Infrastructure and hardware comprise the physical compute systems, specialized chips, and data center components that power artificial intelligence workloads. Truncated covers chip architectures, supply chains, energy requirements, and the hardware deployments that enable large scale computing.
AI-generated summary Researchers at the University of Tokyo and NTT have released a novel AI framework that runs on optical hardware to process information with light. The University of Tokyo researchers say the approach reduces computational demands and could pave the way toward faster and more energy-efficient AI hardware.
AI-generated summary Winbond has available plans for construction of a new plant in Kaohsiung to begin in December, with a meaningful production ramp-up not expected until 2030. Winbond says that a single Nvidia Vera Rubin AI data center rack needs some 600 specialized NOR flash memory chips.
AI-generated summary Aditya Kumar released an overview detailing how to build XLA from source and teach it a new pass. The instructions explain how to add a custom pass to rewrite expressions within the compiler pipeline.
AI-generated summary Hon Hai Precision Industry Co. has released October sales data showing a 46 percent jump in revenue to NT$1.31 trillion, or $39.7 billion. Hon Hai says the sales increase reflects resilient demand for Nvidia Corp. servers that are central to a global artificial intelligence infrastructure build-out.
AI-generated summary Following the midterm results, Democrats secured full control of Congress and made data center limitations available for discussion. Goldman Sachs predicts that data center construction delays will occur in states like Georgia and Texas due to these Democratic gains.
AI-generated summary Goldman Sachs released a note estimating that states likely to tighten rules on data centres account for about 45% of planned US capacity. Goldman Sachs says the political backlash against data centres is a delay story rather than a demand collapse, which may limit further downside if pauses prove brief.
AI-generated summary HPE and AMD have launched new AI and supercomputing infrastructure platforms using AMD EPYC processors and AMD Instinct accelerators to help customers build capable systems. Customers can utilize these integrated technologies to support demanding workloads such as scientific research, digital engineering, and training advanced artificial intelligence models.
AI-generated summary Satellite analysis from research firm Aterio released details indicating that infrastructure delays by third-party cloud providers may explain why Microsoft spent $5 billion less than expected on artificial intelligence last quarter. The research firm Currence estimates that 30% to 50% of planned data center capacity initially scheduled to come online this year could be pushed into 2027 or later due to electricity shortages and other construction issues.
AI-generated summary Amazon and Microsoft shares are available up around seven percent as the market responds to a slower data center buildout following local votes. The market has rejected the high cash demands and spending on data center supply chains and chips as the risk reward ratio proved unconvincing.
AI-generated summary Riot Platforms announced a twenty-eight billion dollar artificial intelligence data center lease agreement covering nine hundred ten megawatts at its Corsicana, Texas facility. CEO Jason Les says nearly all of the company's expected cash flow will be generated from long-term leases with leading players in the artificial intelligence ecosystem.
AI-generated summary Data center permits are available for halt after Governor Greg Abbott directed a key state regulator to stop them in Texas. The change in political leadership in Texas may lead to delays and moratoriums that could stall the ongoing artificial intelligence buildout.
AI-generated summary NVIDIA releases NVIDIA CUDA 13.5 to deliver general availability for NVIDIA Vera Rubin support across major new platforms. NVIDIA says the update provides innovations across libraries, compilers, NVIDIA CUDA Tile, and developer tools to improve scheduling and performance.
AI-generated summary Marvell made available a $115 billion all-stock deal to acquire Lumentum in order to combine optical components with its accelerators. Investors responded by sending Marvell shares down more than 12 percent due to market concerns over power constraints and capital expenditures.
AI-generated summary Aditya Kumar released a detailed technical post pinned to vLLM 0.32.0 explaining the internal architecture and request lifecycle of vLLM V1. Every per-token cost that is not the forward pass is paid in the API server on a single asyncio event loop per process.
AI-generated summary Bernstein released lowered price targets on seven bitcoin miners following an election outcome in Texas. The adjustments follow concerns that a political shift in Texas could introduce stricter rules and delays for data centers.
AI-generated summary CoreWeave Inc. unsecured debt made available for trading fell on Wednesday as bonds tumbled to new lows. The obligations due in 2030, 2031, and 2032 declined following gubernatorial race wins by Democratic candidates who favor slowing down data-center construction.
AI-generated summary ROCm Blogs has made available GLM-5.2 AgentX on AMD Instinct MI355X to boost throughput and interactivity. ROCm Blogs states that cache efficiency, faster decode kernels, careful AgentX validation, and upstream collaboration raised peak throughput 1.72 times and peak P90 interactivity 2.46 times.
AI-generated summary SambaNova announced a partnership with Supermicro to deploy SN50 RDU chips for AI disaggregated inference systems. Supermicro will initially purchase 10,000 SN50 chips from SambaNova and integrate them alongside Nvidia GPUs for customers.
AI-generated summary Nvidia plans to require an 800 VDC power architecture beginning with its Rubin Ultra and Kyber rack architectures in 2027. This shift will result in lower-current and simpler power distribution that enables data centers to use the latest GPUs.
AI-generated summary Dario Gil announced federal technology initiatives and research and development efforts involving quantum computing, artificial intelligence, and fusion energy. The initiatives include a portfolio of two hundred and ninety-seven projects across all fifty states and a goal of electricity-producing fusion-power plants by the mid-2030s.
The top AI news in your email delivered each morning. Choose categories