Microsoft Corp. plans to grow its data center footprint in the United Arab Emirates significantly through a partnership with a local company.

The tech giant announced the initiative today alongside a second major infrastructure investment. Microsoft disclosed that it has commissioned billions of dollars worth of data center capacity from Lambda Labs Inc., a venture-backed cloud startup. Both projects will use Nvidia Corp.’s cutting-edge GB300 artificial intelligence chip.

In 2023, Microsoft partnered with an Abu Dhabi-based company called Group42 to grow its UAE data center presence. The tech giant disclosed today that it will have invested more than $7.3 billion in the partnership by the end of the year. Over half the sum was allocated to capital expenditures, a line item that includes data center infrastructure. 

The project saw Microsoft deploy Nvidia Corp. chips with computing capacity equivalent to 21,500 H100 graphics cards. The H100 was Nvidia’s flagship AI accelerator until 2023. The chipmaker’s newer Blackwell chip, which debuted last March, provides about 30 times more performance for large language model inference.

The investment announced today will see Microsoft spend another $7.9 billion in the UAE. As part of the investment, the company will nearly quadruple its local data center infrastructure’s computing capacity to the equivalent of 81,900 H100 chips.

Microsoft President Brad Smith wrote in a blog post that the tech giant has secured the U.S. Commerce Department approvals necessary to ship the new GPUs to the UAE. Some of the processors will be GB300 superchips, the most advanced AI accelerators currently sold by Nvidia. Each GB300 includes one central processing unit and two Blackwell Ultra graphics cards, each of which provide 50% more performance than a standard Blackwell B200.

Microsoft plans to further grow its UAE infrastructure in the near future. According to the Financial Times, the company expects to secure additional GPU export licenses within the next 12 months. Furthermore, Microsoft estimates that the infrastructure investments will be accompanied by $2.4 billion in local operating expenses and the cost of goods sold.

The other data center investment the company announced today will take the form of a partnership with Lambda, the operator of a cloud platform optimized for AI workloads. In February, the startup disclosed that its cloud included more than a quarter million GPUs. The same month, it raised $480 million from a consortium that includes Nvidia.

The newly announced partnership will see Lambda build several billions dollars’ worth of AI infrastructure for Microsoft. The project will involve tens of thousands of graphics cards, some of which are set to be deployed in GB300 NVL72 appliances. Each GB300 NVL72 includes 36 GB300 chips.

Microsoft previously signed a similar AI infrastructure partnership with CoreWeave Inc., a publicly traded Lambda competitor. Last November, The Information reported that Microsoft was expecting to spend $10 billion on CoreWeave’s platform by the end of the decade.

Photo: Lambda