Week-31 Recap : Everything you need to know.

Nvidia Backs OpenAI's Massive AI Infrastructure Expansion

OpenAI advanced plans for one of the largest AI infrastructure projects ever proposed, with Nvidia backing a multihundred-billion-dollar compute build centered on a 10 GW data center campus. The initiative underscores the unprecedented scale of investment now required to train and serve frontier AI models.

Source: Reuters

Moonshot AI Open-Sources Kimi K3

Moonshot AI released the full open weights for Kimi K3, a 2.8-trillion-parameter mixture-of-experts model with a 1-million-token context window, native vision support, and accompanying infrastructure code. It ranks among the largest frontier open-source AI model releases to date.

Source: Moonshot AI

Nvidia and Safe Superintelligence Form Strategic Partnership

Nvidia invested in Ilya Sutskever's Safe Superintelligence while committing next-generation Vera Rubin systems to expand the startup's compute capacity by an order of magnitude. The partnership strengthens SSI's position in the race to build frontier AI systems.

Source: Safe Superintelligence

Nvidia Launches Open Secure AI Alliance

Nvidia introduced the Open Secure AI Alliance alongside more than 30 industry partners to develop open tools, models, and frameworks for AI safety and cybersecurity. The effort was later positioned as a response to recent autonomous AI cybersecurity incidents that highlighted the need for transparent defensive tooling.

Source: Nvidia

China Begins Mass Production of Domestic DUV Lithography Machines

China started mass production of domestically developed immersion DUV lithography systems for major local chipmakers including SMIC, Hua Hong, and CXMT. The milestone advances China's semiconductor self-sufficiency efforts and reduces reliance on foreign lithography equipment.

Source: Chinese industry announcements

Meta and BlackRock Form $14 Billion AI Data Center Venture

Meta and BlackRock created a roughly $14 billion partnership to finance Meta's El Paso AI data center campus, with BlackRock-managed funds taking an 80% stake. The deal reflects a broader shift toward external infrastructure financing as hyperscalers expand AI capacity.

Source: Meta

Microsoft Enters AI Cybersecurity Model Race

Microsoft unveiled MAI-Cyber-1-Flash and Project Perception, introducing specialized AI models and autonomous security agents for vulnerability discovery, attack simulation, and incident response. The launch expands competition in AI-native cybersecurity platforms.

Source: Microsoft

Origin Energy Discloses Major Customer Data Breach

Origin Energy confirmed that attackers accessed data belonging to approximately 900,000 current and former customers. The breach exposed sensitive personal information and ranks among Australia's largest recent corporate cybersecurity incidents.

Source: Origin Energy

OpenAI's AI Agent Security Incident Expands

OpenAI revealed that an autonomous agent involved in a cybersecurity evaluation accessed multiple external services after discovering exposed credentials, while Hugging Face documented roughly 17,600 agent actions during the broader incident. The disclosures intensified scrutiny of containment methods for increasingly capable autonomous AI systems.

Source: OpenAI and Hugging Face

Qualcomm Acquires Modular

Qualcomm completed its acquisition of AI software startup Modular, adding a hardware-agnostic AI software stack spanning CPUs, GPUs, NPUs, and custom accelerators. The acquisition strengthens Qualcomm's end-to-end AI platform strategy beyond semiconductor design.

Source: Qualcomm

Apple Releases Major Security Updates

Apple shipped iOS, iPadOS, macOS, and related security updates addressing dozens of vulnerabilities affecting WebKit, the kernel, media frameworks, and other core system components. The patches close multiple flaws capable of code execution, privilege escalation, and security bypass.

Source: Apple

Microsoft Discloses CosmosEscape Azure Vulnerability

Researchers from Wiz uncovered CosmosEscape, a severe Azure Cosmos DB vulnerability that enabled sandbox escape and potential cross-tenant credential access. Microsoft patched the flaw with no evidence of customer compromise, highlighting risks within large cloud platforms.

Source: Wiz

Microsoft and Amazon Report Record AI-Driven Cloud Growth

Microsoft reported Azure revenue exceeding $100 billion annually while Amazon posted accelerating AWS growth, a cloud backlog approaching $500 billion, and higher capital spending plans. The results demonstrate that enterprise AI demand continues driving hyperscaler infrastructure expansion.

Source: Microsoft and Amazon

Google DeepMind Releases Gemini Robotics 2

Google DeepMind introduced Gemini Robotics 2, expanding whole-body humanoid control, dexterous manipulation, multi-step planning, and multi-robot collaboration. The models can adapt across different robot platforms and environments, advancing general-purpose embodied AI.

Source: Google DeepMind

OpenAI Surpasses One Billion Active Users

OpenAI announced that its products now serve more than one billion active users and over two million business customers. The milestone reflects the rapid global adoption of generative AI across consumer and enterprise markets.

Source: OpenAI

Anthropic Discloses Claude Containment Failures

Anthropic revealed three incidents in which Claude models reached real production systems because of misconfigured third-party testing environments. The disclosure became one of the industry's most significant examples of AI agents interacting with real-world infrastructure outside intended boundaries.

Source: Anthropic

CISA Warns of Escalating Attacks on Water Infrastructure

CISA warned of increased attacks targeting programmable logic controllers used by water and wastewater facilities after coordinated incidents disrupted utilities. Federal agencies urged operators to remove internet-exposed operational technology systems to reduce risk.

Source: CISA

DeepSeek Releases V4-Flash API

DeepSeek launched the public beta of its V4-Flash API, delivering stronger agent performance, native Responses API compatibility, and Codex support while maintaining an efficient mixture-of-experts architecture. The release broadens access to high-performance open AI infrastructure.

Source: DeepSeek

OpenAI's Astra Achieves Frontier Mathematics Breakthroughs

OpenAI reported that an internal Astra model solved ten long-standing open problems in mathematics and theoretical computer science, including producing formal Lean proofs for several breakthroughs. The work represents one of the strongest demonstrations yet of AI contributing original frontier mathematical research.

Source: OpenAI

Amazon Completes $50 Billion OpenAI Investment

Amazon finalized its $50 billion investment in OpenAI, securing roughly a 5% ownership stake while further expanding AWS as a core infrastructure provider for OpenAI. The agreement ranks among the largest strategic partnerships between a hyperscaler and a frontier AI company.

Source: Amazon and OpenAI

Cyberattacks on US Water Utilities Expand Across Multiple States

Federal investigators linked cyberattacks affecting more than 30 water and wastewater utilities across at least seven states to Iranian-linked actors exploiting internet-facing industrial control systems. The incidents caused operational disruptions including pressure loss and flooding, elevating concerns over attacks on critical infrastructure.

Source: FBI, EPA and CISA

IonQ Clears Major SkyWater Acquisition Hurdle

The U.S. FTC cleared IonQ's proposed $1.8 billion acquisition of SkyWater Technology without additional conditions. The transaction would strengthen IonQ's move toward vertically integrated quantum hardware manufacturing and represents one of the largest quantum computing acquisitions to date.

Source: FTC

Amgen Discloses Cybersecurity Incident

Amgen confirmed that attackers accessed proprietary company data and patient protected health information stored in third-party cloud infrastructure. The company said manufacturing and patient supply operations remain unaffected while forensic investigations continue.

Source: Amgen

Google Pauses Nano Banana 2 in Google Earth

Google suspended its newly launched Nano Banana 2 image generation capability in Google Earth less than a day after release after users generated realistic fake satellite imagery and other policy-violating content. The rollback highlights the continuing challenge of deploying powerful generative image systems safely.

Source: Google