Think Secures Over $8M in MENA’s Largest AI Infrastructure Pre-Seed Round

Saudi artificial intelligence infrastructure startup Think has raised more than $8 million in a pre-seed funding round to expand a platform designed to make deploying and operating AI systems more efficient.

The round was co-led by Saudi venture capital firms RAED Ventures and Wa’ed Ventures, with participation from the venture capital arm of Dhahran Techno Valley and a group of strategic angel investors.

Think described the transaction as the largest AI infrastructure and deep-tech pre-seed round completed in the Middle East and North Africa. That designation is based on the company’s announcement and has not been independently verified through a comprehensive database of every private funding round in the region.

Funding Targets Commercial Expansion

The Riyadh-based company plans to use the new capital to hire employees, increase manufacturing capacity and continue developing its hardware and software. Funding will also support commercial deployments in Saudi Arabia and the company’s planned expansion into other Gulf Cooperation Council markets.

Think expects to broaden its GCC presence over the next 18 months while pursuing selected international markets. The company is already working on proofs of concept, production deployments and partnerships in Saudi Arabia.

Participation from the three institutional investors gives the young company financial backing from firms familiar with Saudi Arabia’s technology ecosystem. RAED Ventures says it invests in early-stage companies across MENA, while Wa’ed Ventures describes itself as a $500 million fund backed by Aramco.

Think Combines Hardware and Orchestration Software

Founded by Ahmed AlSharif and Ammar Enaya, Think is developing an integrated system that combines high-density, liquid-cooled computing hardware with software used to manage graphics processing units, or GPUs. AlSharif previously held roles at Meta, Sony PlayStation Europe and Electronic Arts, while Enaya’s background includes positions at Cisco, HPE Aruba and cybersecurity company Vectra AI.

Its hardware product, known as the AI Node, is built around multi-GPU computing infrastructure. Think pairs those nodes with a proprietary bare-metal orchestration layer called ILM, which is intended to distribute workloads and make more of the installed computing capacity available for AI training and inference.

AI infrastructure has become an increasingly important part of the technology stack as companies move beyond experimenting with generative AI and begin operating models in production. Large-scale deployments require computing capacity, cooling, electricity, networking and software capable of coordinating expensive processors without leaving significant resources idle.

Think is positioning its platform as an alternative for enterprises and government organizations that want greater control over their computing environments and data. Its systems can be installed in data centers, offices, laboratories and edge locations, according to the company, reducing reliance on centralized public cloud infrastructure for some workloads.

The startup said its technology works with widely available GPUs rather than requiring customers to purchase a single type of proprietary inference chip. Future versions are expected to support processors from different vendors within the same environment, potentially giving customers more flexibility when configuring systems for model training and inference.

Think Targets Greater AI Efficiency at Scale

Think reported that its platform maintained GPU utilization above 90% during production benchmark testing, compared with what it described as an industry range of 30% to 50%. The company also claimed that its cost per million tokens was nearly 10 times lower than the average cost of using frontier models offered by Google, OpenAI and Anthropic.

Cost comparisons between locally operated infrastructure and externally hosted AI models can vary considerably. Hardware acquisition, depreciation, electricity, cooling, technical staff, utilization levels and model performance all influence the actual cost of processing AI workloads. Public model providers, meanwhile, charge different rates depending on the model and the volume of input and output tokens.

Think’s next phase will focus on demonstrating how its reported efficiency gains translate across different customer environments and AI workloads. Expanding proofs of concept into recurring commercial deployments could strengthen the company’s position as enterprises look for more flexible and locally controlled AI infrastructure.

Investors Back Sovereign AI Infrastructure

The financing reflects growing investor interest in the infrastructure underpinning AI rather than only the applications built on top of it. Demand for GPUs, power and data center capacity has created opportunities for companies attempting to improve how computing resources are cooled, allocated and used.

RAED Ventures General Partner Wael Nafee said: "The next generation of AI leaders will be defined not only by the models they build, but by the infrastructure that makes AI practical, affordable and sovereign. Think is tackling one of AI's biggest challenges with technology that improves efficiency while giving organisations greater control over their AI capabilities. We believe the team is building a category-defining company from Saudi Arabia with global potential."

Wa’ed Ventures CEO Anas Algahtani added: "Saudi Arabia has a unique opportunity not only to adopt AI, but to build the infrastructure that powers it. Think is resolving one of the industry's biggest challenges by making AI deployment more efficient, scalable and sovereign, and we're proud to support its next stage of growth."

Dhahran Techno Valley Chief Investment Officer Faizan Baig commented: "Sovereign and efficient AI infrastructure is foundational to every country's AI ambitions. Think is tackling one of the sector's most pressing challenges by helping organisations deploy and scale AI while maintaining control over cost, security and data."

Saudi Arabia’s emphasis on locally controlled AI creates a potentially supportive market for Think, particularly among organizations that face data residency, security or operational control requirements. Sovereignty, however, does not depend on hardware location alone. Customers must also consider software dependencies, chip supply, cybersecurity, model ownership and access to specialized technical expertise.

With the pre-seed round completed, Think now has additional capital to move from technical development toward wider commercial execution.