Groq hands LPU distribution in the Middle East to Saudi-based Aljammaz Technologies

Groq, the developer of LPU chips built for inference, has found a distributor for the MENA region. On 17 October 2025, at the GITEX Global exhibition in Dubai, the company announced a partnership with Aljammaz Technologies. Aljammaz is a Saudi value-added distributor based in Riyadh that has been in business since 1996. It will sell both the GroqCloud cloud platform and on-premises GroqRack clusters through a network of system integrators and resellers. The parties did not disclose financial terms, shipment volumes or launch dates.

What exactly Aljammaz gets

The press release refers to a "full set" of Groq inference solutions. It includes two products:

• GroqCloud is a full-stack cloud platform where models are available through an API.
• GroqRack is a compute cluster installed at the customer's site, an option for those who do not want to, or cannot, move data to a public cloud.
• Sales channel — Aljammaz's integrators and resellers in Saudi Arabia and neighboring countries.
• Target audience — enterprises, government organizations and developers.
• Industries named in the release — financial services, healthcare and smart-city initiatives.

The on-premises option matters more here than it might seem. In a region where data sovereignty is often a separate line in the technical requirements, the ability to put a rack in your own data center can decide a tender.

Why the bet on inference

Groq's argument rests on a shift in the market. According to the company, organizations are moving from experiments to production use, and there the main constraints become response speed, cost and power consumption. By Groq's account, the LPU was designed for inference from the ground up rather than adapted from graphics processors. For that reason, the company claims, it can power real-time applications that were impractical on conventional GPU architectures.

This is a marketing frame, and the release contains no independent measurements. It gives no figures for latency, tokens per second or energy efficiency compared with GPUs. Claims of lower latency, cost and power consumption remain claims until the first regional deployments.

The company also notes that Groq founder Jonathan Ross invented the Tensor Processing Unit at Google. Groq itself was founded in 2016.

The distributor's side of the deal

Aljammaz positions itself as a leader among the region's technology distributors. The company works in data centers, cloud, networking and cybersecurity, and its partners include Dell Technologies, HPE, Cisco and Alibaba Cloud. CEO Asim Aljammaz calls the deal part of an AI-First strategy and cites "almost 30 years" of experience in the region.

For Groq, this is a way to scale without building its own army of salespeople. A distributor with established relationships among integrators covers what a startup struggles to build quickly: local distribution, service, and experience with corporate and government procurement. Fahad AlTurief, Groq's vice president and managing director for MENA, describes the goal as helping organizations in the region deploy AI without barriers of latency and cost.

What to watch next

For IT directors and integrators in MENA, the practical takeaway is simple: Groq now has an official purchasing channel, which means you can request pilots and commercial offers through local partners rather than directly from the US. Before budgeting for anything, demand comparative benchmarks from the vendor on your own models and workloads: latency, throughput, rack power consumption and the list of supported models.

The key questions are still open: pricing, GroqRack delivery times, the availability of local capacity, and support. The answers will show whether the agreement becomes a working sales channel or remains a declaration of intent made at a trade show.