As a powerhouse in the Gulf region, Saudi Arabia and the United Arab Emirates have joined the global artificial intelligence (AI) arms race, procuring thousands of high-performance NVIDIA chips crucial for building AI software. AI chips are currently the hottest commodity in Silicon Valley, and their supply is being squeezed by the global AI competition.
In order to accelerate economic development and transformation, Saudi Arabia and the United Arab Emirates are actively working to achieve ambitious AI plans. They have initiated their respective national AI strategies, publicly stating their aim to become global leaders in the field of AI. However, the Gulf countries’ pursuit of AI has also raised concerns about the potential misuse of this technology.
01
Saudi Arabia is building its own large-scale model
It has been revealed that Saudi Arabia has purchased at least 3000 NVIDIA H100 chips through King Abdullah University of Science and Technology, at a total value of approximately $120 million, with delivery expected by the end of 2023.
The H100 chips are priced at $40,000 each, and NVIDIA CEO Jensen Huang has described them as “the world’s first computer chips designed specifically for generative artificial intelligence.” Investigations indicate that the H100 is currently the most essential GPU (graphics processing unit) for training large language models, as it offers the fastest speeds for both inference and training of these models, typically providing the best value for inference performance as well.
Global leading technology companies are racing to acquire scarce chips for artificial intelligence development. Chinese tech giants like Tencent and Alibaba are placing orders for high-performance NVIDIA chips. Meanwhile, U.S. companies at the forefront of advanced large language model development, including Microsoft-backed OpenAI and Google, are also purchasing NVIDIA chips. According to multiple sources close to NVIDIA and its manufacturer TSMC, NVIDIA is set to ship around 550,000 H100 chips worldwide in 2023, primarily to U.S. tech companies.
King Abdullah University of Science and Technology also possesses at least 200 NVIDIA A100 chips and is in the process of building the supercomputer Shaheen III, which is scheduled to be operational this year. This machine will house 700 NVIDIA Grace Hopper superchips, specifically designed for cutting-edge artificial intelligence applications.
Sources closely related to King Abdullah University of Science and Technology have revealed that this state-owned university will use these chips to build its own large language model. Notably, many Chinese researchers are part of the responsible artificial intelligence and data analytics lab developing this model. They have chosen to work at King Abdullah University of Science and Technology due to U.S. government restrictions that prevent them from studying and working in the United States.
02
Joining the global AI competition
In order to reduce dependency on oil and diversify their economies, governments of Gulf countries have been leveraging their oil wealth to enter various industries. With soaring energy prices last year, Saudi Arabia and the United Arab Emirates reaped substantial profits, both managing the world’s largest and most active sovereign wealth funds. Their ample financial resources have enabled them to make significant investments in emerging fields like artificial intelligence and esports, seeking the “new oil”.
Saudi Arabia and the United Arab Emirates have joined the global AI arms race, procuring thousands of NVIDIA AI chips crucial for building AI software.
It has been revealed that Saudi Arabia has purchased at least 3000 NVIDIA H100 chips through King Abdullah University of Science and Technology (KAUST). Sources indicate that this Saudi Arabian university also possesses at least 200 NVIDIA A100 chips. They are developing a supercomputer named Shaheen III, powered by 700 NVIDIA Grace Hopper superchips, set to be operational this year.
Simultaneously, the UAE has secured access to thousands of NVIDIA chips and has launched its own open-source large language model, “Falcon40B”. In a span of two months, “Falcon40B” was trained using 384 A100 chips.
Related:
Recommended Reading:
- TSMC 2nm Plant: Fatal Accident Leads to Immediate Closure
- Unveiling OpenAI’s Self-Developed AI Chips
- How Tesla and NIO Thrive with Minimal SKUs?
- Intel Faces Competition: Nvidia’s CPU Revolution
- Guide to NVIDIA Graphics Card Driver Installation Issues
- SSD Controller Leaders: Revamp of Silicon Motion
- NVIDIA: Navigating U.S. Export Control Challenges
- Innovative Chinese Simultaneous Interpreting Steals CES Show