Think launches Think Grid™, a new alternative to hyperscaler AI compute

• Think Grid provides cloud-like access to Think Fabric's high-efficiency AI infrastructure, without the upfront capital investment of deploying dedicated infrastructure
• Think Grid offers dedicated AI compute at an all-inclusive rate up to 27% lower than comparable hyperscaler configurations, including storage, support and data egress
• Think Grid launches with NVIDIA PRO 6000 Blackwell accelerators and Think's ILM orchestration platform, with support for additional silicon architectures planned

RIYADH, Saudi Arabia, Aug. 31, 2026 /PRNewswire/ -- Think, the AI infrastructure company building for the Age of Efficiency, today announced Think Grid, a new global heterogeneous AI compute service giving enterprises, governments and AI-native organisations access to dedicated, high-performance AI compute without the capital expenditure or operational complexity traditionally associated with deploying and managing dedicated AI infrastructure.

Hosted in a Tier III facility in Riyadh, Think Grid makes the performance and efficiency of Think Fabric available as dedicated capacity through a simple monthly subscription. Alongside Think Fabric's deployable infrastructure model, Grid gives organisations a choice in how they access Think's technology: deploy it within their own infrastructure or consume it as a fully managed, hosted service with no upfront capital investment.

At launch, each Think Grid SuperNode combines four NVIDIA PRO 6000 Blackwell accelerators with 384GB of total VRAM, alongside Think's proprietary ILM orchestration software, high-speed networking, storage, cooling, support and maintenance. ILM pools GPU memory into a unified memory space and orchestrates workloads across available compute, allowing multiple models to share the same infrastructure rather than leaving individual accelerators underutilised.

In production testing, ILM-managed nodes run multiple production models on hardware that conventional deployments dedicate to one, while Think's bare-metal architecture delivers faster time to first token and faster tokens per second while parallelizing training, compared to containerised equivalents.

Unlike traditional cloud infrastructure, where customers typically rent virtualised compute capacity and pay separately for services such as storage, data egress and support, Think Grid provides an entire dedicated bare-metal node for an all-inclusive cost, with no capital outlay or egress charges.

Think Grid is already available to government and enterprise customers in Saudi Arabia, with infrastructure hosted in Riyadh, and is being rolled out globally. For Saudi customers, the Riyadh deployment enables organisations to run AI workloads while keeping their data within the Kingdom, with Think Grid managing the underlying hardware, cooling, maintenance and infrastructure operations.

Analysis of published August 2026 hyperscaler pricing shows Think Grid's all-inclusive monthly rate is up to 27% lower than comparable dedicated Blackwell AI compute configurations, before accounting for storage, data egress and support charges that are typically billed separately.

"For too long, organisations have effectively had two choices when they need serious AI compute: make a significant capital investment in infrastructure of their own, or rent capacity from a hyperscaler and accept the complexity and variable costs that come with it," said Ahmed AlSharif, Founder and CEO of Think. "Think Grid creates another option. By providing the Think Fabric as a service, we're giving customers dedicated, state-of-the-art AI compute, with the software, hardware and operations bonded together under one predictable monthly rate."

At the heart of Think Grid is Think Fabric, Think's unified AI infrastructure platform, which combines its SuperNode-class hardware and ILM orchestration technology. While Think Fabric enables organisations to deploy and operate Think infrastructure directly, Think Grid provides the technology as hosted capacity, removing the need for customers to build or manage the underlying physical infrastructure themselves.

Think Grid can support production inference, pre-training, post-training, AI agents and voice applications, and multi-model workloads.

"The economics of AI infrastructure need to become much simpler," said Ammar Enaya, Co-Founder and COO of Think. "Customers shouldn't need to piece together compute, storage, networking, orchestration, and support and then try to predict what the final bill will look like. With Think Grid, we've packaged the infrastructure required to run production AI into a single service, with a clear cost and dedicated capacity."

Think Grid is initially available to organisations in Saudi Arabia on a monthly rolling basis, with annual pricing options also available. Think is now expanding Grid availability into additional markets, with organisations globally able to register their interest at think-ai.com/grid.

About Think
Think is a Saudi AI infrastructure company building the engine room for the age of efficiency. Its unified approach brings together intelligent software and high-performance hardware to make AI compute more efficient, accessible and economically sustainable.

Think's technology addresses the fundamental bottlenecks limiting AI infrastructure today, including GPU utilisation, memory, cooling, power and cost. Through its Think Fabric infrastructure platform and Think Grid compute service, Think enables enterprises, governments and AI-native organisations to deploy high-performance AI with greater efficiency, flexibility and control.

Founded by Ahmed AlSharif, a technology leader whose experience spans Meta, PlayStation, and EA, and enterprise technology veteran Ammar Enaya, formerly of Cisco, Aruba, and HPE, Think is headquartered in Riyadh, Saudi Arabia.

Find out more at www.think-ai.com and https://www.linkedin.com/company/think-ai-labs/

View original content to download multimedia:https://www.prnewswire.com/news-releases/think-launches-think-grid-a-new-alternative-to-hyperscaler-ai-compute-302864578.html

SOURCE Think