ComputeLabs Research
Edition of · 20 stories
CoreWeave reportedly plans to raise at least $4 billion through a proposed investment-grade loan. Morgan Stanley and Mitsubishi UFJ Financial Group are arranging the potential financing, which would be syndicated to investors in stages.
Hangzhou Landscape Architecture’s controlled subsidiary, Hangwei Tiansuan, signed two procurement contracts. One covers 20,000 NVIDIA RTX PRO 5500 Blackwell GPUs and accessories from Yunjian Information Technology; the other covers 10,000 GPUs from Shanghai Huakun Zhenyu Intelligent Technology.
Alphabet began marketing its first bonds denominated in Hong Kong dollars and Singapore dollars. The reported use of proceeds is general corporate purposes, potentially including repayment of existing debt—not a specifically earmarked AI infrastructure investment.
The reported 73–25 Senate vote was procedural, rather than final passage of the legislation. The source dates that vote to Monday; the November 11 development was House Speaker Mike Johnson’s discussion of the bill’s prospects in the House.
NVIDIA’s board reportedly engaged an external firm to identify potential replacements for founder and CEO Jensen Huang. The Bloomberg-attributed report also said directors were considering an interim CEO while conducting a broader search for a permanent successor.
Supermicro announced that its H15-generation server portfolio is shipping. The systems use sixth-generation AMD EPYC 9006 central processing units, formerly codenamed Venice; H15 identifies the server generation, while EPYC 9006 identifies the CPU family.
South Korean President Lee Jae-myung said Samsung is investing KRW 6 trillion in a high-bandwidth memory packaging plant in Asan, South Korea. The stated facility scope is HBM packaging, not a disclosed GPU production plant or data-center project.
IBM announced Quantum Nighthawk r3 as its next-generation quantum processor. The company says it will be built to accurately execute 10,000 quantum operations, while the accompanying Telegram report specifies a scope of 520 qubits.
Rigetti Computing reported 99.5% median two-qubit fidelity on its 108-qubit system. The result concerns its chiplet-based quantum processing unit (QPU) architecture, not a classical GPU or CPU system.
Data Center Dynamics reported a consortium’s plan to develop a 400-megawatt hyperscale AI data center in Oman. The figure describes the proposed data-center project’s capacity, not a GPU count or a disclosed quantity of commissioned infrastructure.
Microsoft’s cloud operations and innovation head, Noelle Walsh, said the company would bear the electricity costs associated with its data-center expansion. The pledge explicitly excludes reliance on government subsidies or transferring those costs to other electricity consumers.
Microsoft is working with Chevron and investment firm Engine No. 1 on Project Kilby, a large data-center development with associated power infrastructure in West Texas. The source identifies both the computing facilities and supporting electricity project, but does not specify their respective capacities.
ERCOT CEO Pablo Vegas plans to ask the Texas legislature for authority to require data centers to reduce electricity demand earlier in the grid-stress response process. The source describes a proposed expansion of authority, not a rule already enacted.
Utility Dive reported that California ordered utilities to develop a joint framework for orchestrating distributed energy resources (DERs). The underlying order was issued in October, with the supplied article published on November 11.
AWS launched AWS Emulator, a free, open-source local emulator written in Rust. Its stated purpose is to let developers test applications against nearly every AWS service on their own machines.
Dell Technologies announced that the University of Alabama is deploying Dell infrastructure at the core of its new High Performance Computing Center. The deployment covers high-performance computing (HPC), AI, and data infrastructure, rather than a single disclosed server or accelerator purchase.
Meta infrastructure chief and engineering co-head Santosh Janardhan told employees he would leave in early 2027. The report identifies Dave Brown as his successor as infrastructure chief, with Janardhan remaining through the end of the first quarter to support the handover.
Google launched the Gemini Decisions application programming interface (API) with the Decision 1 model, according to the supplied Telegram reports. The API is the announced service interface, while Decision 1 is the named model associated with it.
Google Cloud announced a preview, rather than general availability, of `ai.classify` and `ai.extract` in the AlloyDB AI query engine. The functions are available through the `google_ml_integration` extension, version 1.6.1 or later, on PostgreSQL 18.
University of Florida researchers published a proposal called SANAI for protecting critical neural-network weights against bit-flip attacks. The paper describes a sensitivity-guided, bit-plane-selective scheme, identifying selective weight protection as a central part of the approach.
