🇫🇷Quantum-computing firm Pasqal completed its special-purpose acquisition company merger, receiving approximately $360 million in cash. #
French quantum-computing company Pasqal completed its merger with Bleichroeder Acquisition Corp. II, a special-purpose acquisition company. The completed transaction took Pasqal public.
The merger provided Pasqal with approximately $360 million in cash. The source did not provide the post-merger ticker, redemption level, ownership split, or other dilution details.
This was a completed business combination rather than an announced plan or preliminary agreement. The disclosed cash amount therefore relates to proceeds provided by the closed transaction.
🇺🇸Physics-research startup Physical Superintelligence launched with $58 million in seed funding led by Breakthrough Energy Ventures. #
Matt Pines, Alex Klokus and Dr. Alexander Wissner-Gross launched Physical Superintelligence, or PSI, as an artificial intelligence-focused physics research laboratory. The organization is developing computational models and tools for designing and optimizing physical systems.
PSI announced $58 million in seed funding led by Breakthrough Energy Ventures. Dragon Global and Robot Ventures were also identified as participating investors.
The funding announcement accompanied the company’s launch rather than a later-stage financing. The supplied source did not disclose a valuation, ownership percentages, hardware deployment plan or specific computing-capacity commitment.
- Physical Superintelligence
- Breakthrough Energy Ventures
🇺🇸NVIDIA plans to purchase $1.5 billion of renewable-energy developer SB Energy’s Class N stock through a placement. #
NVIDIA plans to purchase $1.5 billion of SB Energy Class N stock through a placement. The source describes a planned purchase and does not state that the placement has closed.
SB Energy filed a Form S-1 registration statement with the U.S. Securities and Exchange Commission on September 1. The filing describes an integrated business model serving next-generation, AI-driven data infrastructure alongside the company’s renewable-energy activities.
The S-1 identifies risks including uncertain demand for AI compute infrastructure, potential overbuilding of data-center capacity, changes in rack density and cooling requirements, permitting constraints and regulations associated with the energy intensity of AI training and inference. It also notes exposure to cost increases for solar panels, battery-storage systems, transformers, inverters, steel, copper and labor.
The supplied placement report did not specify the purchase price per share, closing date, conditions, voting rights or resulting NVIDIA ownership percentage. Accordingly, the $1.5 billion figure should be treated as the planned investment amount, not confirmed closed proceeds.
🇺🇸 🇮🇷WTI rose 5.2% to $90.22, and Brent gained 4.6% to $94.65, following new U.S. strikes on Iran. #
October West Texas Intermediate crude futures rose $4.46, or 5.20%, to settle at $90.22 per barrel. November Brent crude futures increased $4.16, or approximately 4.60%, to $94.65 per barrel.
The reported market move followed a new round of U.S. strikes on Iranian targets near the Strait of Hormuz. U.S. Central Command said the targets included Islamic Revolutionary Guard Corps air-defense positions, radar systems, maritime assets and facilities, mine-laying capabilities and communications sites.
Other energy contracts also moved sharply. Abu Dhabi Murban crude futures rose 7.30% to $105.64 per barrel, while October New York Mercantile Exchange gasoline and heating-oil futures settled at $3.1351 and $4.6773 per gallon, respectively.
The strikes occurred amid concerns about commercial shipping through the Strait of Hormuz. The Group of Twenty presidency statement separately expressed concern about disruptions to energy trade and the safety of navigation through the strait.
🇺🇸Federal Reserve Governor Michael Barr called inflation too high and supported raising rates if inflation fails to moderate. #
Federal Reserve Governor Michael Barr said U.S. inflation remained too high. He said the persistence of inflation above the Federal Reserve’s target created risks.
Barr favored holding rates steady if policymakers were confident that inflation was moderating. He also said that if inflation did not moderate soon, it would be time to raise interest rates and that he was prepared to support such an increase.
Barr described the labor market as stable with low unemployment and said the economy was growing solidly, supported by artificial-intelligence investment. The Federal Reserve was scheduled to hold its next policy meeting in mid-September.
Dell reported $16.4 billion in second-quarter AI-optimized server revenue and raised its full-year forecast to $74 billion. #
Dell reported fiscal 2027 second-quarter revenue of $46.97 billion, compared with $29.776 billion a year earlier and analyst expectations of approximately $44.8 billion. Net income was $4.13 billion, while adjusted earnings per share were $7.04 and adjusted operating profit was $5.93 billion.
Revenue from servers designed and optimized for artificial-intelligence workloads reached $16.4 billion, versus a reported market expectation of $16.03 billion. Traditional server and networking revenue was $10.53 billion, and Infrastructure Solutions Group revenue totaled $31.78 billion.
Dell raised its full-year AI-optimized server revenue forecast to $74 billion from approximately $60 billion. The company also raised total fiscal-year revenue guidance to approximately $192 billion, plus or minus $2 billion, from its previous range of $165 billion to $169 billion; third-quarter revenue guidance was approximately $49 billion, plus or minus $500 million.
Management said demand was expanding across neocloud, sovereign-cloud and enterprise customers, with the customer count exceeding 6,500. Dell also cited shortages affecting dynamic random-access memory, NAND flash, some central processing units, hard drives, power components, mature-node chips and optical modules, while identifying about 1.2 million 14th-generation or older servers in the installed base as candidates for replacement.
Interconnect-chip supplier Credo Technology reported record first-quarter revenue of $479 million, up 115% year over year. #
Credo Technology reported record fiscal first-quarter revenue of $479 million. Revenue increased 10% sequentially and 115% year over year, marking the company’s seventh consecutive quarter of triple-digit annual growth.
Non-generally accepted accounting principles gross margin was 68%, and non-GAAP operating margin was 48.2%. Non-GAAP net income exceeded $236 million, representing year-over-year growth of 140%, while cash and cash equivalents ended the quarter at $764.3 million after cash expenditure associated with the Dust Photonics acquisition.
Customer concentration remained high: four customers each represented at least 10% of quarterly revenue, accounting for 33%, 28%, 13% and 10%, respectively. The company said it was broadening its hyperscaler and neocloud customer base.
For the second quarter, Credo guided to revenue of $525 million to $535 million and a non-GAAP gross margin of 67% to 69%. Management projected more than 85% full-year revenue growth, optical-interconnect revenue exceeding $600 million and a non-GAAP net margin of approximately 50%; these figures are company guidance rather than reported results.
Mixx Technologies’ SxC optical connector terminates up to 24,576 fibers in one rack unit for frontplane and backplane use. #
Mixx Technologies launched the SxC Connector as an optical connector designed for both frontplane and backplane deployment. The company says one connector technology can extend from the optical engine to the edge of the system enclosure.
The connector terminates as many as 24,576 optical fibers within one rack unit, or 1U. The figure describes fiber termination density rather than server count, accelerator count or aggregate data-center capacity.
Mixx positioned the connector for high-radix scale-up networks and multi-petabit connectivity. The company also announced a live connector demonstration at SEMICON Taiwan.
SK hynix is considering a Japanese factory and identified joint production with Kioxia as one possible option. #
SK Group Chairman Chey Tae-won said SK hynix was considering building a factory in Japan. No final investment decision, factory location, production capacity or construction timetable was disclosed.
Chey identified joint production with Kioxia Holdings as one possible form of cooperation between the two memory-chip companies. He also listed research and development and supply-chain collaboration as possible areas for cooperation.
Chey said Kioxia had several advantages. He added that SK hynix could reduce its indirect stake in Kioxia if the companies could not identify a workable cooperation model, making the production proposal an option rather than a binding agreement.
MISO (Midcontinent Independent System Operator) proposed rules for loads above 50 megawatts, including computational loads with 25 megawatts of IT demand. #
The Midcontinent Independent System Operator proposed reliability requirements for large electricity loads. MISO defines a large load as one exceeding 50 megawatts.
The proposal separately defines a computational load as a large load containing at least 25 megawatts of demand from information-technology equipment. This distinguishes the IT component of a data center from the site’s overall electricity demand, which can also include cooling and other facility systems.
The separate classifications allow MISO to direct particular requirements toward data centers and other computational facilities. The supplied source description did not specify the complete interconnection process, compliance deadlines or penalties under the proposal.
China’s National Energy Administration directed faster approvals and construction for major transmission corridors and other grid projects. #
China’s National Energy Administration (国家能源局) convened a new-grid construction deployment meeting on September 1. The agency called for faster approval and construction of major grid projects, including large transmission corridors, alongside stronger overall coordination and policy implementation.
The agency said electricity demand during the Fifteenth Five-Year Plan period (十五五) was expected to continue growing by approximately 5% annually. It also cited a rising renewable-energy share, increasing demand from advanced industries for green and high-quality electricity, emerging operating models and intensified competition in core technologies.
The planned architecture would coordinate the bulk power grid, distribution networks and flexible smart microgrids. The agency called for accelerated research and deployment involving AI Plus (人工智能+), grid-forming technologies, intelligent dispatch and long-duration energy storage.
The meeting also addressed efficient renewable-energy interconnection and consumption, high-quality power supply, and coordination between computing and electricity. It emphasized bulk-grid stability, project safety and quality, disaster response and wider grid resilience.
- National Energy Administration
The U.S. Energy Department ordered Orlando’s Stanton coal Unit 1 operational from September 2 through November 30. #
U.S. Energy Secretary Chris Wright issued an emergency order concerning Unit 1 at the Stanton Energy Center in Orlando, Florida. The order directs the Orlando Utilities Commission to ensure the coal-fired unit remains available for operation.
The unit had been scheduled to enter an extended cold-shutdown state early, beginning in June 2026. The emergency order instead takes effect on September 2 and remains valid through November 30.
The directive concerns the operational availability of one generating unit, not construction of new generation capacity. The supplied source did not disclose the unit’s generating capacity, anticipated operating schedule or fuel-consumption requirement during the order period.
- Energy Department
- Stanton coal Unit 1
Explosive devices at a Brandenburg substation forced one unit at the nearby Jänschwalde coal plant offline. #
Multiple explosive devices were found at an electrical substation in Brandenburg in eastern Germany on September 1. The incident was described as sabotage affecting power infrastructure.
A local energy-company executive said the disruption caused one unit at the nearby Jänschwalde coal-fired power plant to shut down suddenly. The source did not state the generating capacity of the affected unit.
The report concerned a single plant unit rather than the entire Jänschwalde station. No restoration time, total regional power loss or confirmed perpetrator was provided in the supplied material.
Brookhaven National Laboratory will lead a $14.2 million, three-year project developing a next-generation electric-grid foundation model. #
The U.S. Department of Energy selected Brookhaven National Laboratory to lead a $14.2 million electric-grid artificial-intelligence project. The work will run for three years.
The project’s stated objective is to develop a next-generation foundation model for the electric grid. A foundation model is the project’s AI research output; the $14.2 million figure is project funding, not data-center capital expenditure or grid-generation capacity.
The project was selected through the Department of Energy’s Genesis Mission Phase II Request for Application. The supplied announcement did not specify the model’s parameter count, training-compute allocation or underlying accelerator hardware.
- Brookhaven National Laboratory
NSF established the National Artificial Intelligence Research Resource Operations Center, led by San Diego and Texas advanced-computing centers. #
The U.S. National Science Foundation established the National Artificial Intelligence Research Resource Operations Center, or NSF NAIRR-OC. The center is intended to strengthen U.S. capacity for AI research and AI-enabled scientific discovery.
The operations center will be led by the San Diego Supercomputer Center and the Texas Advanced Computing Center. These organizations are advanced-computing centers rather than commercial cloud providers.
The investment will provide the operational foundation for the broader National Artificial Intelligence Research Resource. NAIRR is described as a national public-private effort designed to expand access to advanced AI research capabilities.
The supplied announcement did not specify a total dollar value, accelerator count, cluster configuration or power capacity for the operations center. Its stated role is operational coordination and support for the national resource.
AI benchmarking consortium MLCommons expanded MLPerf Storage with key-value cache, vector-database, S3 object-storage, and POSIX file-system workloads. #
MLCommons expanded the MLPerf Storage benchmark to test key-value cache and vector-database workloads. These workloads represent storage demands associated with large-scale AI inference and retrieval systems.
The suite also added support for Amazon S3-compatible object storage alongside its existing support for Portable Operating System Interface, or POSIX, file systems. This broadens the benchmark across object-storage and conventional file-system architectures.
MLCommons also released MLPerf Storage v3.0 results. The organization describes the suite as architecture-neutral, representative and reproducible, and said version 3.0 expanded coverage across the full range of AI storage workloads.
The benchmark measures storage-system performance for machine-learning workloads rather than GPU arithmetic performance or model quality. Its new tests therefore focus on how storage platforms handle data-access patterns associated with inference caches and vector databases.
- MLCommons
- MLPerf Storage
- S3 object-storage
- POSIX file-system
Google Cloud added a violation analyzer and dashboard to Virtual Private Cloud Service Controls for organization-wide perimeter monitoring. #
Google Cloud added a violation analyzer and violation dashboard to Virtual Private Cloud Service Controls, or VPC Service Controls. VPC Service Controls establish service perimeters intended to reduce risks involving data exfiltration, compromised accounts and insider threats.
The violation analyzer uses the troubleshooting token or unique identifier from a failed VPC Service Controls request. It generates a report identifying the principals and target resources involved and explains how the request violated the applicable service perimeter, without requiring administrators to construct a Cloud Logging Structured Query Language query.
The violation dashboard aggregates service-perimeter violations across an entire Google Cloud organization into one console view. Administrators can identify trends and spikes in access denials and filter violations by perimeter, project or identity.
BlackLine, a financial-operations-management company, uses the capabilities to manage access levels and ingress and egress policies around its Google Cloud environment. BlackLine said the reports provide access context and support faster resolution of service-perimeter issues.
Google’s TabFM entered BigQuery preview, providing zero-shot tabular regression and classification through SQL across millions of inference rows. #
Google introduced TabFM in BigQuery as a preview service. Developed by Google Research, TabFM is a pretrained foundation model for regression and classification using tabular data.
TabFM uses in-context learning and can perform zero-shot predictions without a separate model-training, tuning or artifact-deployment process. Users provide labeled historical data and new prediction tables through a Structured Query Language statement.
The model is available through the built-in `AI.PREDICT` and `AI.EVALUATE` SQL functions. It automatically determines whether a task is classification or regression from the target label’s data type and handles preprocessing tasks including missing values and categorical encoding.
Google says BigQuery’s distributed inference architecture can process inference tables containing up to millions of rows in minutes. Google also reported that TabFM outperformed classic machine-learning systems and other tabular foundation models in its TabArena classification and regression evaluations.
- Google’s TabFM
- BigQuery
- SQL
Anthropic’s Fable 5.1 keeps enterprise token pricing unchanged while cutting charges for reusing previously processed information by 75%. #
Anthropic announced Fable 5.1 with the same enterprise token prices as Fable 5. The company reduced the charge for recalling or reusing previously processed information by 75%.
Anthropic said reused information can represent more than half of token consumption in long-context and complex tasks. The company therefore described the reduced reuse charge as a way to lower total operating costs while leaving the model’s standard enterprise token price unchanged.
Anthropic reported improvements in long-running programming tasks, including software-project development, code review and work spanning an application’s codebase. It also cited stronger scientific capabilities in experimental design, simulation of results and interpretation of complex charts and tables.
The company said Fable 5.1 has updated classifiers designed to identify risky instructions more accurately and reduce false positives. Anthropic also plans to let enterprise customers retain data for its strongest models in their own cloud infrastructure while continuing to apply safety checks.
Forthcoming OpenAI Astra can autonomously identify and develop zero-day exploits; advanced cybersecurity access initially remains limited to testers. #
OpenAI said its forthcoming Astra model reached a critical cybersecurity-capability threshold under the company’s Preparedness Framework. The model can identify zero-day vulnerabilities and develop exploit programs without human intervention.
OpenAI plans to release Astra soon but did not provide a specific date. Its most advanced cybersecurity functions will initially be available only to a limited group of testers, rather than to all users at launch.
The company subsequently plans to expand defensive access through its Daybreak Blue program. That program will allow vetted testers to use OpenAI’s strongest models for cybersecurity work under accompanying safeguards.
OpenAI said it delayed parts of Astra’s development and release process during recent weeks while strengthening and testing safeguards. The company added monitoring for unauthorized behavior, automatic termination of relevant activity and additional layered protections intended to prevent Astra from taking actions inconsistent with expected behavior.
