Microsoft: AI deals, investors and valuation

FT: Microsoft gives up the option that would have more than doubled the Texas contract (tens of billions); the existing contract remains.

Microsoft places $25 billion of bonds in ten series, its first US issue since 2017, as it ramps up data center spending.

Microsoft gives up acting as anchor customer of the Loughton (Essex) site, about $3 billion, due to delays in the power connection. Approximate date.

Azure remains the primary cloud but OpenAI can sell all its products on any cloud; Microsoft's license through 2032 becomes non-exclusive. Microsoft stops paying revenue share, while OpenAI continues paying it through 2030 with a cap.

During Nadella's Sydney visit Microsoft announced A$25B by end-2029 for cloud and AI capacity (+140%), security and skills, tied to an MOU with the Australian government.

After OpenAI's exit, Microsoft takes the capacity at Nscale's Narvik site, with more than 30,000 Nvidia Vera Rubin GPUs.

Microsoft will invest $10bn in Japan in 2026-2029 in AI infrastructure, cybersecurity and skills; SoftBank and Sakura Internet will supply GPU compute via Azure.

Microsoft joins the additional tranche of the round closed at an $852 billion post-money valuation; its contribution was not disclosed.

Samsung is negotiating 3-5 year memory deals with Google and Microsoft; Microsoft's prepayment could exceed $10bn, deducted if purchases fall short.

Nscale takes over the 8 GW Monarch campus and Microsoft signs a letter of intent for 1.35 GW of Vera Rubin compute.

Korean media report SK Hynix is the sole supplier of HBM3E for Microsoft's Maia 200 AI accelerator.

After meeting Modi, Nadella announced $17.5B in India for 2026-2029, Microsoft's largest in Asia, including a Hyderabad cloud region and sovereign cloud offerings.

Microsoft announced C$19B in Canada for 2023-2027, including over C$7.5B in the next two years to expand Azure regions, plus a five-point sovereignty plan.

The Information: Microsoft is discussing with Broadcom moving the design of its Maia chips, currently handled by Marvell.

Part of the three-way deal with Nvidia (up to $10 billion, covered elsewhere). Claude comes to Microsoft Foundry and Copilot.

Counterpart to the Nvidia-Microsoft investment: Anthropic commits to buy $30 billion of Azure compute, with additional capacity of up to 1 GW.

Microsoft will invest about $10B at Start Campus in Sines with Nscale and Nvidia, deploying 12,600 GB300 GPUs at a campus planned for 1.2 GW.

Microsoft and G42 add 200 MW of capacity via G42 unit Khazna, part of Microsoft's $15.2B UAE investment through 2029.

Five-year AI cloud contract at IREN's Childress campus (Texas).

Multiyear contract for AI infrastructure with tens of thousands of GPUs (including GB300 NVL72); amount undisclosed.

In the October 2025 recapitalization OpenAI commits to purchase an additional $250 billion of Azure services; Microsoft loses its right of first refusal on compute.

OpenAI becomes a public benefit corporation; Microsoft holds about 27% on a diluted basis (worth ~$135 billion at the time), with IP rights to models and products through 2032.

Orders signed between September 2025 and April 2026 running to 2033 (Texas, Portugal, UK, Norway); initially ~200,000 GB300s for ~$23 billion.

During Trump's state visit Microsoft pledged £22B (about $30B) in the UK for 2025-2028. Roughly half is capex for AI infrastructure, including the Nscale-built supercomputer at Loughton (23,040 Nvidia GB300 GPUs).

GPU capacity from Nebius's data center in Vineland (New Jersey), over 100,000 GB300; up to $19.4 billion with additional services.

In Bengaluru Satya Nadella announced $3B over two years to expand Azure and AI capacity in India.

Microsoft takes a minority stake in G42; G42 uses Azure for its AI services.

Microsoft's cumulative investments in OpenAI from 2019 to 2023 (including the multibillion-dollar tranche of January 2023), largely in the form of Azure credits.