英伟达CEO黄仁勋：数据中心正在进入（被部署到）太空。 ...

Core Insights - Kepler Communications has launched the world's first commercially operational optical data relay network with distributed on-orbit computing capabilities, enhancing its service offerings beyond mere connectivity to include scalable, cloud-like processing in space [1][2]. Group 1: Network Capabilities - The Kepler Network integrates optical connectivity, distributed on-orbit computing, and secure payload hosting into a unified space-based infrastructure, enabling real-time data transport and advanced applications such as AI-driven Earth observation analytics and autonomous network operations [2][3]. - The network utilizes NVIDIA-powered edge computing to process data in space, allowing customers to generate insights at the edge and reduce downlink requirements, thus facilitating scalable, cloud-native workloads across the constellation [2][3]. Group 2: On-Orbit Compute Deployment - The initial on-orbit compute capability is powered by 40 NVIDIA Jetson Orin modules distributed across 10 satellites, marking the first integration of constellation-scale edge computing within a commercially operational optical data relay network [3][4]. - Each satellite acts as a compute-enabled node, supporting AI and other accelerated workloads, and is interconnected through Kepler's real-time optical communications network [3][4]. Group 3: Architectural Advantages - The architecture allows for both single-node execution and clustered, distributed computing models, enabling dynamic scaling of workloads across the constellation [4]. - In the event of a node failure, workloads can be seamlessly shifted to other nodes, ensuring continuity of service [4]. Group 4: Strategic Vision - The CEO of Kepler Communications emphasized that this architecture alleviates long-standing constraints in space operations, enabling real-time data processing and decision-making in orbit, which enhances resilience and supports new mission architectures for customers [5]. - The GPU-enabled on-orbit compute platform is designed to support multiple concurrent customers with secure isolation between workloads, facilitating various in-space applications [5]. Group 5: Company Overview - Kepler Communications operates the first commercial optical data relay constellation with 33 satellites launched to date, achieving significant milestones in space communications and on-orbit computing capabilities [6]. - The company is headquartered in Toronto, Canada, and aims to enable communications for the future space economy [6].

Core Insights - Delta has introduced AI-based digital twin applications utilizing NVIDIA Omniverse to enhance building automation and smart manufacturing solutions, showcasing the benefits of real-time simulation and predictive analysis capabilities [1][2][3] Building Automation - The integration of physics-based simulations for HVAC, lighting, and natural shading has led to a potential energy savings improvement of up to 20% [2][4] - Delta's AI digital twins combine real-time environmental data with photorealistic simulations, optimizing energy performance and occupant comfort [3][4] Smart Manufacturing - Delta's DIATwin system employs NVIDIA Omniverse to streamline the design, validation, and scaling of smart manufacturing processes, reducing change orders and commissioning timelines [5][6][7] - The use of NVIDIA PhysX for accurate production line simulations enhances robot path optimization and accelerates production line deployment [7][8] Operational Efficiency - The AI-driven digital twin technology supports decentralized manufacturing with centralized management, facilitating the transition to autonomous factories [8] - Delta's collaboration with NVIDIA allows for real-time system behavior simulations and the modeling of complex scenarios, improving operational resilience and energy efficiency [3][5]

Core Insights - Supermicro has unveiled a new system portfolio powered by the NVIDIA Vera Rubin platform, aimed at transforming data centers into AI factories to meet the growing demand for AI workloads [1][2] Group 1: Product Offerings - The new systems include NVIDIA Vera Rubin NVL72, NVIDIA HGX Rubin NVL8, and NVIDIA Vera CPU systems, designed to accelerate customer time-to-market [1][2] - Supermicro's DCBBS technology stack is engineered to support these systems, providing a fast path for deploying next-generation AI factories at scale [2][3] - The NVIDIA Vera Rubin NVL72 is designed to deliver up to 3.6 Exaflops of inference, 75TB of fast memory, and 1.6 PB/s of HBM4 bandwidth, targeting up to 10x throughput per watt compared to NVIDIA Blackwell solutions [6][5] Group 2: Infrastructure and Technology - Supermicro's DCBBS approach allows data center operators to deploy validated, pre-engineered rack solutions, reducing time-to-online and minimizing integration risk [3][4] - The DCBBS includes a full suite of validated liquid-cooling infrastructure, essential for the upcoming Vera Rubin platforms, which will be fully liquid-cooled [4][5] - The 2U HGX Rubin NVL8 system supports up to 72 Rubin GPUs per rack, providing flexibility in CPU selections, including NVIDIA Vera CPUs and next-generation AMD and Intel x86 processors [7][8] Group 3: Market Position and Strategy - Supermicro emphasizes the necessity for organizations to establish AI factories to remain competitive, as the demand for inference workloads reshapes data center infrastructure [2] - The company is committed to delivering first-to-market innovations for enterprise, cloud, AI, and 5G Telco/Edge IT infrastructure [13][14] - Supermicro is investing in both its current Blackwell lineup and next-generation systems to ensure customers have the right platform for their AI infrastructure needs [11]

Core Insights - Delta is showcasing its integrated ecosystem of energy-saving power, cooling, and microgrid solutions at NVIDIA GTC, aimed at enhancing the 800 VDC architecture for next-generation AI factories [1][3] Power Solutions - Delta has developed new 800 VDC In-Row 660kW Power Racks featuring a total of 480kW of embedded Battery Backup Units (BBU) and achieving AC-DC efficiency up to 98% [2][4] - The new power racks are supported by 18.5kW AC/DC Power Supply Units (PSU) designed to optimize GPU workloads against high-frequency dynamic distortion [4] Cooling Solutions - Delta's thermal management portfolio includes a 2.4MW liquid-to-liquid (L2L) Coolant Distribution Unit (CDU) that provides 2,400kW of cooling capacity, designed specifically for AI factories powered by 800 VDC architectures [4][5] - The new 3MW L2L CDU offers up to 3,000kW of cooling capacity and is scalable to operate in groups for optimal cooling [4] Microgrid Solutions - Delta's Microgrid Solution integrates Solid State Transformers (SST) and Solid Oxide Fuel Cells (SOFC), achieving efficiency as high as 98.5% while converting medium voltage AC to 800 VDC [5] - The SOFC provides resilient onsite power generation with lower CO2 emissions, significantly reducing 'time-to-power' from years to months [5] System Integration and Readiness - Delta's solutions are designed to support the foundational infrastructure needed for AI data centers, addressing both performance and sustainability challenges [6] - The company emphasizes its system integration capabilities to enhance power and thermal management in data centers and energy infrastructure [3][6]

Supermicro Launches Seven AI Data Platform Solutions with NVIDIA and Leading Ecosystem Partners to Accelerate Enterprise AI Innovation Accessibility StatementSkip Navigation SAN JOSE, Calif., March 16, 2026 /PRNewswire/ -- Supermicro, Inc. (NASDAQ: SMCI), a Total IT Solution Provider for AI, Cloud, Storage, and 5G/Edge, today announced seven fully integrated AI Data Platform solutions designed to accelerate enterprise AI adoption. Built on NVIDIA reference architectures and powered by Supermicro's high-perf ...

Skip to main content Exclusive news, data and analytics for financial market professionalsLearn more aboutRefinitiv FILE PHOTO: The logo of Swiss pharmaceutical company Roche is pictured on the company's headquarters in Basel February 4, 2009. REUTERS/Christian Hartmann/File Photo Purchase Licensing Rights, opens new tab BERLIN, March 16 (Reuters) - Swiss drugmaker Roche (ROG.S), opens new tabsaid on Monday it had expanded its artificial intelligence computing capacity with more than 2,100 Nvidia chips to s ...

Core Insights - Uber and Nvidia plan to deploy a fleet of robotaxis powered by Nvidia's autonomous driving software, starting in Los Angeles and San Francisco in 2027, with expansion to 28 cities globally by 2028 [1][5]. Group 1: Company Plans and Partnerships - The rollout will begin with data-collection vehicles to train the system on city-specific driving conditions, transitioning to operator-supervised launches and eventually fully driverless Level 4 operations [4]. - Uber's partnership with Nvidia is part of a broader strategy to create a "multi-player" autonomous vehicle ecosystem, allowing various developers and automakers to bring robotaxi services to market [5][6]. - Uber has previously partnered with Lucid Group and Nuro to deploy robotaxis built on Lucid vehicles and powered by Nuro's self-driving software [6]. Group 2: Competitive Landscape - The competition in the autonomous ride-hailing sector is intensifying, with Alphabet's Waymo currently leading with fully driverless rides in cities like Phoenix, San Francisco, and Los Angeles [2][3]. - Tesla is also entering the market with plans for its own robotaxi service, leveraging its large vehicle manufacturing capacity and a camera-based approach to autonomy [3].

Vertiv Brings Converged Physical Infrastructure to NVIDIA Vera Rubin DSX AI Factories Accessibility StatementSkip Navigation Simulation-ready power and cooling infrastructure models, designed to accelerate deployment and reduce execution risk COLUMBUS, Ohio, March 16, 2026 /PRNewswire/ -- Vertiv (NYSE: VRT), a global leader in critical digital infrastructure, today announced its role in advancing converged physical infrastructure designs for the NVIDIA Vera Rubin DSX AI factory reference designand the NVIDI ...

Core Insights - Skild AI's AI model will enhance robotics on Foxconn's assembly lines in Houston, marking a significant step in the deployment of generalized physical AI [1][2] - The partnership with ABB Robotics and Universal Robots aims to integrate Skild AI's software into industrial robots, providing a general-purpose "brain" for enhanced adaptability [2][3] Company Developments - Skild AI, supported by Nvidia and SoftBank, is focusing on overcoming limitations of current robotics systems that are typically designed for single tasks [2] - The CEO of Skild AI emphasized the potential for scalability through partnerships with robotic OEMs that have extensive deployments [3] Industry Trends - The U.S. is investing approximately $1.2 trillion in domestic manufacturing, particularly in electronics, pharmaceuticals, and semiconductors, with automation being crucial for advanced manufacturing reshoring [4] - Nvidia plans to invest $500 billion in AI infrastructure over the next three to four years, highlighting the need for more autonomous factories [5] Financial Highlights - SoftBank's acquisition of ABB's robotics business for $5.38 billion is expected to close by mid-to-late 2026, indicating significant consolidation in the robotics sector [6] - Skild AI raised $1.4 billion in funding, valuing the company at over $14 billion, showcasing strong investor confidence [6]