Aleria plans 8,640 Blackwell GPUs and 28 Vera Rubin racks

By Marcus Bennett

Aleria will deploy 8,640 Nvidia Grace Blackwell Ultra GPUs in the United States and 28 racks of Nvidia DGX Vera Rubin NVL72 in the United Arab Emirates, according to the UAE-based sovereign AI infrastructure provider.

The US build will operate in a 25MW data center, which Aleria has not disclosed. The company also plans to expand the US configuration from 8,640 GPUs to 16,000. Aleria says the UAE Vera Rubin deployment will be among the first in the region. It has not disclosed details of the data center that will house those racks.

DDN's high-performance storage architecture will support both deployments. Eric Leandri, Aleria's CEO, said: “We did not come to market with a promise. We came with working infrastructure. Our sovereign AI structure is live in the UAE, running national workloads, and the results are what is driving this expansion.”

Leandri described Blackwell Ultra in the US and DGX Vera Rubin in the UAE as the next chapter of infrastructure the company has already proven. His wording places the announcement after a working UAE deployment. The company says that structure already runs national workloads.

Deployment footprint

The 25MW figure defines the facility's power capacity, but not its location. Aleria has not identified the site. In April 2026, Aleria Technologies LLC signed an agreement with Bitcoin miner Bitari to convert a cryptomining data center in Wheeler, Texas, into an AI facility.

That agreement provides a named conversion project linked to Aleria. It does not, by itself, confirm that Wheeler will host the 25MW Blackwell deployment. The announcement therefore leaves the US site's location unresolved.

Aleria serves governments and enterprises from its own data centers. Customers can also bring the solution on-premises, the company says. Aleria has not provided details about its existing data-center footprint. Public comparisons between that operating base and the announced configurations remain limited.

Storage and caching

In a separate announcement this week, DDN said European AI cloud company Nebul selected DDN Infinia for storage and key-value cache to help its GPUs serve AI models. The selection combines storage with a cache that makes processed model context reusable across Nebul's GPU fleet.

Ankur Arora, DDN's senior regional director for the Middle East and Africa, said the collaboration combines “the performance and scale required for AI-first environments, where organizations cannot afford to compromise on security, reliability, or data sovereignty.”

“Every time a GPU rebuilds context it has already processed, it is spending cycles repeating work,” said Alex Bouzari, CEO and co-founder at DDN. “By making that context available across Nebul's GPU fleet, Infinia helps deliver faster inference and more productive use of GPU capacity while keeping data within the sovereign environment.”

“AI sovereignty should not come at the expense of performance,” said Arnold Juffer, CEO of Nebul. “With DDN Infinia, we can reuse model context across our GPU infrastructure instead of spending valuable GPU cycles recomputing it. That makes inference faster and more efficient while keeping the complete AI workload within the sovereign environment.”

What does this change for AI buyers?

For Aleria's government and enterprise customers, the announced change is more compute capacity in two jurisdictions, backed by DDN storage. The 16,000-GPU figure remains a future expansion target. The 25MW US location, UAE facility details and existing operating footprint remain undisclosed, and Aleria gave no deployment timetable.

For Nebul, the claimed benefit is operational rather than geographic. Reusing context across GPUs avoids repeated computation and increases the useful work each GPU can perform. In both cases, delivery will determine the commercial result: the US and UAE builds must materialize, the announcement does not say whether Wheeler becomes the 25MW site, and Aleria must reach 16,000 GPUs if that target is to count as capacity.