ComputeLabs Research
Edition of · 20 stories
CScale emerged from stealth on September 30 and announced a $145 million Series C, bringing its stated cumulative funding to $188 million. The announcement identifies development and commercialization of optical interconnects for AI scale-up as the purpose of the financing.
HPE announced a $1.2 billion order from Vultr for AMD Helios AI Rack by HPE systems. The announcement identifies this as HPE’s first AMD Helios order and places the planned deployments at Vultr’s US cloud data-center locations.
Micron (美光科技) expects approximately $11.5 billion of capital expenditure in fiscal 2027’s first quarter and approximately $25 billion in its first half, according to the supplied company-attributed Telegram report. The first-quarter amount is included within the first-half projection, not additional to it.
The Senate failed to advance the Republican-backed legislation after a 57–43 procedural vote, below the required 60-vote threshold. The detailed report says only four Democratic senators supported advancement.
The supplied financialjuice message reports that the Federal Trade Commission (FTC) opened a probe involving Anthropic, OpenAI, and other companies. Its attribution is to CNBC citing an FTC spokesperson, rather than to an FTC document included in the source material.
The supplied earnings summaries report Micron (美光科技) fiscal 2026 fourth-quarter adjusted revenue of $54.23 billion, versus an estimate of $51.49 billion. Cloud Memory revenue was $16.28 billion, versus $15.14 billion estimated; Mobile & Client revenue was $13.11 billion, versus $12.95 billion; and Automotive & Embedded revenue was $6.82 billion, versus $4.73 billion.
CoreWeave plans to offer NVIDIA Vera central processing units (CPUs) for agentic AI, according to financialjuice. This is a CPU offering, distinct from the company’s separately announced availability of NVIDIA Vera Rubin NVL72 systems.
Eurofiber and Delft Networks have deployed dark fiber between Rotterdam and Delft, according to Data Center Dynamics. The source explicitly describes the deployment as a closed-loop configuration.
The supplied reports identify Project Star as a $22.3 billion energy-infrastructure campus in Encinal, Texas, jointly developed by Related Companies and NextEra Energy, with Lewis Energy Group participating. The US Department of Commerce and South Korean government announced strategic investment in the project.
LG Electronics (LG电子) plans to invest approximately KRW 150 billion to expand AI-data-center cooling-system production in the United States and South Korea. The report attributes the investment plan to a company statement.
PJM delayed its backstop power procurement after the Federal Energy Regulatory Commission (FERC) only partially approved the proposed arrangement. Utility Dive identifies cost allocation as the central concern in the commission’s review.
Ameren Missouri filed a 20-year resource plan proposing 10.6 gigawatts of gas-fired generation additions. The figure represents proposed generation capacity, not a reported increase in already operating supply.
The supplied Telegram report says Constellation Energy and Amazon reached a 20-year power purchase agreement (PPA). The contractual term is the principal commercial detail provided.
CoreWeave announced the availability of NVIDIA Vera Rubin NVL72 on its cloud on September 30. Its announcement identifies Cognition as the first customer anywhere running production workloads on the system; that first-customer claim is company-attributed.
VDURA announced general availability of Data Platform V12, identifying AI factories and neoclouds as target users of its high-performance storage infrastructure. The disclosed release features include multi-tenancy, application programming interface (API)-first automation, and Context-Aware Tiering.
Google Cloud made Spanner Omni, its deploy-anywhere version of Spanner, generally available for production workloads. Google says it provides the same core capabilities as managed Spanner while supporting deployment on customers’ own infrastructure and other clouds.
Google Kubernetes Engine (GKE) introduced Pod snapshots to save a workload’s running state, including CPU and GPU memory, and restore it on demand. The feature concerns workload-state preservation rather than a new processor or additional physical compute capacity.
Google’s DeepMind introduced Gemini 4 Argon, initially making it available to selected trusted cybersecurity defenders through the Fairwind program. The company described later access for developers, enterprises, consumers, paid API customers, and Google AI Ultra users as planned expansion following further safety work, not availability already delivered to all those groups.
OpenAI and Synopsys (新思科技; ticker SNPS in the supplied report) signed a binding multiyear agreement to jointly develop GPT-Synopsys, a specialized semiconductor-design model. The reported agreement includes revenue sharing and joint go-to-market cooperation, but no revenue-sharing percentages, minimum payments, or total contract value are supplied.
Anthropic announced general availability of Claude for Government, according to the supplied messages. A separate company-attributed report explicitly identifies US federal and state governments as the covered government users.
