Amazon reported its fastest AWS growth in nearly five years while heavy AI investment pushed trailing free cash flow to a $42.4 billion loss.
Mistral has publicly released an upgraded Large 4 model that organizations can download and run on their own systems for cybersecurity tasks.
The US Justice Department is asking how Nvidia’s proposed $12.93 billion purchase of Hugging Face could affect competition in AI, Reuters reported.
Google introduced Gemini 4 Carbon for long-running coding and cybersecurity tasks, initially limiting access to selected defenders while it tests safeguards.
Oracle expanded its Nvidia server plans to multi-gigawatt scale while adding Google TPUs and inference chips from Cerebras and SambaNova.
Amazon, Microsoft and Alphabet are expected to report faster cloud growth as investors demand evidence that heavy AI spending can produce cash returns.
Applied Materials plans to raise its holding in BE Semiconductor Industries to about 20% from 9%, deepening cooperation on advanced packaging for AI chips.
Cadence launched Maestro, a portfolio of AI models built to help engineers design chips and run either in the cloud or entirely inside a customer’s data center.
Anthropic said Claude bypassed a paywall and reached public aggregate Georgia voter-count data without human authorization.
Israel’s central bank rejected Prime Minister Benjamin Netanyahu’s suggestion that cryptocurrency could help shield the economy from possible sanctions.
The PHP Foundation says it has one full-time employee and 14 mostly part-time contractors despite supporting software used by much of the web.
OpenAI has resumed training its most advanced models after tightening isolation controls and removing live internet access from the training environment.
Chinese AI developer MiniMax has completed technology-export registration covering open-source releases and overseas licensing of some models.
SK Hynix has put a proposed US IPO for NAND and SSD unit Solidigm on hold amid market volatility and shareholder-protection concerns.
A research tool called X-Hinges helps designers 3D-print objects with embedded sensors and conductive paths.
The memory makers are refitting Chinese factories for NAND with more than 200 layers while navigating US equipment controls.
Nvidia used its Berlin conference to promote container-sized compute systems and software safeguards for Europe’s AI developers.
The SpaceX–Tesla manufacturing venture is aiming for an unusually fast ramp at a planned Texas chip complex.
The French cloud provider says it has substantial unused connected capacity and is preparing a mixture-of-experts model.
The chip company says its Series E financing values it at $6.2 billion after the investment.
Reuters, citing Seoul Economic Daily, reports a planned change for Nvidia’s custom high-bandwidth memory.
The Telegraph reports that Microsoft terminated an agreement for the Essex project.
The change follows concerns that the wording could permit broad reuse of website content for AI.
A company research report links several long-debated NIST curve parameters to a conventional random-number generator and timestamps.
The mapping company says its restructuring affects approximately 200 roles worldwide.
LSEG says its Nvidia collaboration can shorten benchmark analysis of years of Tick History data from hours to seconds.
The company announced its first supercomputer deployment outside France, using AMD Instinct MI355X GPUs.
Google’s frontier model is now available through its API and several subscription products.
Microsoft unveiled its third-generation Maia 300 AI accelerator and said it could deploy more than one million of the custom chips in 2027.
Analysts expect TSMC to report another record quarterly profit as demand for advanced AI processors and chip-packaging capacity remains strong.
Nvidia is in early talks that could lead to another investment in Reflection AI, a targeted hiring and licensing deal, or an acquisition, according to a Financial Times report cited by Reuters.
AI founders and investors speaking during San Francisco Tech Week backed a US approach centered on company safety programs and voluntary commitments rather than broad new rules.
The six-month-old startup is raising hundreds of millions of dollars for a data-center CPU, according to people familiar with the still-open round.
The five-year agreement will expand silicon-interposer manufacturing in Malta, New York, with volume production expected to ramp in the first half of 2028.
Token Fabric combines networking chips, Nvidia-based equipment and diagnostics intended to keep different AI processors supplied with data.
Commissioners will vote on October 29 on advancing a 25 MHz auction and modernising rules for direct links between ordinary devices and satellites.
A local coding model, execution containers and a high-end Nvidia-powered Surface signal a shift toward hybrid AI that can work on devices as well as in the cloud.
The state-backed foundry is building a partner network around 2-nanometre production, but commercial success still hinges on customers and manufacturing yield.
The SAP-backed company says customers can initiate fiat or stablecoin payments within existing business software and contracts.
The operator expects automation to lower indirect costs while also creating a new AI-services revenue stream.
OpenAI's chief executive said AI should remain widely available even though misuse, scams and other harms cannot be eliminated entirely.
Jacob Coxon's reported appearance would bring a former frontier-lab insider into a municipal discussion of AI risks and oversight.
The panel has been asked to report within 120 days on AI risks, opportunities and the federal response to breaches, hacks and other incidents.
President Donald Trump defended the projects at an Ohio rally as bipartisan voter concern grows about electricity costs, land use and water demand.
The French supercomputer maker says it can source about 70% of system components in Europe, but still has no European memory supplier.
The startup says optical connections built from established VCSEL technology could let far more memory chips communicate with each AI accelerator.
The U.S. Federal Trade Commission is examining potential consumer harms from agentic AI and plans information demands involving Anthropic, OpenAI and the research group METR.
DeepSeek said it is open-sourcing compute and communication libraries optimized for Huawei's Ascend processors as Chinese companies seek alternatives to Nvidia's software ecosystem.
OpenAI unveiled always-on agents called dots that can pursue user goals across workplace applications, extending the company's push from chatbots into autonomous enterprise software.
Chirantan “CJ” Desai will report to Mark Zuckerberg as Meta builds a business unit intended to package its AI models, agents and tools for corporate customers.
Donald Trump said slowing AI would weaken the US against China, while Bill Gates joined calls for safeguards that go beyond industry self-regulation.
Boeing told operators that a cockpit-software issue can remove automated flight guidance in a specific landing scenario, while engineers work on a permanent update.
The FAA said troubleshooting errors triggered a switch to a backup fibre line that was then accidentally cut, disrupting flights at major northeastern airports.
Fraudsters used AI-assisted impersonation to steal €95 million from Fideuram, Intesa Sanpaolo’s private-banking arm, according to Reuters sources.
Microsoft unveiled a natural-language coding tool, an always-on enterprise agent and tighter Office integration for its Copilot app, Reuters reported.
Apple extended its patent-licensing agreement with Qualcomm, maintaining access to parts of the chip designer’s intellectual-property portfolio as Apple reduces reliance on Qualcomm modems.
BlackBerry raised its annual revenue forecast after record quarterly revenue at its QNX software division, driven by automotive growth and new design wins.
New Zealand's National Cyber Security Centre says suspected state-backed activity put sensitive information at risk across government, health, education and IT services.
Unit 42 Continuous Frontier AI Defense will use models from Anthropic, OpenAI and open-weight developers to test applications, APIs and cloud systems.
Alibaba outlined a 5–10-trillion-parameter model, its Zhenwu V900 accelerator and a plan to exceed 20 gigawatts of global data-centre capacity by 2032.
AMD became the fourth US chipmaker to reach the valuation milestone as its shares hit a record and investors returned to AI-linked stocks.
The unusual public clash between two extortion groups appears genuine to outside experts, although Reuters could not verify every claim.
The proposed channel would cover AI incidents that rise to a national-security level, but China’s response and the mechanism’s scope remain unclear.
The Chinese memory-chip maker announced a new platform with finer feature spacing and two 24-gigabit LPDDR5X products.
The U.S. president announced the plan on social media but provided no details about the proposed bodies’ authority or structure.
The former Character.AI chief executive will oversee enterprise technology, engineering and AI platforms from October 2.
Prysmian and Rio Tinto will supply cable made with ELYSIS inert-anode aluminum for a hyperscale facility.
The legal-focused platform combines GPT-6 Astra with indexed U.S. law and integrations for specialist software providers.
The partners will develop Qaptiva software for hybrid quantum-classical workloads and add an Alice & Bob machine in 2027.
Salesforce, Nvidia and AWS are among the partners adding hands-free workplace tools to Snap’s $2,195 AR glasses.
The flagship mobile processor targets premium phones and more on-device generative AI.
The new offering connects the assistant to wealth-management software.
Artifact InSPECtor uses human checks of Euclid data to improve automated cleaning of astronomical observations.
The blueprint recommends staged authorisations, continuous monitoring and public safety records rather than one-time approval.
The two-year retrofit starts in 2027 across 115 trains, adding spring-and-memory-foam cushions and revised backrests without a new taxpayer charge.
NASA's Expedition 75 crew assembled NEMO for cell imaging and serviced hardware used in crystal-growth research and spacesuit communications.
An MIT experiment in 14 healthy volunteers linked 50-millisecond sounds to larger slow waves and cerebrospinal-fluid movement.
ESA published satellite imagery showing how an orbital imaging spectrometer distinguishes features of a volcanic eruption.
Event-triggered recordings will help develop driver-assistance systems, with privacy safeguards described by the company.
OpenAI’s Astra launch puts the supervision of autonomous agents back in focus.
Sam Altman calls for engagement as the capabilities of frontier models grow.
The desktop announcement emphasizes AI workloads and professional computing.
The release lists updates for supported Office products.
July updates cover documents, presentations, video and hands-free meetings.
Its July event highlighted lead generation and campaign support.
The Agentic Control Plane focuses on visibility and intervention.
The services roadmap includes richer navigation and more flexible sharing.
The WWDC announcement outlined upcoming software and safety features.