Sources: Saudi Arabia bought 3,000+ Nvidia H100 GPUs and the UAE purchased thousands of Nvidia GPUs; Nvidia expects to ship ~550K H100 GPUs globally in 2023
it's the $40,000 chip that has sparked a frenzied spending spree Rob Thubron / TechSpot : Saudi Arabia and UAE invest in thousands of Nvidia chips, aiming to become global AI powerhouses Alex Heath / The Verge : Countries are scrambling to buy Nvidia's AI chips. X: @alishihabi : “Many Chinese nationals with AI expertise have chosen to work at Kaust because they have been prevented from studying and working in the US after graduating from Chinese universities on the US entity list, according to two Kaust sources”. https://www.ft.com/... Rachel Ziemba / @reziemba : Good piece on UaE/KSA chip investments to foster their AI goals. Have to imagine that some of the stockpiling also influenced by US restrictions on Chinese purchases of highest level NVIDIA and other chips. https://www.ft.com/... @tariqkrim : According to today's FT story, Saudi and UAE has bought thousands of H100 Tensor Core GPUs, Nvidia's superfast AI chip that has a price tag of about $40,000 . I wonder how many of these chips are actually running in Europe ? https://www.ft.com/... @researchqf : Still think this is underestimated - LT strategic importance of AI for governments (and enterprises). Many of these. I'd prefer H100 # not being leaked but there's still plenty of room for alpha beyond ST timeframe H100 specific numbers. Even for $NVDA. https://www.ft.com/... [image] Marc Lehman / @markflowchatter : Goes along with UBS raising their price target on $NVDA today to $540 ahead of earnings next week Maya Zehavi / @mayazi : AI is a geopolitical race: * a race for Nvidia chips * a race for training data * a race to own LLM modes * race to roll out biometric proof of humanity ID https://www.ft.com/... Adam Tooze / @adam_tooze : Saudi's LLM is being developed by AI outfits at Kaust, primarily staffed by Chinese researchers. Many Chinese with AI expertise have chosen to work at Kaust because they are prevented from working in US after graduating from Chinese unis on US entity list. https://www.ft.com/... Leo Strauss / @drleostrauss : Nvidia H100 chips a decisive component in the global AI race. (Newly announced H200 as well). 🇨🇳, 🇸🇦,🇦🇪. Tracking and understanding their utilization a key natsec and political economy metric beyond competitive tech business models frames. [image]
Context & Ripple Effects
Earlier coverage described the A100 as a critical generative-AI input with Nvidia holding an estimated 95% share of machine-learning GPUs; this report shows that demand for its successor-class hardware was extending beyond cloud and technology companies to state-backed buyers. The scale is notable against Nvidia's stated plan to ship about 550,000 H100s worldwide in 2023.
Nvidia had also been broadening its AI stack with the DGX GH200 platform and data-center networking tools. Large national GPU purchases make that surrounding infrastructure increasingly consequential, not just the chips themselves.
First-order effects
- Saudi Arabia and the UAE secure sizable allocations of scarce H100 capacity, while Nvidia adds major sovereign customers to its 2023 shipment mix.
- The buyers must turn the GPU purchases into operating AI capacity, increasing the immediate importance of data-center systems and networking around the accelerators.
Second-order effects
- Other H100 buyers face tighter competition for a finite 2023 supply pool, strengthening Nvidia's leverage over allocation and reinforcing the value of alternatives and cloud access.
- The purchases pull demand toward the broader AI-infrastructure stack; Nvidia's earlier A100 leadership in generative AI had already made accelerators a central bottleneck.
Third-order effects
- If governments continue procuring frontier compute directly, AI infrastructure may increasingly be treated as strategic national capacity rather than solely as commercial cloud equipment.
- That shift would make access rules and supplier concentration more important to AI deployment: later coverage of different A100/H100 cloud-rental costs in China despite export controls illustrates how policy can shape effective compute access.
The trend: This is an early instance of sovereign AI procurement turning advanced GPU supply into a form of strategic national infrastructure.