ComputeLabs Research
Edition of · 18 stories
JPMorgan and Goldman Sachs completed the sale of Volta Infrastructure Holdings Ltd.’s $5 billion leveraged loan on Thursday, according to the report citing people familiar with the transaction. Marketing began the previous week, and the banks increased the return offered to investors and revised the financing terms.
Some existing and prospective investors discussed valuing Anthropic at approximately $1.5 trillion after its Wednesday investor day in San Francisco. The reports attribute these discussions to people familiar with the matter; they do not establish an agreed initial public offering (IPO) valuation or an offering price.
Parallel Technology disclosed proposed purchases by two controlled subsidiaries, according to the Telegram summary of its announcement. The equipment is described as GPU compute servers, rather than standalone GPUs, racks, or a specified amount of data-center power capacity.
September’s headline Producer Price Index (PPI) increased 6.3% year-over-year, compared with a reported forecast and previous reading of 5.4%. The monthly increase was 1.3%, against a 0.6% forecast and a previous increase of 0.4%.
TSMC’s third-quarter net profit increased 69% from a year earlier, according to the supplied daily recap. The same source described the quarter as its sixth consecutive quarter of record net profit, but did not provide the absolute profit amount.
Huagong Tech said on an investor-interaction platform that it had jointly developed the products with supply-chain partners. The announcement identifies near-packaged optics (NPO) products at 3.2 terabits per second and 6.4 terabits per second.
Huahai Qingke introduced the ion implanter at the Bay Area semiconductor exhibition, where domestic suppliers presented new equipment, materials, and tools. The report places the launch within broader efforts by Chinese semiconductor-equipment companies to expand into higher-end products.
Financialjuice reported that ByteDance was in talks to obtain MediaTek AI chips for data centers. The supplied message identifies the parties and intended deployment setting, but provides no further transaction details.
Utility Dive reports that American Electric Power reached an agreement to buy up to 1.6 gigawatts of Mainspring linear generators. The stated figure is generation capacity covered by the deal, not an announced amount of operating data-center load.
American Electric Power expects to spend approximately $4 billion on a Bloom Energy fuel-cell project in Wyoming, according to Utility Dive. The spending amount is an expectation, not a statement that the entire sum has already been invested.
Utility Dive reports that reliability standards addressing AI data centers and cryptocurrency miners have moved closer to adoption. The proposed scope covers computational loads of 50 megawatts or more connected at voltage levels of at least 100 kilovolts.
The International Energy Agency’s governing board agreed to accelerate the release of remaining volumes under its March 2026 collective action responding to the Strait of Hormuz crisis. The accelerated release includes more than 20 million barrels of diesel to be made available as soon as possible.
Google Cloud announced a preview integration in which Temporal Cloud, the managed version of the open-source Temporal execution platform, sizes Cloud Run worker pools according to task-queue workloads. Unlike request-serving applications, Temporal workers poll for work, and the integration does not require changes to existing worker code.
Visko and Google Cloud engineers deployed Orbis’s complete live inference pipeline—including video generation, super-resolution upscaling, and synchronized audio—on a single Cloud Tensor Processing Unit (TPU) slice. Google reports that the collaboration moved from kickoff to production handover in 30 days, following the start of work in August 2026.
Google Cloud made the new Virtual Private Cloud Service Controls (VPC-SC) capabilities generally available at no additional cost. Administrators can now add folders to security-perimeter membership, automatically extending their policies to contained projects and nested subfolders.
Muse Spark Open is a coding-optimized mixture-of-experts (MoE) model with 816 billion total parameters, 42 billion active parameters, and a context window of 1 million tokens, according to NVIDIA’s developer announcement. Total and active parameter counts describe different aspects of the model and should not be treated as equivalent measures of its size.
NVIDIA released Nemotron 3.5 Super VL as an open 120-billion-parameter model for specialized, always-on agents. Its developer announcement emphasizes multimodal reasoning and enterprise workflows involving document reading and checking information across sources.
Google announced that its Quantum AI team experimentally validated magic state cultivation on the Willow quantum processor. The reported results were an approximately 40-fold reduction in errors for particular quantum states and 99.99% accuracy in the prepared states, with the research published in Science.
