract up to one gigawatt of additional compute power”, again naming no counterparty for the gigawatt.

“For example, in May 2026, we entered into Cloud Services Agreements with Anthropic PBC (“Anthropic”), an AI research and development public benefit corporation, with respect to access to compute capacity across COLOSSUS and COLOSSUS II. Compute capacity provided includes approximately 325,000 NVIDIA GPUs, backed by hyperscale-class CPUs, exabyte-scale storage and high-speed networking and interconnects purpose-built for AI workloads. Pursuant to these agreements, the customer has agreed to pay us $1.25 billion per month through May 2029, with capacity ramping in May and June 2026 at a reduced fee. After the initial three-month period, the agreements may be terminated by either party upon 90 days’ notice.”

SpaceX’s final prospectus covers Anthropic’s access across Colossus and Colossus II, a wider scope than the Colossus I facility linked here. It describes payments of about $1.25 billion per month through May 2029 for capacity that includes approximately 325,000 NVIDIA GPUs, with reduced fees during the May–June 2026 capacity ramp. After an initial three months, either party may terminate on 90 days’ notice. The filing divides neither the payments nor the GPUs between the sites.

“with respect to access to compute capacity across COLOSSUS and COLOSSUS II”

The COLOSSUS II half of the agreement SpaceX’s final prospectus records across both sites; the Colossus 1 half is its own record. The filing divides neither the $1.25 billion a month nor the approximately 325,000 NVIDIA GPUs between the sites.

“We’ve signed an agreement with SpaceX to use all of the compute capacity at their Colossus 1 data center. This gives us access to more than 300 megawatts of new capacity (over 220,000 NVIDIA GPUs) within the month.”

Anthropic: an agreement “to use all of the compute capacity at their Colossus 1 data center”, giving “more than 300 megawatts of new capacity (over 220,000 NVIDIA GPUs) within the month”. More than 300 MW is a floor, so no capacity value is recorded. SpaceX’s prospectus of 12 June puts the agreement across Colossus 1 and Colossus 2 at about 325,000 GPUs.

“Executed 401 MW Anthropic Lease at Justified Data Campus”

Justified Data Campus, Hawesville, Kentucky. TeraWulf’s Q2 2026 business update, furnished as Exhibit 99.2 to its results filing, lists 401 MW for delivery between the second half of 2027 and the first half of 2028 and puts the lease’s long-term contracted cash flows at about $19 billion over the 20-year term. TeraWulf’s 6 July announcement gave the same figure as contracted revenue over the initial lease term.

“approximately 401 MW of critical IT load for high-performance computing ('HPC') operations ... Delivery of the leased capacity is expected to occur in phases beginning in late 2027 and concluding in early 2028. ... Anthropic's obligation to pay rent under the Justified Data Campus Lease will commence upon the delivery of the applicable leased premises and continue for a term of 20 years”

Justified Data Campus, Hawesville, Kentucky. The lease names Anthropic PBC as tenant and TeraWulf subsidiary Raylan Data LLC as landlord. Two five-year renewal options could extend the term to 30 years; delivery was expected to begin in late 2027. A separate Kentucky utility proceeding specifies 482 MW of maximum contract demand for Justified DataPower LLC. That utility demand measures a different quantity from this lease’s 401 MW of critical IT load, and the utility proceeding does not name Anthropic.

“Fluidstack is announcing a $50 billion American computing infrastructure deal with Anthropic, building data centers in New York and Texas, with more sites to come. These facilities are custom built for Anthropic”

Fluidstack’s announcement of the same arrangement, the same day: a $50 billion deal building data centres in New York and Texas, “custom built for Anthropic”.

“Today, we are announcing a $50 billion investment in American computing infrastructure, building data centers with Fluidstack in Texas and New York, with more sites to come. These facilities are custom built for Anthropic”

Anthropic announces a $50 billion investment in American computing infrastructure, building data centres with Fluidstack in Texas and New York, “custom built for Anthropic”. Neither party’s announcement says who will own, operate or lease them, so the form of control is not stated.