The U.S. dominates global data center capacity (if you have any doubts, check this quarter’s stats from LandGate, linked below). So for the rest of the world, data sovereignty means deploying their own local or regional infrastructure and figuring out their relationship to U.S. companies and technology. Today, we see some of how that is playing out in Canada with Bell and Cohere, with NTT Data and Mistral AI in Europe, and in Malaysia’s work to become a regional HPC hub, with a sovereign LLM in the works as well. But technological sovereignty can be a tricky tightrope to walk, especially with current geopolitical tensions. Exhibit A: Nvidia, which is betting big on being able to sell its H20 chip in China, but still faces political pressures at home. Let’s zoom in.
Kelly Hill
Executive Editor
RCRTech
AI Infrastructure Top 3
Bell taps Cohere for AI: Bell and Cohere have partnered to deliver sovereign, security-focused AI solutions to Canadian government and business clients, combining Bell’s infrastructure with Cohere’s agentic platform and AI models.
Sovereign private AI: NTT Data and Mistral AI will deliver private, compliant AI systems tailored for regulated sectors, giving enterprises local control over data, infrastructure and model deployment – independent from U.S.-based hyperscalers.
Malaysia’s AI power move: Malaysia partners with Nvidia and YTL Power to invest RM10b in AI infrastructure, build a sovereign LLM, and position itself as ASEAN’s green-powered hub for high-performance computing and semiconductors.
AI-Powered Telecom Infrastructure
Supermicro, in collaboration with NVIDIA, delivers AI-powered infrastructure tailored for telcos, enhancing operational efficiency, network management, and customer experiences. Explore now
AI Today: What You Need to Know
Nvidia bets on China: Nvidia has reportedly ordered 300,000 H20 chips from manufacturer TSMC, due to strong demand in China. Jeffries had recently estimated that Nvidia’s H20 stockpile was unlikely to meet Chinese demand.
Opposing AI chip sales: Twenty national security experts and former government officials wrote to the Secretary of Commerce, opposing the Trump administration’s approval of sales of some AI chips (like Nvidia’s H20) to China.
AI at Mayo Clinic: Mayo Clinic deploys Nvidia’s Blackwell DGX B200 servers and DGX SuperPOD platform to build scalable AI infrastructure for multimodal and generative foundation models in pathology, drug discovery, and precision medicine.
NatWest banks on AI: U.K. bank NatWest has signed a five-year deal with AWS and Accenture to build a new AI cloud platform to simplify its bank-wide operations. A partnership with OpenAI is already delivering chatbot upgrades.
Power, processors and people: An AI data center planned in Wyoming would consume more energy than the state’s entire population to start — then scale up to five times that size. Wyoming is the least populous state, but a net exporter of energy.
Clouds on the water: You could bring water to the data center, or … the data center to the water? A Turkish company that builds floating power plants is expanding its concept to data centers in shipyards, betting that they’ll be faster to build.
Funding Fireworks: GPU reseller Fireworks AI — backed by the likes of Nvidia, AMD and Databricks Ventures — is reportedly in talks for another funding round that could value the company at $4 billion.
By the numbers: LandGate’s Q2 report on U.S. data centers says that total energy consumption for the quarter reached an estimated 224 TWh. Capex hit a growth rate of nearly 18%, with hyperscalers pouring in more than $300 billion.
California dreamin’: San Jose wants more than just tech HQs: It wants the data centers, too. The city is teaming up with PG&E to supercharge its grid in support of major data center projects underway.
The only thing to fear: Humans have quick, complex, instinctual reactions to things like sudden loud noises. Researchers found that a robot trained to mimic a human “fear response” navigated an unfamiliar environment more safely.