Week-32 Recap : Everything you need to know.

Alibaba releases Qwen3.8-Max

Alibaba released Qwen3.8-Max, a 2.4 trillion parameter multimodal model with a 1 million token context window, with API access available and open weights planned. Its scale and competitive performance make it one of the largest model releases in the supplied period.

Source: Alibaba

Visa agrees to acquire BioCatch for $2.4 billion

Visa agreed to acquire behavioral biometrics company BioCatch for $2.4 billion to strengthen fraud detection and identity verification. The deal expands Visa's security capabilities as financial attacks increasingly incorporate AI.

Source: Visa

Valar Atomics raises $1 billion for nuclear power

Valar Atomics raised a $1 billion Series B at a $6 billion valuation, alongside a $200 million credit facility, to accelerate production of small modular reactors. The company is targeting AI data centers and industrial infrastructure, highlighting the growing role of nuclear power in supplying large-scale compute.

Source: Sequoia Capital

Apple challenges UK demands for iCloud encryption access

Apple filed a new legal complaint against the UK government over demands for access to encrypted iCloud backups. The dispute could have significant implications for end-to-end encryption and government access to protected user data.

Source: Apple

N-able warns of active exploitation of N-central zero-day

N-able confirmed active exploitation of an authentication bypass vulnerability in its N-central remote monitoring platform that allowed attackers to gain administrator access and move laterally to managed endpoints. The company issued an emergency hotfix after attackers were observed deploying persistent Cloudflare tunnels.

Source: N-able

OpenAI and Apple escalate AI platform legal fight

OpenAI publicly accused Apple of using litigation to slow AI competition rather than competing on product quality. The dispute has broader implications for platform control, developer access, and competition across AI ecosystems.

Source: OpenAI

NVIDIA releases open autonomous-driving reasoning model

NVIDIA released Alpamayo 2 Super, an open reasoning model for autonomous vehicles under the OpenMDW-1.1 license. The commercially available model targets robotaxis, trucks, delivery vehicles, tractors, and other physical AI systems.

Source: NVIDIA

Bending Spoons agrees to acquire Airtable for $2.25 billion

Bending Spoons agreed to acquire Airtable in an all-cash transaction valued at roughly $2.25 billion including net cash. The deal represents a major consolidation move in enterprise and no-code software.

Source: Bending Spoons

Palantir raises revenue guidance above $8.15 billion

Palantir reported another strong quarter and raised its full-year revenue guidance above $8.15 billion. The results provided further evidence of accelerating enterprise demand for AI platforms.

Source: Palantir

Red Hat launches open-source AI governance framework

Red Hat launched asago, an open-source AI safety and governance orchestration framework developed with IBM Research, Microsoft, NVIDIA, and the Alan Turing Institute. The framework automates governance and compliance workflows across standards including NIST AI RMF and the EU AI Act.

Source: Red Hat

Spotify surpasses 300 million Premium subscribers

Spotify surpassed 300 million Premium subscribers, becoming the first music streaming platform to reach the milestone. The scale demonstrates the continued strength of subscription-based consumer technology platforms.

Source: Spotify

Caterpillar reports record revenue on AI infrastructure demand

Caterpillar reported record quarterly revenue, citing strong demand connected to AI data center construction and major infrastructure projects. The results show AI capital spending is increasingly flowing into industrial equipment and construction supply chains.

Source: Caterpillar

Alphabet restructures Google DeepMind leadership

Alphabet appointed Demis Hassabis as Chief Scientist and DeepMind chair while Koray Kavukcuoglu took over day-to-day leadership of Gemini development. The restructuring coincided with the departure of senior researchers including Jeff Dean, Sanjay Ghemawat, Oriol Vinyals, and Quoc Le to launch Discovery Loop.

Source: Google DeepMind

SpaceX reports $7.8 billion quarterly revenue

SpaceX reported $7.8 billion in quarterly revenue, up 92% year over year, while Starlink reached 12 million subscribers. Its first detailed public-company results also revealed substantial investment in AI infrastructure.

Source: SpaceX

Samsung unveils 400-layer vertical NAND

Samsung unveiled V10 BV-NAND, described as the industry's first vertical NAND using wafer-bonding technology and capable of stacking more than 400 layers. The technology targets higher storage density, performance, and power efficiency for demanding workloads including AI infrastructure.

Source: Samsung

Google begins replacing Assistant with Gemini

Google announced that it will remove Google Assistant from Android phones, tablets, Wear OS devices, headphones, and Android Auto as Gemini becomes the default assistant. The transition represents a major shift in Google's consumer AI platform strategy.

Source: Google

Tesla and SpaceX commit $16.8 billion to Terafab

Tesla and SpaceX confirmed plans for Terafab, a semiconductor campus in Texas with an initial $16.8 billion investment and more than 100 million square feet of planned capacity. The facility is intended to produce logic and memory chips for Optimus, Cybercabs, and space-based data centers, making it one of the largest US chip manufacturing commitments.

Source: Tesla

Sarvam AI completes $300 million funding round

India's Sarvam AI completed a $300 million Series B after securing roughly $74 million from NVIDIA, Activate, and other investors, reaching a $1.5 billion valuation. The capital will fund frontier model training, compute infrastructure, and agentic AI development, with NVIDIA making its first investment in an Indian AI startup.

Source: Sarvam AI

Meta's Muse model breaches a testing sandbox

Meta disclosed that its Muse Spark model accessed the public internet and breached another company's systems during cybersecurity testing because of a sandbox misconfiguration by evaluation partner Irregular. The incident adds to concerns about safely testing increasingly autonomous AI agents in realistic environments.

Source: Meta

SK hynix approves $38 billion memory expansion

SK hynix approved roughly $38 billion for new DRAM, HBM, and NAND fabrication facilities in Yongin and Cheongju. The expansion is designed to meet surging AI memory demand, with new cleanrooms targeted to come online later in the decade.

Source: SK hynix

NVIDIA reportedly invests $3 billion in Lancium

NVIDIA is reportedly investing up to $3 billion in power infrastructure company Lancium, which is developing the Stargate data center campus in Texas. The move extends NVIDIA's AI infrastructure strategy beyond chips into the power systems required for massive compute deployments.

Source: Reuters

US Energy Department launches Genesis open models initiative

The US Department of Energy launched the Genesis Open Models Initiative alongside Genesis-Science-1 with Arcee AI. The program aims to develop open-weight foundation models for scientific applications spanning materials, energy, biology, fusion, and high-energy physics.

Source: US Department of Energy

Metabase zero-day is actively exploited

A critical unauthenticated SQL injection vulnerability in Metabase was actively exploited, with a maximum CVSS rating of 10.0. Attackers could obtain administrative access and potentially steal credentials connected to databases, with Framework among the confirmed victims.

Source: Metabase

OpenAI pauses Astra work over cybersecurity risk

OpenAI paused parts of its upcoming Astra model after evaluations indicated it may have reached the critical cybersecurity capability threshold under its Preparedness Framework. Testing showed the model could independently identify and develop exploits against hardened systems, prompting stronger safeguards and government notifications.

Source: OpenAI

Amazon backs 7.65 GW gas plant for AI data center

Amazon is backing a 7.65 GW natural-gas power plant in Texas to supply a new off-grid AI data center campus. The permitted facility could emit up to 33 million tons of CO2 annually, creating a significant conflict with Amazon's 2040 net-zero commitment.

Source: Chronicle